Review

Per- and poly-fluoroalkyl substances: A review of sources, properties, chromatographic detection, and toxicological implications

  • Received: 25 September 2024 Revised: 28 January 2025 Accepted: 26 March 2025 Published: 02 April 2025
  • Emerging per and poly-fluoroalkyl substances (PFASs) are very resistant to degradation and have negative impacts on human and environmental health at very low concentrations. The initial stage in removing PFASs from contaminated locations is their detection and quantification. Particularly utilized in this context are Gas Chromatography-Mass Spectrometry (GC-MS), Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS), and High-Performance Liquid Chromatography (HPLC). In this review, we seek to contribute to our understanding of the state-of-the-art in emerging PFAS by offering a complete analysis of PFAS in environments, taking into consideration their sources, classes, and properties. Polyfluoroalkyl ether substances (PFAES), short-chain PFA, replacement PFA, and fluorotelomer-based substances can bioaccumulate in living species, making their detection even more necessary. We intend to provide researchers with an overview of the current state of research on PFASs in environments, encompassing the toxicological effects and their detection and quantification methods, serving as a guide for current and future studies.

    Citation: Eliasu Issaka, Mabruk Adams, Enock Adjei Agyekum, Josephine Baffoe, Blessing Tornyeava. Per- and poly-fluoroalkyl substances: A review of sources, properties, chromatographic detection, and toxicological implications[J]. AIMS Environmental Science, 2025, 12(2): 321-351. doi: 10.3934/environsci.2025015

    Related Papers:

    [1] Ran Zhou, Yanghan Ou, Xiaoyue Fang, M. Reza Azarpazhooh, Haitao Gan, Zhiwei Ye, J. David Spence, Xiangyang Xu, Aaron Fenster . Ultrasound carotid plaque segmentation via image reconstruction-based self-supervised learning with limited training labels. Mathematical Biosciences and Engineering, 2023, 20(2): 1617-1636. doi: 10.3934/mbe.2023074
    [2] WenJun Zhou, Tianfei Wang, Yuhang He, Shenghua Xie, Anguo Luo, Bo Peng, Lixue Yin . Contrast U-Net driven by sufficient texture extraction for carotid plaque detection. Mathematical Biosciences and Engineering, 2023, 20(9): 15623-15640. doi: 10.3934/mbe.2023697
    [3] Chenkai Chang, Fei Qi, Chang Xu, Yiwei Shen, Qingwu Li . A dual-modal dynamic contour-based method for cervical vascular ultrasound image instance segmentation. Mathematical Biosciences and Engineering, 2024, 21(1): 1038-1057. doi: 10.3934/mbe.2024043
    [4] Keying Du, Liuyang Fang, Jie Chen, Dongdong Chen, Hua Lai . CTFusion: CNN-transformer-based self-supervised learning for infrared and visible image fusion. Mathematical Biosciences and Engineering, 2024, 21(7): 6710-6730. doi: 10.3934/mbe.2024294
    [5] Xiao Zou, Jintao Zhai, Shengyou Qian, Ang Li, Feng Tian, Xiaofei Cao, Runmin Wang . Improved breast ultrasound tumor classification using dual-input CNN with GAP-guided attention loss. Mathematical Biosciences and Engineering, 2023, 20(8): 15244-15264. doi: 10.3934/mbe.2023682
    [6] Jingyao Liu, Qinghe Feng, Yu Miao, Wei He, Weili Shi, Zhengang Jiang . COVID-19 disease identification network based on weakly supervised feature selection. Mathematical Biosciences and Engineering, 2023, 20(5): 9327-9348. doi: 10.3934/mbe.2023409
    [7] Dehua Feng, Xi Chen, Xiaoyu Wang, Xuanqin Mou, Ling Bai, Shu Zhang, Zhiguo Zhou . Predicting effectiveness of anti-VEGF injection through self-supervised learning in OCT images. Mathematical Biosciences and Engineering, 2023, 20(2): 2439-2458. doi: 10.3934/mbe.2023114
    [8] Yanghan Ou, Siqin Sun, Haitao Gan, Ran Zhou, Zhi Yang . An improved self-supervised learning for EEG classification. Mathematical Biosciences and Engineering, 2022, 19(7): 6907-6922. doi: 10.3934/mbe.2022325
    [9] Jin Zhang, Nan Ma, Zhixuan Wu, Cheng Wang, Yongqiang Yao . Intelligent control of self-driving vehicles based on adaptive sampling supervised actor-critic and human driving experience. Mathematical Biosciences and Engineering, 2024, 21(5): 6077-6096. doi: 10.3934/mbe.2024267
    [10] Ansheng Ye, Xiangbing Zhou, Kai Weng, Yu Gong, Fang Miao, Huimin Zhao . Image classification of hyperspectral remote sensing using semi-supervised learning algorithm. Mathematical Biosciences and Engineering, 2023, 20(6): 11502-11527. doi: 10.3934/mbe.2023510
  • Emerging per and poly-fluoroalkyl substances (PFASs) are very resistant to degradation and have negative impacts on human and environmental health at very low concentrations. The initial stage in removing PFASs from contaminated locations is their detection and quantification. Particularly utilized in this context are Gas Chromatography-Mass Spectrometry (GC-MS), Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS), and High-Performance Liquid Chromatography (HPLC). In this review, we seek to contribute to our understanding of the state-of-the-art in emerging PFAS by offering a complete analysis of PFAS in environments, taking into consideration their sources, classes, and properties. Polyfluoroalkyl ether substances (PFAES), short-chain PFA, replacement PFA, and fluorotelomer-based substances can bioaccumulate in living species, making their detection even more necessary. We intend to provide researchers with an overview of the current state of research on PFASs in environments, encompassing the toxicological effects and their detection and quantification methods, serving as a guide for current and future studies.



    Carotid atherosclerotic plaque rupture leading to the formation of a thrombus is one of the most significant causes of strokes [1,2]. For individuals with a high risk of having a stroke, early detection and timely classification of carotid atherosclerotic plaques are clinically significant. Attributable to the convenience, efficiency, and affordability of ultrasound, carotid ultrasound examination has been widely used in carotid plaque classification [3]. In clinical practice, we primarily rely on doctors to manually identify carotid plaques based on neck images, which is time-consuming and labor-intensive. We need to make efforts to detect and classify plaques early to assess the risk level of ischemic stroke in patients and take appropriate preventive and treatment measures. Recently, artificial intelligence-aided carotid plaque diagnosis has shown great potential in helping to reduce the workload of doctors and alleviate the burden of medical care.

    Over the past few years, deep learning (DL) algorithms have proven effective in carotid plaque ultrasound image classification. These algorithms enhance the classification capability by constructing multiple nonlinear neural network layers to discover the complex structures in high-dimensional data. Lekadir et al. [4] proposed a convolutional neural network (CNN) that can automatically extract the optimal information for identifying different plaque constituents from images. Zhan et al. [5] designed the clustered PCA network based on the PCAnet with the principal component analysis (PCA) vector as the convolution kernel to extract the features of patches effectively. Ma et al. [6,7] employed a deep residual network for carotid plaque classification and the spatial pyramid pooling (SPP) was redesigned, and the multilevel strip pooling (MSP) was proposed for carotid plaque longitudinal echo classification. Zreik et al. [8] proposed a multitask recurrent convolutional neural network applied to coronary artery multiplanar reformatted (MPR) images to perform an automatic analysis. After a detailed analysis of the DL-based segmentation technique in the carotid artery ultrasound images [9], Huang et al. proposed a novel boundary-delineation network to extract the vascular wall in carotid ultrasound [10] and a nested attention-guided deep learning model (named NAG-Net) for accurate segmentation of carotid lumen-intima interface and media-adventitia interface [11]. Cai et al. [12] designed a machine-learning approach based on internal carotid artery blood flow to predict cerebral perfusion status. Although the above algorithms have achieved excellent performance, their performance heavily relies on the availability of a substantial amount of labeled images. However, it is challenging to acquire numerous labeled carotid plaque ultrasound images in clinical practice, and manual annotation costs are very high. Therefore, the shortage of labeled image quantity is a limitation for deep learning algorithms in the clinical carotid plaque classification.

    To ease the issue of limited labeled image quantity, self-supervised learning (SSL) methods provide a new solution and show good performance [13]. For instance, Bai et al. [14] presented an SSL approach for cardiac magnetic resonance (MR) image segmentation based on the prediction of anatomical positions. Koohbanani et al. [15] demonstrated the SSL method called Self-Path that effectively leverages contextual, multi-resolution, and semantic features to enhance region adaptation for histopathological image patch classification. Abbet et al. [16] introduced an SSL method that integrates clustering metrics with the representation of organizational regions to acquire potential histopathological patterns. Hervella et al. [17] presented the SSL approach for multimodal reconstruction tasks and conducted a series of experiments using multimodal retinal and fluorescein angiography that provided complementary fundus information. In order to better exploit unlabeled images, Chen et al. [18] proposed a strategy of incorporating context restoration as a pretext task and validated its effectiveness in medical imaging tasks. The above works demonstrate that SSL methods can effectively improve the performance of CNNs, but the key lies in how to extract more features through pretext tasks.

    Most existing works are mainly focused on how to construct pretext tasks and they don't consider using multiple pretext tasks to pretrain the network. However, different pretext tasks can make the network acquire richer visual feature presentations from different aspects. The model should be more accurate and robust if these different features can be fused. In this study, we propose a self-supervised fusion network for carotid plaque ultrasound image classification named Fusion-SSL. The method utilizes two SSL tasks (classifying image block order [19] and predicting image rotation angle [20]) for the network pretraining. A dual-branch residual neural network is developed to fuse feature presentations learned by the two tasks, which can extract richer visual boundary shape and contour information than a single task. The Fusion-SSL method further enhances the classification performance with limited labeled training images. Investigation on carotid ultrasound images shows that the developed Fusion-SSL algorithm provides high accuracy and agreement in classification results, which is superior to the performance of a single SSL network. This suggests that Fusion-SSL may be suitable for clinical use.

    In this study, ultrasound experts with decades of vascular imaging experience acquired 1270 carotid plaque ultrasound images from 844 patients at Zhongnan Hospital (Wuhan, China). Carotid ultrasound images were collected by an Acuson SC2000 (Siemens, Erlangen, Germany) ultrasound system with a 5–12 MHz linear array probe (9L4). The study obtained approval from the Institutional Review Board (IRB) of Zhongnan Hospital (Wuhan, China) and written informed consent from all patients.

    The appearance of carotid ultrasound images depends on the image acquisition and is affected by the device, operator, and patient. Moreover, the region of interest (ROI) of carotid plaques in ultrasound images varies in size, which does not satisfy the requirement of a uniform-sized input for CNN. Direct cropping and scaling of ultrasound images may lead to information loss, image distortion, scale inconsistency, and image artifacts, which may affect subsequent analysis and applications. To address the issue of changed tissue appearance and improve the ability of extracting detailed and overall information, the following image preprocessing steps are used in this study to meet the input requirements of CNN: (i) Obtain the ROI image for each plaque by the segmented patch boundary, and set the pixels outside the boundary to zero to get the preprocessed plaque image. (ii) Pad the preprocessed patch image with zeros and generate a modified square plaque image based on the longer side of the original plaque rectangle. (iii) Normalize the size of the square plaque image obtained in step (ii) to a fixed size of 224 × 224. We used a linear scaling operation between the minimum and maximum values of the images as a standard normalization method to enhance the comparability and reliability. The normalization formula is shown in Eq (2.1). The original and processed images are shown in Figure 1.

    y=xxminxmaxxmin (2.1)
    Figure 1.  Three types of carotid plaque. (a–c) represent the original ultrasound images of hyperechoic plaque, hypoechoic plaque, and mixed-echoic plaque; (d–f) represent the three preprocessed plaque images; (g–i) represent the three square plaque images.

    where x is the pixel value of the carotid plaque ultrasound image, xmin and xmax are, respectively, the minimum and maximum pixel values of that carotid plaque ultrasound image, and y is the normalized carotid plaque ultrasound image.

    The real-label data in this study was produced following the European Carotid Plaque Study Group standard, which classified carotid plaque echogenicity into three categories: hyperechoic plaque, hypoechoic plaque, and mixed-echoic plaque. Hypoechoic plaque and mixed-echoic plaque are prone to rupture [21]. This classification was executed by a clinical expert (co-author F. Wang) with over a decade of experience in using carotid ultrasound images to assess atherosclerosis. Initially, he classified 1270 plaques into three types according to echogenicity and reclassified them after three months. He calculated the kappa value (k = 0.747), which indicates a high degree of consistency between the two classifications. Finally, all carotid plaque ultrasound images were classified into three types: hyperechoic plaque, hypoechoic plaque, and mixed-echoic plaque. The resolution of these carotid plaque ultrasound images was 1024 × 768 with a pixel size of 0.13 mm.

    Although DL has demonstrated effectiveness in carotid plaque ultrasound image classification, its capability heavily depends on numerous labeled datasets. SSL has partially alleviated this issue, but extracting rich features through a single SSL pretext task is difficult. Therefore, we fused two pretext tasks to pretrain the network, then, a dual-branch residual neural network was constructed to extract feature maps from the two tasks and fuse them. We aim to obtain richer visual features of carotid plaques by learning from boundary shapes and contour information. This approach not only addresses the issue of insufficient labeled images, but also further enhances the capacity of carotid plaque image classification.

    However, not all feature fusion is helpful in improving the performance of the model. Some features extracted by different SSL methods will produce redundancy and inconsistent problems. After comparing, we chose two reliable SSL methods, classifying image block order and predicting image rotation angle, which can provide diverse feature learning, complementary feature extraction, and enhanced robustness, thus improving the performance and reliability of the model. First, they both enrich the diversity of features. Classifying image block order randomly introduces different contextual information and predicting image rotation angle introduces different perspectives and transformations. Second, the features extracted by these two methods are complementary. Classifying image block order can emphasize the long-distance dependence in the image and investigate global information, helping the model to learn the semantic information of the image. Predicting image rotation angle can capture the local orientation information in the image and retain the information of the original image, helping the model to understand the geometry and structure in the image. Therefore, we propose the idea of combining these two SSL methods to leverage their strengths and compensate for their respective limitations. By simultaneously learning the tasks of classifying image block order and predicting image rotation angles, our network can extract richer image features from various auto-generated targets and better capture the semantic information and structure of the images. This fusion approach is expected to enhance the performance of SSL in image classification tasks and decrease the requirement for labeled images, thereby reducing the cost and time consumption of data labeling.

    The Fusion-SSL method proposed in this study mainly consists of two stages. (1) Pre-training of self-supervised models: Model A and model B are obtained by training two different SSL methods with the same CNN, and then both model A and model B are migrated to the feature extraction layer in the dual-branch residual network for parameter initialization. (2) Dual-branch residual network feature fusion in plaque classification: Build a dual-branch residual network (DBResNet) to fuse the two self-supervised network models obtained in step (1). The DBResNet contains three parts: a feature extraction layer, a feature fusion layer, and a fully connected layer. 1) Transfer models A and B to the two parallel branch networks of the feature extraction layer in DBResNet (both the first and second branches are composed of ResNet-101, but the ResNet-101 architecture is only kept up to the conv5_x residual structure and the subsequent average pooling layer and fully connected layer are removed) to initialize the parameters for each branch and obtain two features (FA and FB) after training; 2) FA and FB are fused through the feature fusion layer to obtain the fused feature ˆF; 3) Input the fused feature ˆF to the fully connected layer to obtain the final classification result. The overall flowchart of Fusion-SSL is shown in Figure 2.

    Figure 2.  The overall flowchart of Fusion-SSL.

    To learn richer visual features about carotid plaques, the proposed Fusion-SSL method in this study combines two different SSL methods for feature learning.

    (1) The first SSL method used in this study is the "classifying image block order" method. Its process is shown in Figure 3. First, the original carotid plaque ultrasound dataset X is reorganized into a new carotid plaque ultrasound dataset R by the self-supervised "classifying image block order" method, and corresponding pseudo-labels are generated. The label set of the original image dataset X is set as 1, while the label set of the disordered image dataset R is set as 0 (the real labels of the dataset are not used here). Second, the X dataset and R dataset are preprocessed, and then the two datasets after the preprocessing operation are combined to call the expanded X dataset. The label set of the X dataset is Y and its value is shown in Eq (2.2). The classification categories for this dataset are binary, which include the original images and the disordered images. Finally, the images of the dataset X with their corresponding label set Y are convolved and pooled to serve as the input for ResNet-101, and the optimal network model A is obtained after multiple training iterations.

    yi={1,(XiX)0,(XiR) (2.2)
    Figure 3.  Flowchart of feature extraction for classifying image block order method.

    where yi represents the value of Y label set and Xi represents the i-th image in X dataset.

    (2) The second SSL method used in this study is the "predicting image rotation angles" method, and the process is shown in Figure 4. First, the original carotid plaque ultrasound dataset X is transformed into rotational angles of 0, 90,180, and 270 degrees by the self-supervised "predicting image rotation angle" method. At the same time, data preprocessing operations are performed on the new carotid plaque ultrasound dataset R after angular transformation. Carotid plaque ultrasound datasets with four different rotation angle transformations R1, R2, R3, and R4 are obtained as shown in Figure 4. The label set of the dataset R1 is set to 0, the label set of the dataset R2 is set to 1, the label set of the dataset R3 is set to 2, and the label set of the dataset R4 is set to 3. R1, R2, R3, and R4 were then merged into a new dataset R with four categories, namely, four different angles. The label set of the R dataset is the Y and its value is shown in Eq (2.3). Next, the images in the dataset R with their corresponding label set Y are convolved and pooled to serve as the input to ResNet-101, and the final network model B is obtained after multiple training iterations.

    yi={0,(RiR1)1,(RiR2)2,(RiR3)3,(RiR4) (2.3)
    Figure 4.  Flowchart of feature extraction for predicting image rotation angle method.

    where yi represents the value of Y label set and Ri represents the i-th image in R dataset.

    The overall framework of DBResNet consists of three parts: the feature extraction layer, the feature fusion layer, and the fully connected layer.

    (1) The feature extraction layer of DBResNet has two parallel branches, each of which consists of the network structure after deleting the average pooling layer and the fully connected layer in ResNet-101, and its structure is shown in Figure 5. The first layer is a 7 × 7 convolutional layer with a stride of 2 and 64 convolutional filters. The second layer is a 3 × 3 max pooling layer with a stride of 2. After passing through the first two layers, a 56 × 56, 64-channel output is obtained. The subsequent layers are well-known residual modules, including 3 conv2 blocks, 4 conv3 blocks, 23 conv4 blocks, and 3 conv5 blocks. Due to their complexity and extensive utilization, they will not be elaborated here. Finally, two 7 × 7, 2048-channel feature maps (FA and FB) are obtained after multiple convolution operations.

    Figure 5.  Flowchart of DBResNet.

    (2) The selective kernel network (SKNet) [22] presented by Li et al. adopts a nonlinear method to fuse features from different kernels, achieving adjustment of receptive fields with different sizes. It contains three operations: The split operation generates several channels with varying sizes of kernel according to the different receptive field sizes of the neurons; the fuse operation integrates information from several channels to acquire a global and comprehensible representation for weight selection; and the select operation fuses the feature maps of different kernel sizes based on the weights obtained from the selection. Therefore, the Fusion-SSL method proposed in this study uses the SK module as a feature fusion layer of DBResNet to learn richer visual features. 1) As shown in Figure 5, the feature fusion layer of DBResNet differs from the SK module in the following locations: The network layer removes the Split operation in the SK module and combines the feature maps FA and FB after the feature extraction layer as inputs, and the feature FA and FB, respectively, carry information from two SSL methods. 2) The fusion result is fused from the two branches by elemental summation. First, the two feature maps are added to obtain F (F=FA+FB). Then, inspired by channel-attention, to achieve the weighted attention of different channel features and suppress redundant features, making the model pay more attention to features that are more important to the current task, the feature map is embedded with global information by applying a simple global average pooling operation (Fgp) to generate channel statistics s as shown in Eq (2.4), where C is the feature dimension of the s in Figure 5. The feature z is then obtained through a fully connected layer (Ffc). 3) Calculate the weight ac and bc of each layer of different sensory field information using the softmax function according to the direction of the channel, where b is a redundant matrix, and in the case of two branches, b=1a, as shown in Eq (2.5). Finally, the elements of ac and bc are multiplied and summed with the corresponding features FA and FB to obtain the final output ˆF, as shown in Eq (2.6).

    sc=Fgp(Fc)=1H×WHi=1Wj=1Fc(i,j) (2.4)
    aC=eAczeAcz+eBcz,bC=eBczeAcz+eBcz (2.5)
    ˆF=acFA+bcFB,aC+bC=1 (2.6)

    (3) The fully connected layer comprises a global average pooling layer and a fully connected layer to predict the carotid plaque class (hyperechoic plaque, hypoechoic plaque, and mixed-echoic plaque).

    The hardware environment of this study is NVIDIA GeForce GT 710 graphics card and 24G memory. The software environment is the Windows 10 operating system, PyTorch 2.0.1 DL framework, Python 3.8.8, CUDA 11.7, and PyCharm 2023 community version.

    For the pretraining part of the Fusion-SSL method, the details of the dataset division are as follows:

    (1) The total number of carotid plaque ultrasound image dataset X expanded by the "classifying image block order" pretext task is 2540. Next, the expanded carotid plaque ultrasound image dataset X was randomly divided into a training set and a validation set in a ratio of 0.8:0.2. The number of images for the divided training sets and the validation sets were 2032 and 508.

    (2) The total number of carotid plaque ultrasound image datasets R expanded by the "predicting image rotation angle" pretext task is 5080. Next, the expanded carotid plaque ultrasound image dataset R was randomly divided into a training set and a validation set in a ratio of 0.8:0.2. The number of images for the divided training sets and the validation sets were 4064 and 1016.

    For the DBResNet part of the Fusion-SSL method, the dataset is divided as follows:

    The dataset used in this part is the original carotid ultrasound dataset X, which has a total number of 1270. The dataset was then randomly divided into training set, validation set, and test set based on the number of patients in the ratio of 0.6:0.2:0.2. The number of images in the partitioned training set, validation set, and test set were 764,260 and 246.

    The network training parameters for this experiment were epoch, batch size, and learning rate. The selection of these parameters is based on the datasets, task, model complexity, computational resources, and other conditions.

    (1) Adjustment of epoch: Increasing the number of epochs increases the time of model training, but improves the model accuracy. However, excessive epochs may lead to overfitting of the model. Therefore, the most appropriate epoch can be selected by gradually increasing the number of epochs and observing the model performance. In our experiment, the "classifying image block order" pretext task was trained for 25 epochs, while the "predicting image rotation angle" pretext task was trained for 150 epochs. The remaining experiments were all trained for 50 epochs.

    (2) Adjustment of batch size: A larger batch size can accelerate training speed, especially on the graphics processing unit (GPU), but may reduce accuracy. A smaller batch size may lead to a more stable convergent behavior, reducing the probability of overfitting. Therefore, multiple attempts on the model are needed to find the most appropriate batch size. As shown in Table 1, we finally chose 8 as the best batch size.

    Table 1.  Performance evaluation of the Fusion-SSL method on 100% labeled training images for different batch size.
    batch size Accuracy
    (%)
    Precision
    (%)
    Sensitivity
    (%)
    Specificity
    (%)
    G-Mean
    (%)
    F1 score
    (%)
    Kappa
    4 86.0 ± 2.3 85.7 ± 1.9 85.8 ± 2.1 92.7 ± 1.1 89.1 ± 1.6 85.5 ± 2.1 0.78 ± 0.03
    8 87.6 ± 1.0 87.4 ± 1.2 87.2 ± 1.3 93.4 ± 0.7 90.2 ± 1.0 87.2 ± 1.1 0.80 ± 0.02
    16 87.2 ± 2.0 87.0 ± 1.7 86.9 ± 2.0 93.1 ± 1.3 89.8 ± 1.8 86.9 ± 2.0 0.79 ± 0.03

     | Show Table
    DownLoad: CSV

    (3) Adjustment of learning rate: A smaller learning rate can make the model more stable but may require more iterations to converge. The large learning rate can accelerate the weight update speed but will cause the model to jump in the iteration and miss the optimal solution. Therefore, multiple experiments need to be performed to find a suitable learning rate. As shown in Table 2, we finally chose 0.0001 as the best learning rate.

    Table 2.  Performance evaluation of the Fusion-SSL method on 100% labeled training images for different learning rate.
    learning rate Accuracy
    (%)
    Precision
    (%)
    Sensitivity
    (%)
    Specificity
    (%)
    G-Mean
    (%)
    F1 score
    (%)
    Kappa
    0.01 86.2 ± 1.9 85.4 ± 2.0 86.3 ± 2.2 92.9 ± 1.0 89.5 ± 1.6 85.7 ± 2.1 0.78 ± 0.03
    0.001 85.8 ± 1.3 85.1 ± 1.8 85.2 ± 1.2 92.6 ± 0.6 88.7 ± 1.0 84.9 ± 1.1 0.77 ± 0.02
    0.0001 87.6 ± 1.0 87.4 ± 1.2 87.2 ± 1.3 93.4 ± 0.7 90.2 ± 1.0 87.2 ± 1.1 0.80 ± 0.02
    0.00001 84.8 ± 0.3 84.5 ± 1.0 84.0 ± 0.5 91.7 ± 0.4 87.5 ± 0.5 84.0 ± 0.3 0.76 ± 0.01

     | Show Table
    DownLoad: CSV

    After multiple experiments, the final parameter settings for different experiments in this study were as follows:

    (1) During the pretraining phase, the CNN was ResNet-101, the optimizer was Adam optimizer, with a batch size of 8 and a learning rate of 0.0001, and the loss function was cross-entropy loss. The "classifying image block order" pretext task was trained for 25 epochs, while the "predicting image rotation angle" pretext task was trained for 150 epochs.

    (2) During the training of the DBResNet, the Adam optimizer was used with a batch size of 8 and a learning rate of 0.0001, the loss function was cross-entropy loss, and the training was performed for 50 epochs.

    (3) During the training process of the other comparative experiments, the Adam optimizer was used with a batch size of 8 and a learning rate of 0.0001, the loss function was cross-entropy loss, and the training was performed for 50 epochs.

    Lastly, all experimental results in this study were obtained through 5 random experiments.

    To evaluate the performance of the Fusion-SSL method, we adopted several evaluation metrics, including accuracy, precision, sensitivity, specificity, G-Mean, F1 score [23], and Kappa coefficient [23]. In addition, to compare the performance of different classifiers, we also evaluated the performance of the Fusion-SSL method using the receiver operating characteristic curve (ROC) and the area under the ROC curve, Precision-Recall (PR) curve and the area under the PR curve. Based on the multiclassification confusion matrix, we determined the true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) for each of the three plaque classifications and then used them to obtain the following metrics. The equations are as follows:

    Accuracy=1KKk=1TPk+TNkTPk+TNk+FPk+FNk×100% (2.7)
    Precision=1KKk=1TPkTPk+FPk×100% (2.8)
    Sensitivity=1KKk=1TPkTPk+FNk×100% (2.9)
    Specificity=1KKk=1TNkTNk+FPk×100% (2.10)
    GMean=Sensitivity×Specificity×100% (2.11)
    F1=2×Precision×SensitivityPrecision+Sensitivity×100% (2.12)
    kappa=c×sKkpk×tks2Kkpk×tk (2.13)

    Here k is the k-th class, and K is the total number of classes. The calculation of Kappa is based on the confusion matrix C in Figure 8 of reference [23], where c=KkCkk is the total number of correctly predicted elements, s=KiKjCij is the total number of elements, pk=KiCki is the number of times the class K was predicted (column sum), and tk=KiCik is the number of times the class K truly occurred (row sum).

    This experiment evaluated the performance of popular network models in CNN (ShuffleNet-V2 [24], MobileNet-V3 [25], EfficientNet-B0 [26], and ResNet-101 [27]) to select the best model for plaque classification applications. To test the performance of CNNs on small training datasets, this experiment chose 40 and 100% of the labeled training images as examples. The experimental results are shown in Table 3. Among all the networks shown in Table 3, ResNet-101 has the best evaluation metrics for all the evaluation metrics in the case of 40 and 100% labeled training images. Therefore, ResNet-101 was used in this study as the base network for subsequent experiments. The deep structure, residual connection, pretrained model, and convolution pooling of ResNet-101 provide better feature extraction and initial weight, helping our model to reduce the load, accelerate convergence, and achieve better performance.

    Table 3.  Performance evaluation of different networks on 40 and 100% labeled training images.
    Model Proportion Accuracy
    (%)
    Precision
    (%)
    Sensitivity
    (%)
    Specificity
    (%)
    F1 score
    (%)
    Kappa
    ShuffleNet-V2 40% 64.1 ± 1.9 61.9 ± 2.9 62.1 ± 2.6 79.4 ± 1.1 60.6 ± 2.7 0.41 ± 0.03
    100% 69.0 ± 2.2 68.7 ± 3.3 66.6 ± 2.6 82.0 ± 1.1 65.4 ± 3.5 0.49 ± 0.04
    MobileNet-V3 40% 69.1 ± 1.3 71.3 ± 3.6 66.3 ± 2.0 81.8 ± 0.9 62.9 ± 3.7 0.48 ± 0.02
    100% 73.7 ± 0.7 73.9 ± 0.7 72.0 ± 0.6 85.0 ± 0.4 72.3 ± 0.4 0.57 ± 0.01
    EfficientNet-B0 40% 70.0 ± 0.7 69.0 ± 1.6 67.8 ± 0.5 82.9 ± 0.4 66.7 ± 1.2 0.51 ± 0.01
    100% 79.3 ± 0.9 79.2 ± 0.7 78.4 ± 1.1 88.6 ± 0.6 78.6 ± 1.0 0.67 ± 0.02
    ResNet-101 40% 78.4 ± 1.0 77.4 ± 1.1 77.2 ± 2.1 88.3 ± 0.9 76.8 ± 1.7 0.65 ± 0.02
    100% 83.3 ± 0.9 83.1 ± 0.5 82.1 ± 1.6 91.1 ± 0.7 82.4 ± 1.2 0.73 ± 0.02

     | Show Table
    DownLoad: CSV

    To show that Fusion-SSL can effectively relieve the issue of limited labeled images, this study compared Fusion-SSL with different SSL methods on diverse ratios of labeled training images, including 10, 20, 30, 40, and 100%.

    Table 4 and Figures 6 and 7 show the accuracy, precision, sensitivity, specificity, F1 score, and Kappa coefficient of SSL methods and the Fusion-SSL method trained with different proportions of labeled training images. The results indicate that the performance of the Fusion-SSL method is superior to individual SSL methods on all proportions of labeled training images.

    Table 4.  Performance comparison between Fusion-SSL and single SSL methods on labeled images of different proportions.
    Method Proportion Accuracy
    (%)
    Precision
    (%)
    Sensitivity
    (%)
    Specificity
    (%)
    G-Mean
    (%)
    F1 score
    (%)
    Kappa
    Ordering 10% 68.4 ± 2.4 67.4 ± 4.0 66.0 ± 2.2 81.8 ± 1.5 69.9 ± 3.7 64.3 ± 3.8 0.48 ± 0.04
    20% 75.1 ± 1.5 75.0 ± 1.3 73.8 ± 1.0 86.2 ± 0.8 79.1 ± 1.0 74.0 ± 0.8 0.60 ± 0.02
    30% 79.5 ± 2.3 79.0 ± 2.7 78.3 ± 3.5 89.0 ± 1.6 83.1 ± 2.8 78.2 ± 3.0 0.67 ± 0.04
    40% 81.2 ± 1.7 80.9 ± 1.5 80.0 ± 2.4 89.7 ± 1.2 84.3 ± 2.1 80.1 ± 2.2 0.70 ± 0.03
    100% 84.5 ± 1.1 84.4 ± 1.3 83.9 ± 1.5 91.6 ± 0.6 87.6 ± 1.1 84.1 ± 1.4 0.75 ± 0.02
    Rotating 10% 67.8 ± 1.8 63.8 ± 9.5 63.8 ± 2.6 80.5 ± 1.5 63.5 ± 1.7 60.7 ± 5.9 0.45 ± 0.04
    20% 71.1 ± 3.0 70.5 ± 4.8 68.6 ± 4.3 83.0 ± 2.3 71.6 ± 7.0 67.3 ± 6.5 0.52 ± 0.06
    30% 76.8 ± 1.7 77.5 ± 1.3 75.8 ± 2.7 86.9 ± 1.4 80.2 ± 3.0 75.8 ± 2.8 0.62 ± 0.03
    40% 80.4 ± 1.7 80.4 ± 1.6 79.0 ± 1.9 89.2 ± 1.1 83.5 ± 1.6 79.4 ± 1.9 0.68 ± 0.03
    100% 85.3 ± 0.9 85.4 ± 1.0 84.5 ± 1.3 91.9 ± 0.6 88.0 ± 1.0 84.9 ± 1.0 0.76 ± 0.02
    SimSiam 10% 73.1 ± 1.5 75.3 ± 2.1 72.6 ± 2.3 85.3 ± 0.9 77.9 ± 2.2 72.1 ± 2.2 0.57 ± 0.02
    20% 76.6 ± 1.5 76.4 ± 2.1 74.9 ± 2.0 86.9 ± 1.0 79.5 ± 2.1 74.7 ± 2.4 0.63 ± 0.03
    30% 78.2 ± 2.4 78.0 ± 2.5 78.2 ± 2.2 88.4 ± 1.1 83.0 ± 1.8 77.9 ± 2.3 0.65 ± 0.03
    40% 80.5 ± 1.2 81.0 ± 1.1 80.5 ± 2.0 89.3 ± 0.8 84.4 ± 1.2 80.5 ± 1.3 0.69 ± 0.02
    100% 85.3 ± 1.5 85.5 ± 1.2 85.2 ± 1.9 92.2 ± 0.8 88.5 ± 1.4 85.2 ± 1.6 0.77 ± 0.02
    Fusion-SSL
    (Our)
    10% 75.0 ± 2.6 75.5 ± 3.4 72.5 ± 3.5 86.2 ± 1.6 77.9 ± 3.6 72.8 ± 4.3 0.59 ± 0.05
    20% 80.9 ± 0.4 80.6 ± 0.8 79.5 ± 0.9 89.4 ± 0.3 83.6 ± 0.9 79.3 ± 1.0 0.69 ± 0.01
    30% 83.1 ± 1.5 82.9 ± 1.1 81.6 ± 1.8 90.6 ± 1.0 85.6 ± 1.6 81.9 ± 1.6 0.73 ± 0.03
    40% 84.5 ± 1.3 84.2 ± 0.9 83.9 ± 2.1 91.7 ± 0.9 87.5 ± 1.6 83.9 ± 1.6 0.75 ± 0.02
    100% 87.6 ± 1.0 87.4 ± 1.2 87.2 ± 1.3 93.4 ± 0.7 90.2 ± 1.0 87.2 ± 1.1 0.80 ± 0.02

     | Show Table
    DownLoad: CSV
    Figure 6.  Evaluation metrics of Fusion-SSL and single SSL methods on 100% labeled images.
    Figure 7.  Radar chart of evaluation metrics of Fusion-SSL and single SSL methods on 100% labeled images.

    For 10, 20, 30, 40, and 100% labeled training images, compared to the Ordering [20] method, the accuracy of Fusion-SSL was improved by 6.6, 5.8, 3.6, 3.3, and 3.1%. Compared to the Rotating [21] method, the accuracy of Fusion-SSL was improved by 7.2, 9.8, 6.3, 4.1, and 2.3%. Compared to the SimSiam [28] method, the accuracy of Fusion-SSL was improved by 1.9, 4.3, 4.9, 4.0, and 2.3%. Similarly, other evaluation metrics also yielded similar results.

    Figures 8 and 9 demonstrate the ROC curves and PR curves, further indicating the superior performance of Fusion-SSL on different scales of labeled training images. Table 4 and Figures 8 and 9 also show that the performance of Fusion-SSL on 40% labeled training images (accuracy = 0.845, precision = 0.842, sensitivity = 0.839, specificity = 0.917, G-Mean = 0.875, F1 = 0.839, kappa = 0.75, ROC-AUC = 0.940, PR-AUC = 0.880) is very close to the performance of the predicting image rotation angle method (Rotating) on 100% labeled training images (accuracy = 0.853, precision = 0.854, sensitivity = 0.845, specificity = 0.919, G-Mean = 0.880, F1 = 0.849, kappa = 0.76, ROC-AUC = 0.940, PR-AUC = 0.890).

    Figure 8.  ROC curves of Fusion-SSL and single SSL methods on the labeled training images of different proportions, where the numbers 40 and 100 represent the proportions of the labeled training images and the numbers in parentheses represent the AUC.
    Figure 9.  PR curves of Fusion-SSL and single SSL methods on the labeled training images of different proportions, where the numbers 40 and 100 represent the proportions of the labeled training images and the numbers in parentheses represent the AUC.

    Strokes are the primary cause of mortality and morbidity in upper-middle-income and high-income countries. The rupture of atherosclerotic plaques is one of the most significant causes of strokes and acute coronary syndrome. In this research, we proposed a Fusion-SSL method for automatic carotid plaque classification, which improved classification accuracy on a small set of labeled training images and demonstrated consistency with expert classification.

    Regarding SSL and supervised learning in carotid plaque ultrasound image classification, although existing supervised neural networks have provided great classification performance for carotid plaques, their performance heavily relies on the availability of a substantial number of labeled training samples. Unfortunately, obtaining these samples was difficult due to the time and effort required for manual data annotation. The experiments have found that for popular CNNs (ShuffleNet-V2, EfficientNet-B0, MobileNet-V3, and ResNet-101), carotid plaque classification performance decreased as the number of labeled training images decreased (Table 3). To efficiently relieve the issue of limited labeled carotid plaque ultrasound images, our study employed the SSL method. For example, the classifying image block order method on 40% labeled training images (accuracy = 0.812, precision = 0.809, sensitivity = 0.800, specificity = 0.890, F1 = 0.801, kappa = 0.70) performed close to the ResNet-101 method on 100% labeled training images (accuracy = 0.833, precision = 0.831, sensitivity = 0.821, specificity = 0.911, F1 = 0.824, kappa = 0.73). Taking accuracy as an example, the classifying image block order method on 40% labeled training images(accuracy = 0.812) accounts for 97.5% of the accuracy produced by the ResNet-101 method with 100% labeled training images(accuracy = 0.833). These results suggest that SSL methods can reduce the workload of manual annotation and have great potential for clinical use.

    We compared Fusion-SSL with a single SSL method. Although a single SSL method can effectively relieve the issue of limited labeled training images and enhance the performance of the basic CNN, the challenge lies in how to learn richer plaque features through pretext tasks. To address this issue and improve the performance of CNN in carotid plaque classification, this paper redirected the research focus toward the fusion of self-supervised networks. We proposed a Fusion-SSL method to fuse two different SSL methods. To validate the effectiveness of Fusion-SSL, a comparison was made between the Fusion-SSL method and single SSL methods. The experimental results in Table 4 showed that the performance of Fusion-SSL was superior to the single SSL methods in different cases (labeled training images ranging from 10 to 100%). This suggests that Fusion-SSL can acquire more valuable information about carotid plaques than a single SSL method.

    Although feature fusion has advantages in improving the model performance, it also has some disadvantages and limitations, which need to be balanced and processed in practical application. With the increase in dataset size and network depth, feature fusion may increase memory usage and computation time. Additionally, if there are too many features or a high correlation among features, it may lead to model overfitting, which consumes time and computational resources to perform multiple experiments to avoid. SSL methods have enormous potential in carotid plaque ultrasound image classification and deserve more in-depth studies. However, due to inherent limitations, the proposed Fusion-SSL method in this study only combines two simple SSL methods, and the complex features of the plaque have not been fully understood yet. In subsequent research, we need to explore more pretext tasks to extract the visual features of carotid plaques and integrate more SSL methods to enhance the classification performance of carotid plaques with a few labeled training images.

    In this paper, we proposed a self-supervised fusion network for carotid plaque ultrasound image classification, which enabled us to obtain richer features of carotid plaque to enhance classification performance. The experimental results demonstrated that the proposed method effectively mitigates the impact of limited labeled images on the accuracy of carotid plaque classification and outperforms single SSL methods with significant improvements in classification performance under different proportions of labeled training images. This indicates that our method is beneficial for the clinical recognition of carotid plaques and stroke warning, providing a new research approach in the field of carotid plaque ultrasound image classification.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    This work was supported by the National Natural Science Foundation of China under grant No. 62201203, the Natural Science Foundation of Hubei Province, China under grant No. 2021CFB282, and the Doctoral Scientific Research Foundation of Hubei University of Technology and China under grant No. BSDQ2020064.

    The authors declare there is no conflict of interest.



    [1] Elmore SA, Boorman GA (2023) Environmental Toxicologic Pathology and Human Health. Haschek and Rousseaux's Handbook of Toxicologic Pathology, Volume 3: Environmental Toxicologic Pathology and Major Toxicant Classes 3–32. https://doi.org/10.1016/B978-0-443-16153-7.00001-0
    [2] Sadia M, Nollen I, Helmus R, et al. (2023) Occurrence, Fate, and Related Health Risks of PFAS in Raw and Produced Drinking Water. Environ Sci Technol 57: 3062–3074. https://doi.org/10.1021/acs.est.2c06015 doi: 10.1021/acs.est.2c06015
    [3] Lohmann R, Cousins IT, Dewitt JC, et al. (2020) Are Fluoropolymers Really of Low Concern for Human and Environmental Health and Separate from Other PFAS? Environ Sci Technol 54: 12820–12828. https://doi.org/10.1021/acs.est.0c03244 doi: 10.1021/acs.est.0c03244
    [4] Yan S, Murtadha B, Foster G, et al. (2023) Development and testing of novel functionalized polymeric thin-films for equilibrium passive sampling of PFAS compounds in water. Chemical Engineering Journal 477. https://doi.org/10.1016/j.cej.2023.146734 doi: 10.1016/j.cej.2023.146734
    [5] Bayode AA, Emmanuel SS, Akinyemi AO, et al. (2024) Innovative techniques for combating a common enemy forever chemicals: A comprehensive approach to mitigating per- and polyfluoroalkyl substances (PFAS) contamination. Environ Res 261. https://doi.org/10.1016/j.envres.2024.119719 doi: 10.1016/j.envres.2024.119719
    [6] Bayode AA, Emmanuel SS, Akinyemi AO, et al. (2024) Innovative techniques for combating a common enemy forever chemicals: A comprehensive approach to mitigating per- and polyfluoroalkyl substances (PFAS) contamination. Environ Res 261. https://doi.org/10.1016/j.envres.2024.119719 doi: 10.1016/j.envres.2024.119719
    [7] Park M, Wu S, Lopez IJ, et al. (2020) Adsorption of perfluoroalkyl substances (PFAS) in groundwater by granular activated carbons: Roles of hydrophobicity of PFAS and carbon characteristics. Water Res 170: 115364. https://doi.org/10.1016/j.watres.2019.115364 doi: 10.1016/j.watres.2019.115364
    [8] Tian Y, Zhou Q, Zhang L, et al. (2023) In utero exposure to per-/polyfluoroalkyl substances (PFASs): Preeclampsia in pregnancy and low birth weight for neonates. Chemosphere 313: 137490. https://doi.org/10.1016/j.chemosphere.2022.137490 doi: 10.1016/j.chemosphere.2022.137490
    [9] Ashley-Martin J, Hammond J, Velez MP (2024) Assessing preconception exposure to environmental chemicals and fecundity: Strategies, challenges, and research priorities. Reproductive Toxicology 125. https://doi.org/10.1016/j.reprotox.2024.108578 doi: 10.1016/j.reprotox.2024.108578
    [10] Teymoorian T, Munoz G, Sauvé S (2025) PFAS contamination in tap water: Target and suspect screening of zwitterionic, cationic, and anionic species across Canada and beyond. Environ Int 195. https://doi.org/10.1016/j.envint.2025.109250 doi: 10.1016/j.envint.2025.109250
    [11] Merino N, Qu Y, Deeb RA, et al. (2016) Degradation and Removal Methods for Perfluoroalkyl and Polyfluoroalkyl Substances in Water. Environ Eng Sci 33: 615–649. https://doi.org/10.1016/j.envint.2025.109250 doi: 10.1016/j.envint.2025.109250
    [12] Berhanu A, Mutanda I, Taolin J, et al. (2023) A review of microbial degradation of per- and polyfluoroalkyl substances (PFAS): Biotransformation routes and enzymes. Science of the Total Environment 859. https://doi.org/10.1016/j.scitotenv.2022.160010 doi: 10.1016/j.scitotenv.2022.160010
    [13] Kuppan N, Padman M, Mahadeva M, et al. (2024) A comprehensive review of sustainable bioremediation techniques: Eco friendly solutions for waste and pollution management. Waste Management Bulletin 2: 154–171. https://doi.org/10.1016/j.wmb.2024.07.005 doi: 10.1016/j.wmb.2024.07.005
    [14] Chukwuonye GN, Alqattan ZA, Jones M, et al. (2024) Toxic layering and compound extremes: Per- and polyfluoroalkyl substances (PFAS) exposure in rural, environmental justice copper mining communities. Science of the Total Environment 957. https://doi.org/10.1016/j.scitotenv.2024.177767 doi: 10.1016/j.scitotenv.2024.177767
    [15] Wang F, Xiang L, Sze-Yin Leung K, et al. (2024) Emerging contaminants: A One Health perspective. Innovation 5.
    [16] Groffen T, Bervoets L, Jeong Y, et al. (2021) A rapid method for the detection and quantification of legacy and emerging per- and polyfluoroalkyl substances (PFAS) in bird feathers using UPLC-MS/MS. Journal of Chromatography B 1172: 122653. https://doi.org/10.1016/j.jchromb.2021.122653 doi: 10.1016/j.jchromb.2021.122653
    [17] Dodds JN, Kirkwood-Donelson KI, Boatman AK, et al. (2024) Evaluating Solid Phase Adsorption Toxin Tracking (SPATT) for passive monitoring of per- and polyfluoroalkyl substances (PFAS) with Ion Mobility Spectrometry-Mass Spectrometry (IMS-MS). Science of The Total Environment 947: 174574. https://doi.org/10.1016/j.scitotenv.2024.174574 doi: 10.1016/j.scitotenv.2024.174574
    [18] Manojkumar Y, Pilli S, Rao PV, et al. (2023) Sources, occurrence and toxic effects of emerging per- and polyfluoroalkyl substances (PFAS). Neurotoxicol Teratol 97: 107174. https://doi.org/10.1016/j.ntt.2023.107174 doi: 10.1016/j.ntt.2023.107174
    [19] De Silva AO, Armitage JM, Bruton TA, et al. (2021) PFAS Exposure Pathways for Humans and Wildlife: A Synthesis of Current Knowledge and Key Gaps in Understanding. Environ Toxicol Chem 40: 631–657. https://doi.org/10.1002/etc.4935 doi: 10.1002/etc.4935
    [20] Gaines LGT (2023) Historical and current usage of per- and polyfluoroalkyl substances (PFAS): A literature review. Am J Ind Med 66: 353–378. https://doi.org/10.1002/ajim.23362 doi: 10.1002/ajim.23362
    [21] Brunn H, Arnold G, Kö rner W, et al. (2023) PFAS: forever chemicals—persistent, bioaccumulative and mobile. Reviewing the status and the need for their phase out and remediation of contaminated sites. Environ Sci Eur 35. https://doi.org/10.1186/s12302-023-00721-8 doi: 10.1186/s12302-023-00721-8
    [22] Cookson ES, Detwiler RL (2022) Global patterns and temporal trends of perfluoroalkyl substances in municipal wastewater: A meta-analysis. Water Res 221. https://doi.org/10.1016/j.watres.2022.118784 doi: 10.1016/j.watres.2022.118784
    [23] Jha G, Kankarla V, McLennon E, et al. (2021) Per- and Polyfluoroalkyl Substances (PFAS) in Integrated Crop–Livestock Systems: Environmental Exposure and Human Health Risks. Int J Environ Res Public Health 18: 12550. https://doi.org/10.3390/ijerph182312550 doi: 10.3390/ijerph182312550
    [24] Lin H, Lao JY, Wang Q, et al. (2022) Per- and polyfluoroalkyl substances in the atmosphere of waste management infrastructures: Uncovering secondary fluorotelomer alcohols, particle size distribution, and human inhalation exposure. Environ Int 167. https://doi.org/10.1016/j.envint.2022.107434 doi: 10.1016/j.envint.2022.107434
    [25] Wang P, Zhang M, Li Q, et al. (2021) Atmospheric diffusion of perfluoroalkyl acids emitted from fluorochemical industry and its associated health risks. Environ Int 146. https://doi.org/10.1016/j.envint.2020.106247 doi: 10.1016/j.envint.2020.106247
    [26] Buck RC, Korzeniowski SH, Laganis E, et al. (2021) Identification and classification of commercially relevant per- and poly-fluoroalkyl substances (PFAS). Integr Environ Assess Manag 17: 1045–1055. https://doi.org/10.1002/ieam.4450 doi: 10.1002/ieam.4450
    [27] Zhang X, Liu J, Zhang H, et al. (2025) Highly selective guanidine-linked covalent organic framework for efficient removal of perfluoroalkyl carboxylic acids from water samples. Sep Purif Technol 357: 130039. https://doi.org/10.1016/j.seppur.2024.130039 doi: 10.1016/j.seppur.2024.130039
    [28] Xu C, Yin S, Liu Y, et al. (2019) Prenatal exposure to chlorinated polyfluoroalkyl ether sulfonic acids and perfluoroalkyl acids: Potential role of maternal determinants and associations with birth outcomes. J Hazard Mater 380. https://doi.org/10.1016/j.jhazmat.2019.120867 doi: 10.1016/j.jhazmat.2019.120867
    [29] Brendel S, Fetter É, Staude C, et al. (2018) Short-chain perfluoroalkyl acids: environmental concerns and a regulatory strategy under REACH. Environ Sci Eur 30: 1–11. https://doi.org/10.1186/s12302-018-0134-4 doi: 10.1186/s12302-018-0134-4
    [30] Marques E, Pfohl M, Wei W, et al. (2022) Replacement per- and polyfluoroalkyl substances (PFAS) are potent modulators of lipogenic and drug metabolizing gene expression signatures in primary human hepatocytes. Toxicol Appl Pharmacol 442. https://doi.org/10.1016/j.taap.2022.115991 doi: 10.1016/j.taap.2022.115991
    [31] Roth K, Yang Z, Agarwal M, et al. (2021) Exposure to a mixture of legacy, alternative, and replacement per- and polyfluoroalkyl substances (PFAS) results in sex-dependent modulation of cholesterol metabolism and liver injury. Environ Int 157. https://doi.org/10.1016/j.envint.2021.106843 doi: 10.1016/j.envint.2021.106843
    [32] Pan Z, Miao W, Wang C, et al. (2021) 6:2 Cl-PFESA has the potential to cause liver damage and induce lipid metabolism disorders in female mice through the action of PPAR-γ. Environmental Pollution 287: 117329. https://doi.org/10.1016/j.envpol.2021.117329 doi: 10.1016/j.envpol.2021.117329
    [33] Pervez MN, Jiang T, Mahato JK, et al. (2024) Surface Modification of Graphene Oxide for Fast Removal of Per- and Polyfluoroalkyl Substances (PFAS) Mixtures from River Water. ACS ES and T Water 4: 2968–2980. https://doi.org/10.1021/acsestwater.4c00187 doi: 10.1021/acsestwater.4c00187
    [34] Xu C, Yin S, Liu Y, et al. (2019) Prenatal exposure to chlorinated polyfluoroalkyl ether sulfonic acids and perfluoroalkyl acids: Potential role of maternal determinants and associations with birth outcomes. J Hazard Mater 380. https://doi.org/10.1016/j.jhazmat.2019.120867 doi: 10.1016/j.jhazmat.2019.120867
    [35] Gomez-Ruiz B, Gómez-Lavín S, Diban N, et al. (2017) Boron doped diamond electrooxidation of 6:2 fluorotelomers and perfluorocarboxylic acids. Application to industrial wastewaters treatment. Journal of Electroanalytical Chemistry 798: 51–57. https://doi.org/10.1016/j.jelechem.2017.05.033 doi: 10.1016/j.jelechem.2017.05.033
    [36] Daramola O (2021) Metabolism of 6: 2 Fluorotelomer Alcohol by CYP 2A6.
    [37] Riedel TP, Wallace MAG, Shields EP, et al. (2021) Low temperature thermal treatment of gas-phase fluorotelomer alcohols by calcium oxide. Chemosphere 272. https://doi.org/10.1016/j.chemosphere.2021.129859 doi: 10.1016/j.chemosphere.2021.129859
    [38] Kassotis CD, Vandenberg LN, Demeneix BA, et al. (2020) Endocrine-disrupting chemicals: economic, regulatory, and policy implications. Lancet Diabetes Endocrinol 8: 719–730. https://doi.org/10.1016/S2213-8587(20)30128-5 doi: 10.1016/S2213-8587(20)30128-5
    [39] Manojkumar Y, Pilli S, Rao PV, et al. (2023) Sources, occurrence and toxic effects of emerging per- and polyfluoroalkyl substances (PFAS). Neurotoxicol Teratol 97. https://doi.org/10.1016/j.ntt.2023.107174 doi: 10.1016/j.ntt.2023.107174
    [40] Thompson D, Zolfigol N, Xia Z, et al. (2024) Recent progress in per- and polyfluoroalkyl substances (PFAS) sensing: A critical mini-review. Sensors and Actuators Reports 100189. https://doi.org/10.1016/j.snr.2024.100189 doi: 10.1016/j.snr.2024.100189
    [41] Sharma N, Kumar V, Sugumar V, et al. (2024) A comprehensive review on the need for integrated strategies and process modifications for per- and polyfluoroalkyl substances (PFAS) removal: Current insights and future prospects. Case Studies in Chemical and Environmental Engineering 9. https://doi.org/10.1016/j.cscee.2024.100623 doi: 10.1016/j.cscee.2024.100623
    [42] Lu Y, Gao J, Nguyen HT, et al. (2021) Occurrence of per- and polyfluoroalkyl substances (PFASs) in wastewater of major cities across China in 2014 and 2016. Chemosphere 279. https://doi.org/10.1016/j.chemosphere.2021.130590 doi: 10.1016/j.chemosphere.2021.130590
    [43] Valencia A, Ordonez D, Sadmani AHMA, et al. (2023) Comparing the removal and fate of long and short chain per- and polyfluoroalkyl substances (PFAS) during surface water treatment via specialty adsorbents. Journal of Water Process Engineering 56. https://doi.org/10.1016/j.jwpe.2023.104345 doi: 10.1016/j.jwpe.2023.104345
    [44] Giannelli Moneta B, Feo ML, Torre M, et al. (2023) Occurrence of per- and polyfluorinated alkyl substances in wastewater treatment plants in Northern Italy. Science of the Total Environment 894. https://doi.org/10.1016/j.scitotenv.2023.165089 doi: 10.1016/j.scitotenv.2023.165089
    [45] Zhang YH, Ding TT, Huang ZY, et al. (2023) Environmental exposure and ecological risk of perfluorinated substances (PFASs) in the Shaying River Basin, China. Chemosphere 339. https://doi.org/10.1016/j.chemosphere.2023.139537 doi: 10.1016/j.chemosphere.2023.139537
    [46] Jeong Y, Da Silva KM, Iturrospe E, et al. (2022) Occurrence and contamination profile of legacy and emerging per- and polyfluoroalkyl substances (PFAS) in Belgian wastewater using target, suspect and non-target screening approaches. J Hazard Mater 437. https://doi.org/10.1016/j.jhazmat.2022.129378 doi: 10.1016/j.jhazmat.2022.129378
    [47] Zarębska M, Bajkacz S, Hordyjewicz-Baran Z (2024) Assessment of legacy and emerging PFAS in the Oder River: Occurrence, distribution, and sources. Environ Res 118608. https://doi.org/10.1016/j.envres.2024.118608 doi: 10.1016/j.envres.2024.118608
    [48] Labine LM, Oliveira Pereira EA, Kleywegt S, et al. (2022) Comparison of sub-lethal metabolic perturbations of select legacy and novel perfluorinated alkyl substances (PFAS) in Daphnia magna. Environ Res 212. https://doi.org/10.1016/j.envres.2022.113582 doi: 10.1016/j.envres.2022.113582
    [49] EPA- USA (2024) United States Environmental Protection Agency, Per- and Polyfluoroalkyl Substances (PFAS) | US EPA, 2024. Available from: https://www.epa.gov/pfas.
    [50] Brunn H, Arnold G, Kö rner W, et al. (2023) PFAS: forever chemicals—persistent, bioaccumulative and mobile. Reviewing the status and the need for their phase out and remediation of contaminated sites. Environ Sci Eur 35. https://doi.org/10.1186/s12302-023-00721-8 doi: 10.1186/s12302-023-00721-8
    [51] Moretti S, Brambilla G, Maffucci F, et al. (2024) Occurrence and pattern of legacy and emerging per- and Poly-FluoroAlkyl substances (PFAS) in eggs of loggerhead turtle Caretta caretta from western Mediterranean. Environmental Pollution 343. https://doi.org/10.1016/j.envpol.2023.123257 doi: 10.1016/j.envpol.2023.123257
    [52] Ankley GT, Cureton P, Hoke RA, et al. (2021) Assessing the Ecological Risks of Per- and Polyfluoroalkyl Substances: Current State-of-the Science and a Proposed Path Forward. Environ Toxicol Chem 40: 564–605. https://doi.org/10.1002/etc.4869 doi: 10.1002/etc.4869
    [53] Cousins I, DeWitt J, Glüge J, et al. Strategies for grouping per-and polyfluoroalkyl substances (PFAS) to protect human and environmental health. pubs.rsc.org.
    [54] Guelfo JL, Korzeniowski S, Mills MA, et al. (2021) Environmental Sources, Chemistry, Fate, and Transport of Per- and Polyfluoroalkyl Substances: State of the Science, Key Knowledge Gaps, and Recommendations Presented at the August 2019 SETAC Focus Topic Meeting. Environ Toxicol Chem 40: 3234–3260. https://doi.org/10.1002/etc.5182 doi: 10.1002/etc.5182
    [55] McCord J, Strynar M (2019) Identification of Per- and Polyfluoroalkyl Substances in the Cape Fear River by High Resolution Mass Spectrometry and Nontargeted Screening. Environ Sci Technol 53: 4717–4727. https://doi.org/10.1021/acs.est.8b06017 doi: 10.1021/acs.est.8b06017
    [56] Strynar M, Dagnino S, McMahen R, et al. (2015) Identification of Novel Perfluoroalkyl Ether Carboxylic Acids (PFECAs) and Sulfonic Acids (PFESAs) in Natural Waters Using Accurate Mass Time-of-Flight Mass Spectrometry (TOFMS). Environ Sci Technol 49: 11622–11630. https://doi.org/10.1021/acs.est.5b01215 doi: 10.1021/acs.est.5b01215
    [57] Tsianou M, Bedrov D, Alexandridis P (2023) Surfactants in the Environment: Self-Assembly of PFAS Pollutants in Solution and at Interfaces. ACS Symposium Series 1457: 443–462. https://doi.org/10.1021/bk-2023-1457.ch016 doi: 10.1021/bk-2023-1457.ch016
    [58] Yao J, Sheng N, Guo Y, et al. (2022) Nontargeted Identification and Temporal Trends of Per- and Polyfluoroalkyl Substances in a Fluorochemical Industrial Zone and Adjacent Taihu Lake. Environ Sci Technol 56: 7986–7996. https://doi.org/10.1021/acs.est.2c00891 doi: 10.1021/acs.est.2c00891
    [59] Jin B, Zhu Y, Zhao W, et al. (2023) Aerobic Biotransformation and Defluorination of Fluoroalkylether Substances (ether PFAS): Substrate Specificity, Pathways, and Applications. Environ Sci Technol Lett 10: 755–761. https://doi.org/10.1021/acs.estlett.3c00411 doi: 10.1021/acs.estlett.3c00411
    [60] Frigerio G, Cafagna S, Polledri E, et al. (2022) Development and validation of an LC–MS/MS method for the quantitation of 30 legacy and emerging per- and polyfluoroalkyl substances (PFASs) in human plasma, including HFPO-DA, DONA, and cC6O4. Anal Bioanal Chem 414: 1259–1278. https://doi.org/10.1007/s00216-021-03762-1 doi: 10.1007/s00216-021-03762-1
    [61] Liu J, international SA-E, 2013 undefined (2013) Microbial degradation of polyfluoroalkyl chemicals in the environment: a review. Elsevier. https://doi.org/10.1016/j.envint.2013.08.022
    [62] Cousins I, DeWitt J, Glüge J, et al. The high persistence of PFAS is sufficient for their management as a chemical class. pubs.rsc.org.
    [63] Meegoda J, Kewalramani J, … BL-… journal of environmental, et al. A review of the applications, environmental release, and remediation technologies of per-and polyfluoroalkyl substances. mdpi.com.
    [64] Walkowiak-Kulikowska J (2022) Poly/Perfluorinated Alkyl Substances (PFASs)–Synthetic Methods, Properties and Applications. https://doi.org/10.1039/9781839167591-00022
    [65] Olsen GW (2015) PFAS Biomonitoring in Higher Exposed Populations. Molecular and Integrative Toxicology 77–125. https://doi.org/10.1007/978-3-319-15518-0_4 doi: 10.1007/978-3-319-15518-0_4
    [66] Guelfo JL, Korzeniowski S, Mills MA, et al. (2021) Environmental sources, chemistry, Fate, and transport of per‐and Polyfluoroalkyl Substances: State of the science, key knowledge gaps, and recommendations. Wiley Online Library 40: 3234–3260. https://doi.org/10.1002/etc.5182 doi: 10.1002/etc.5182
    [67] Yu X, Nishimura F, Hidaka T (2018) Effects of microbial activity on perfluorinated carboxylic acids (PFCAs) generation during aerobic biotransformation of fluorotelomer alcohols in activated sludge. Science of the Total Environment 610–611: 776–785. https://doi.org/10.1016/j.scitotenv.2017.08.075 doi: 10.1016/j.scitotenv.2017.08.075
    [68] Manojkumar Y, Pilli S, Rao PV, et al. (2023) Sources, occurrence and toxic effects of emerging per- and polyfluoroalkyl substances (PFAS). Neurotoxicol Teratol 97. https://doi.org/10.1016/j.ntt.2023.107174 doi: 10.1016/j.ntt.2023.107174
    [69] Choi YJ, Helbling DE, Liu J, et al. (2022) Microbial biotransformation of aqueous film-forming foam derived polyfluoroalkyl substances. Science of the Total Environment 824. https://doi.org/10.1016/j.scitotenv.2022.153711 doi: 10.1016/j.scitotenv.2022.153711
    [70] Sharma N, Kumar V, Sugumar V, et al. (2024) A comprehensive review on the need for integrated strategies and process modifications for per- and polyfluoroalkyl substances (PFAS) removal: Current insights and future prospects. Case Studies in Chemical and Environmental Engineering 9. https://doi.org/10.1016/j.cscee.2024.100623 doi: 10.1016/j.cscee.2024.100623
    [71] Dong S, Yan PF, Liu C, et al. (2023) Assessing aerobic biotransformation of 8:2 fluorotelomer alcohol in aqueous film-forming foam (AFFF)-impacted soils: Pathways and microbial community dynamics. J Hazard Mater 446. https://doi.org/10.1016/j.jhazmat.2022.130629 doi: 10.1016/j.jhazmat.2022.130629
    [72] Fiedler H, Kennedy T, Henry BJ (2021) A Critical Review of a Recommended Analytical and Classification Approach for Organic Fluorinated Compounds with an Emphasis on Per- and Polyfluoroalkyl Substances. Integr Environ Assess Manag 17: 331–351. https://doi.org/10.1002/ieam.4352 doi: 10.1002/ieam.4352
    [73] Guelfo JL, Korzeniowski S, Mills MA, et al. (2021) Environmental Sources, Chemistry, Fate, and Transport of Per- and Polyfluoroalkyl Substances: State of the Science, Key Knowledge Gaps, and Recommendations Presented at the August 2019 SETAC Focus Topic Meeting. Environ Toxicol Chem 40: 3234–3260. https://doi.org/10.1002/etc.5182 doi: 10.1002/etc.5182
    [74] Dasu K, Xia X, Siriwardena D, et al. (2022) Concentration profiles of per- and polyfluoroalkyl substances in major sources to the environment. J Environ Manage 301. https://doi.org/10.1016/j.jenvman.2021.113879 doi: 10.1016/j.jenvman.2021.113879
    [75] Ritscher A, Wang Z, Scheringer M, et al. (2018) Zürich Statement on Future Actions on Per- and Polyfluoroalkyl Substances (PFASs). Environ Health Perspect 126. https://doi.org/10.1289/EHP4158 doi: 10.1289/EHP4158
    [76] Issaka E, Adams M, Baffoe J, et al. (2024) Covalent organic frameworks: a review of synthesis methods, properties and applications for per- and poly-fluoroalkyl substances removal. Clean Technologies and Environmental Policy 2024 1–28.
    [77] Holmquist H, Schellenberger S, van der Veen I, et al. (2016) Properties, performance and associated hazards of state-of-the-art durable water repellent (DWR) chemistry for textile finishing. Environ Int 91: 251–264. https://doi.org/10.1016/j.envint.2016.02.035 doi: 10.1016/j.envint.2016.02.035
    [78] Li YF, Ho CY, Liu YJ, et al. (2024) Enhance electrocoagulation-flotation (ECF) removal efficiency perfluorohexanoic acid (PFHxA) by adding surfactants. J Environ Chem Eng 12. https://doi.org/10.1016/j.jece.2023.111773 doi: 10.1016/j.jece.2023.111773
    [79] Manouras E, Ioannou D, Zeniou A, et al. (2024) Superhydrophobic and oleophobic Nylon, PES and PVDF membranes using plasma nanotexturing: Empowering membrane distillation and contributing to PFAS free hydrophobic membranes. Micro and Nano Engineering 24. https://doi.org/10.1016/j.mne.2024.100269 doi: 10.1016/j.mne.2024.100269
    [80] Kourtchev I, Sebben BG, Bogush A, et al. (2023) Per- and polyfluoroalkyl substances (PFASs) in urban PM2.5 samples from Curitiba, Brazil. Atmos Environ 309. https://doi.org/10.1016/j.atmosenv.2023.119911 doi: 10.1016/j.atmosenv.2023.119911
    [81] Ji B, Zhao Y, Yang Y, et al. (2023) Curbing per- and polyfluoroalkyl substances (PFASs): First investigation in a constructed wetland-microbial fuel cell system. Water Res 230. https://doi.org/10.1016/j.watres.2022.119530 doi: 10.1016/j.watres.2022.119530
    [82] Yamijala SSRKC, Shinde R, Hanasaki K, et al. (2022) Photo-induced degradation of PFASs: Excited-state mechanisms from real-time time-dependent density functional theory. J Hazard Mater 423. https://doi.org/10.1016/j.jhazmat.2021.127026 doi: 10.1016/j.jhazmat.2021.127026
    [83] Yadav S, Ibrar I, Al-Juboori RA, et al. (2022) Updated review on emerging technologies for PFAS contaminated water treatment. Chemical Engineering Research and Design 182: 667–700. https://doi.org/10.1016/j.cherd.2022.04.009 doi: 10.1016/j.cherd.2022.04.009
    [84] Ducrotoy J-P (2024) Chemical Introductions to the Systems: Point Source Pollution (Persistent Chemicals). Reference Module in Earth Systems and Environmental Sciences. https://doi.org/10.1016/B978-0-323-90798-9.00083-4 doi: 10.1016/B978-0-323-90798-9.00083-4
    [85] Popli S, Badgujar PC, Agarwal T, et al. (2022) Persistent organic pollutants in foods, their interplay with gut microbiota and resultant toxicity. Science of the Total Environment 832. https://doi.org/10.1016/j.scitotenv.2022.155084 doi: 10.1016/j.scitotenv.2022.155084
    [86] Krafft MP, Riess JG (2015) Per- and polyfluorinated substances (PFASs): Environmental challenges. Curr Opin Colloid Interface Sci 20: 192–212. https://doi.org/10.1016/j.cocis.2015.07.004 doi: 10.1016/j.cocis.2015.07.004
    [87] Koskue V, Monetti J, Rossi N, et al. (2022) Fate of pharmaceuticals and PFASs during the electrochemical generation of a nitrogen-rich nutrient product from real reject water. J Environ Chem Eng 10. https://doi.org/10.1016/j.jece.2022.107284 doi: 10.1016/j.jece.2022.107284
    [88] Ducrotoy J-P (2024) Chemical Introductions to the Systems: Point Source Pollution (Persistent Chemicals). Reference Module in Earth Systems and Environmental Sciences. https://doi.org/10.1016/B978-0-323-90798-9.00083-4 doi: 10.1016/B978-0-323-90798-9.00083-4
    [89] Liu F, Pignatello JJ, Sun R, et al. (2024) A Comprehensive Review of Novel Adsorbents for Per- and Polyfluoroalkyl Substances in Water. ACS ES & T Water. https://doi.org/10.1021/acsestwater.3c00569 doi: 10.1021/acsestwater.3c00569
    [90] Oliver DP, Navarro DA, Baldock J, et al. (2020) Sorption behaviour of per- and polyfluoroalkyl substances (PFASs) as affected by the properties of coastal estuarine sediments. Science of The Total Environment 720: 137263. https://doi.org/10.1016/j.scitotenv.2020.137263 doi: 10.1016/j.scitotenv.2020.137263
    [91] Mikhael E, Bouazza A, Gates WP, et al. (2024) Unlocking the sorption mechanism of perfluoroalkyl acids (PFAAs) on geosynthetics: Case of the geotextile components of geosynthetic clay liners. Geotextiles and Geomembranes 52: 59–71. https://doi.org/10.1016/j.geotexmem.2023.09.002 doi: 10.1016/j.geotexmem.2023.09.002
    [92] Umeh AC, Hassan M, Egbuatu M, et al. (2023) Multicomponent PFAS sorption and desorption in common commercial adsorbents: Kinetics, isotherm, adsorbent dose, pH, and index ion and ionic strength effects. Science of the Total Environment 904. https://doi.org/10.1016/j.scitotenv.2023.166568 doi: 10.1016/j.scitotenv.2023.166568
    [93] Qi Y, Cao H, Pan W, et al. (2022) The role of dissolved organic matter during Per- and Polyfluorinated Substance (PFAS) adsorption, degradation, and plant uptake: A review. J Hazard Mater 436. https://doi.org/10.1016/j.jhazmat.2022.129139 doi: 10.1016/j.jhazmat.2022.129139
    [94] Issaka E, Wariboko MA, Mohammed A, et al. (2023) Trends in enzyme mimics for enhanced catalytic cascade systems for bio-sensing of environmental pollutants -A review. Chemical Engineering Journal Advances 15. https://doi.org/10.1016/j.ceja.2023.100510 doi: 10.1016/j.ceja.2023.100510
    [95] M.B. B, Rhakho N, Jena SR, et al. (2023) Detection of PFAS via surface-enhanced Raman scattering: Challenges and future perspectives. Sustainable Chemistry for the Environment 3: 100031. https://doi.org/10.1016/j.scenv.2023.100031 doi: 10.1016/j.scenv.2023.100031
    [96] Deng Z-H, Cheng C-G, Wang X-L, et al. (2018) Preconcentration and Determination of Perfluoroalkyl Substances (PFASs) in Water Samples by Bamboo Charcoal-Based Solid-Phase Extraction Prior to Liquid Chromatography–Tandem Mass Spectrometry. Molecules 23: 902. https://doi.org/10.3390/molecules23040902 doi: 10.3390/molecules23040902
    [97] Miserli K, Athanasiou V, Boti V, et al. (2023) Determination of PFAS in wastewaters and natural waters by solid phase extraction and UHPLC LTQ/Orbitrap MS for assessing occurrence and removals. Case Studies in Chemical and Environmental Engineering 8: 100505. https://doi.org/10.1016/j.cscee.2023.100505 doi: 10.1016/j.cscee.2023.100505
    [98] Tang S, Qin X, Lv Y, et al. (2022) Adsorption of three perfluoroalkyl sulfonate compounds from environmental water and human serum samples using cationic porous covalent organic framework as adsorbents and detection combination with MALDI-TOF MS. Appl Surf Sci 601: 154224. https://doi.org/10.1016/j.apsusc.2022.154224 doi: 10.1016/j.apsusc.2022.154224
    [99] Nassazzi W, Lai FY, Ahrens L (2022) A novel method for extraction, clean-up and analysis of per- and polyfluoroalkyl substances (PFAS) in different plant matrices using LC-MS/MS. Journal of Chromatography B 1212: 123514. https://doi.org/10.1016/j.jchromb.2022.123514 doi: 10.1016/j.jchromb.2022.123514
    [100] Piva E, Fais P, Cecchetto G, et al. (2021) Determination of perfluoroalkyl substances (PFAS) in human hair by liquid chromatography-high accurate mass spectrometry (LC-QTOF). Journal of Chromatography B 1172: 122651. https://doi.org/10.1016/j.jchromb.2021.122651 doi: 10.1016/j.jchromb.2021.122651
    [101] Dvorakova D, Jurikova M, Svobodova V, et al. (2023) Complex monitoring of perfluoroalkyl substances (PFAS) from tap drinking water in the Czech Republic. Water Res 247: 120764. https://doi.org/10.1016/j.watres.2023.120764 doi: 10.1016/j.watres.2023.120764
    [102] Forster ALB, Zhang Y, Westerman DC, et al. (2023) Improved total organic fluorine methods for more comprehensive measurement of PFAS in industrial wastewater, river water, and air. Water Res 235: 119859. https://doi.org/10.1016/j.watres.2023.119859 doi: 10.1016/j.watres.2023.119859
    [103] Simon F, Gehrenkemper L, von der Au M, et al. (2022) A fast and simple PFAS extraction method utilizing HR–CS–GFMAS for soil samples. Chemosphere 295: 133922. https://doi.org/10.1016/j.chemosphere.2022.133922 doi: 10.1016/j.chemosphere.2022.133922
    [104] Wang Y, Munir U, Huang Q (2023) Occurrence of per- and polyfluoroalkyl substances (PFAS) in soil: Sources, fate, and remediation. Soil and Environmental Health 1. https://doi.org/10.1016/j.seh.2023.100004 doi: 10.1016/j.seh.2023.100004
    [105] Lukić Bilela L, Matijošytė I, Krutkevičius J, et al. (2023) Impact of per- and polyfluorinated alkyl substances (PFAS) on the marine environment: Raising awareness, challenges, legislation, and mitigation approaches under the One Health concept. Mar Pollut Bull 194: 115309. https://doi.org/10.1016/j.marpolbul.2023.115309 doi: 10.1016/j.marpolbul.2023.115309
    [106] Joerss H, Menger F (2023) The complex 'PFAS world' - How recent discoveries and novel screening tools reinforce existing concerns. Curr Opin Green Sustain Chem 40. https://doi.org/10.1016/j.cogsc.2023.100775 doi: 10.1016/j.cogsc.2023.100775
    [107] Ganesan S, Chawengkijwanich C, Gopalakrishnan M, et al. (2022) Detection methods for sub-nanogram level of emerging pollutants – Per and polyfluoroalkyl substances. Food and Chemical Toxicology 168: 113377. https://doi.org/10.1016/j.fct.2022.113377 doi: 10.1016/j.fct.2022.113377
    [108] Orazbayeva D, Muratuly A, Bektassov M, et al. (2022) Chromatographic determination of pesticides in soil: Current trends in analysis and sample preparation. Trends in Environmental Analytical Chemistry 35: e00174. https://doi.org/10.1016/j.teac.2022.e00174 doi: 10.1016/j.teac.2022.e00174
    [109] Fontanals N, Pocurull E, Borrull F, et al. (2021) Clean-up techniques in the pressurized liquid extraction of abiotic environmental solid samples. Trends in Environmental Analytical Chemistry 29. https://doi.org/10.1016/j.teac.2020.e00111 doi: 10.1016/j.teac.2020.e00111
    [110] Oflu S, Erarpat S, Zaman BT, et al. (2023) Quantification of trace fenuron in waste water samples by matrix matching calibration strategy and gas chromatography–mass spectrometry after simultaneous derivatization and preconcentration. Environ Monit Assess 195. https://doi.org/10.1007/s10661-023-11575-1 doi: 10.1007/s10661-023-11575-1
    [111] Hassan MdT-A, Chen X, Fnu PIJ, et al. (2024) Rapid detection of per- and polyfluoroalkyl substances (PFAS) using paper spray-based mass spectrometry. J Hazard Mater 465: 133366. https://doi.org/10.1016/j.jhazmat.2023.133366 doi: 10.1016/j.jhazmat.2023.133366
    [112] Kourtchev I, Hellebust S, Heffernan E, et al. (2022) A new on-line SPE LC-HRMS method for the analysis of Perfluoroalkyl and Polyfluoroalkyl Substances (PFAS) in PM2.5 and its application for screening atmospheric particulates from Dublin and Enniscorthy, Ireland. Science of The Total Environment 835: 155496. https://doi.org/10.1016/j.scitotenv.2022.155496 doi: 10.1016/j.scitotenv.2022.155496
    [113] Maciel-Silva FW, Viganó J, Castro LEN, et al. (2022) Pressurized liquid extraction coupled in-line with SPE and on-line with HPLC (PLE-SPExHPLC) for the recovery and purification of anthocyanins from SC-CO2 semi-defatted Aç aí (Euterpe oleracea). Food Research International 160. https://doi.org/10.1016/j.foodres.2022.111711 doi: 10.1016/j.foodres.2022.111711
    [114] Gurrani S, Prakasham K, Zii Ying JL, et al. (2023) A low-cost eco-friendly fast drug extraction (FaDEx) technique for environmental and bio-monitoring of psychoactive drug in urban water and sports-persons' urine samples. Environ Res 217. https://doi.org/10.1016/j.envres.2022.114787 doi: 10.1016/j.envres.2022.114787
    [115] van Leeuwen SPJ, de Boer J (2007) Extraction and clean-up strategies for the analysis of poly- and perfluoroalkyl substances in environmental and human matrices. J Chromatogr A 1153: 172–185. https://doi.org/10.1016/j.chroma.2007.02.069 doi: 10.1016/j.chroma.2007.02.069
    [116] Omidoyin KC, Jho EH (2024) Environmental occurrence and ecotoxicological risks of plastic leachates in aquatic and terrestrial environments. Science of the Total Environment 954. https://doi.org/10.1016/j.scitotenv.2024.176728 doi: 10.1016/j.scitotenv.2024.176728
    [117] Xie Z, Hu Y, Lin J, et al. (2023) Calix[4]arene-based covalent organic frameworks with host-guest recognition for selective adsorption of six per- and polyfluoroalkyl substances in food followed by UHPLC-MS/MS detection. J Hazard Mater 459: 132198. https://doi.org/10.1016/j.jhazmat.2023.132198 doi: 10.1016/j.jhazmat.2023.132198
    [118] Chow YN, Foo KY (2023) Insights into the per- and substances-contaminated paper mill processing discharge: Detection, phytotoxicity, bioaccumulative profiling, and health risk verification. J Clean Prod 384: 135478. https://doi.org/10.1016/j.jclepro.2022.135478 doi: 10.1016/j.jclepro.2022.135478
    [119] Al Amin Md, Sobhani Z, Chadalavada S, et al. (2020) Smartphone-based / Fluoro-SPE for selective detection of PFAS at ppb level. Environ Technol Innov 18: 100778. https://doi.org/10.1016/j.eti.2020.100778 doi: 10.1016/j.eti.2020.100778
    [120] G H, G S, R.S. R, et al. (2023) Early detection of emerging persistent perfluorinated alkyl substances (PFAS) along the east coast of India. Science of The Total Environment 902: 166155. https://doi.org/10.1016/j.scitotenv.2023.166155 doi: 10.1016/j.scitotenv.2023.166155
    [121] Choi H, Bae I-A, Choi JC, et al. (2018) Perfluorinated compounds in food simulants after migration from fluorocarbon resin-coated frying pans, baking utensils, and non-stick baking papers on the Korean market. Food Additives & Contaminants: Part B 11: 264–272. https://doi.org/10.1080/19393210.2018.1499677 doi: 10.1080/19393210.2018.1499677
    [122] Park N, Kho Y, Kim J (2021) Levels of Perfluorinated Compounds in Liquid Milk Products in Korea. Journal of Food Hygiene and Safety 36: 310–315. https://doi.org/10.13103/JFHS.2021.36.4.310 doi: 10.13103/JFHS.2021.36.4.310
    [123] Wang J, Wang H (2018) One-step fabrication of coating-free mesh with underwater superoleophobicity for highly efficient oil/water separation. Surf Coat Technol 340: 1–7. https://doi.org/10.1016/j.surfcoat.2018.02.036 doi: 10.1016/j.surfcoat.2018.02.036
    [124] Souza MCO, Saraiva MCP, Honda M, et al. (2020) Exposure to per- and polyfluorinated alkyl substances in pregnant Brazilian women and its association with fetal growth. Environ Res 187. https://doi.org/10.1016/j.envres.2020.109585 doi: 10.1016/j.envres.2020.109585
    [125] Chen MH, Ng S, Hsieh CJ, et al. (2017) The impact of prenatal perfluoroalkyl substances exposure on neonatal and child growth. Science of the Total Environment 607–608: 669–675. https://doi.org/10.1016/j.scitotenv.2017.06.273 doi: 10.1016/j.scitotenv.2017.06.273
    [126] Cao W, Liu X, Liu X, et al. (2018) Perfluoroalkyl substances in umbilical cord serum and gestational and postnatal growth in a Chinese birth cohort. Environ Int 116: 197–205. https://doi.org/10.1016/j.envint.2018.04.015 doi: 10.1016/j.envint.2018.04.015
    [127] Liu B, Lu X, Jiang A, et al. (2024) Influence of maternal endocrine disrupting chemicals exposure on adverse pregnancy outcomes: A systematic review and meta-analysis. Ecotoxicol Environ Saf 270. https://doi.org/10.1016/j.ecoenv.2023.115851 doi: 10.1016/j.ecoenv.2023.115851
    [128] Souza MCO, Saraiva MCP, Honda M, et al. (2020) Exposure to per- and polyfluorinated alkyl substances in pregnant Brazilian women and its association with fetal growth. Environ Res 187. https://doi.org/10.1016/j.envres.2020.109585 doi: 10.1016/j.envres.2020.109585
    [129] Kumara MK, Bhattacharyya D (2022) Per- and poly-fluoroalkyl substances (PFASs) in drinking water and related health effects. Current Developments in Biotechnology and Bioengineering: Sustainable Treatment Technologies for Per- and Poly-fluoroalkyl Substances 71–103. https://doi.org/10.1016/B978-0-323-99906-9.00016-4
    [130] Temkin AM, Hocevar BA, Andrews DQ, et al. (2020) Application of the Key Characteristics of Carcinogens to Per and Polyfluoroalkyl Substances. International Journal of Environmental Research and Public Health 2020, Vol 17, Page 1668 17: 1668. https://doi.org/10.3390/ijerph17051668
    [131] Girardi P, Merler E (2019) A mortality study on male subjects exposed to polyfluoroalkyl acids with high internal dose of perfluorooctanoic acid. Environ Res 179. https://doi.org/10.1016/j.envres.2019.108743 doi: 10.1016/j.envres.2019.108743
    [132] Yang Z, Liu H yu, Yang Q yun, et al. (2022) Associations between exposure to perfluoroalkyl substances and birth outcomes: A meta-analysis. Chemosphere 291. https://doi.org/10.1016/j.chemosphere.2021.132909 doi: 10.1016/j.chemosphere.2021.132909
    [133] UNEP Stockholm Convention (2019) UNEP, The New POPs under the Stockholm Convention, 2019. Available from: http://www.pops.int/SearchResults/tabid/37/Default.aspx?Search = PFOA http://www.pops.int/TheConvention/ThePOPs/TheNewPOPs/tabid/2511/Default.asp.
    [134] Sinclair GM, Long SM, Jones OAH (2020) What are the effects of PFAS exposure at environmentally relevant concentrations? Chemosphere 258: 127340. https://doi.org/10.1016/j.chemosphere.2020.127340 doi: 10.1016/j.chemosphere.2020.127340
    [135] Wanninayake DM (2021) Comparison of currently available PFAS remediation technologies in water: A review. J Environ Manage 283: 111977. https://doi.org/10.1016/j.jenvman.2021.111977 doi: 10.1016/j.jenvman.2021.111977
    [136] Brendel S, Fetter É, Staude C, et al. (2018) Short-chain perfluoroalkyl acids: environmental concerns and a regulatory strategy under REACH. Environ Sci Eur 30: 9. https://doi.org/10.1186/s12302-018-0134-4 doi: 10.1186/s12302-018-0134-4
    [137] Brusseau ML (2018) Assessing the potential contributions of additional retention processes to PFAS retardation in the subsurface. Science of The Total Environment 613–614: 176–185. https://doi.org/10.1016/j.scitotenv.2017.09.065 doi: 10.1016/j.scitotenv.2017.09.065
  • Reader Comments
  • © 2025 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(864) PDF downloads(142) Cited by(0)

Figures and Tables

Figures(6)  /  Tables(1)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog