Cardiac arrest (CA) is a fatal acute event. The development of new CA early warning system based on time series of vital signs from electronic health records (EHR) has great potential to reduce CA damage. In this process, recursive architecture-based deep learning, as a powerful tool for time series data processing, enables automatically extract features from various monitoring clinical parameters and to further improve the performance for acute critical illness prediction. However, the unexplainable nature and excessive time caused by black box structure with poor parallelism are the limitations of its development, especially in the CA clinical application with strict requirement of emergency treatment and low hidden dangers. In this study, we present an explainable and efficient deep early warning system for CA prediction, which features are captured by an efficient temporal convolutional network (TCN) on EHR clinical parameters sequence and explained by deep Taylor decomposition (DTD) theoretical framework. To demonstrate the feasibility of our method and further evaluate its performance, prediction and explanation experiments were performed. Experimental results show that our method achieves superior CA prediction accuracy compared with standard national early warning score (NEWS), in terms of overall AUROC (0.850 Vs. 0.476) and F1-Score (0.750 Vs. 0.450). Furthermore, our method improves the interpretability and efficiency of deep learning-based CA early warning system. It provides the relevance of prediction results for each clinical parameter and about 1.7 times speed enhancement for system calculation compared with the long short-term memory network.
Citation: Qinhua Tang, Xingxing Cen, Changqing Pan. Explainable and efficient deep early warning system for cardiac arrest prediction from electronic health records[J]. Mathematical Biosciences and Engineering, 2022, 19(10): 9825-9841. doi: 10.3934/mbe.2022457
[1] | Mario Coccia, Igor Benati . Negative effects of high public debt on health systems facing pandemic crisis: Lessons from COVID-19 in Europe to prepare for future emergencies. AIMS Public Health, 2024, 11(2): 477-498. doi: 10.3934/publichealth.2024024 |
[2] | Yosef Mohamed-Azzam Zakout, Fayez Saud Alreshidi, Ruba Mustafa Elsaid, Hussain Gadelkarim Ahmed . The magnitude of COVID-19 related stress, anxiety and depression associated with intense mass media coverage in Saudi Arabia. AIMS Public Health, 2020, 7(3): 664-678. doi: 10.3934/publichealth.2020052 |
[3] | Pamila Sadeeka Adikari, KGRV Pathirathna, WKWS Kumarawansa, PD Koggalage . Role of MOH as a grassroots public health manager in preparedness and response for COVID-19 pandemic in Sri Lanka. AIMS Public Health, 2020, 7(3): 606-619. doi: 10.3934/publichealth.2020048 |
[4] | Srikanth Umakanthan, Anuradha Chauhan, Madan Mohan Gupta, Pradeep Kumar Sahu, Maryann M Bukelo, Vijay Kumar Chattu . COVID-19 pandemic containment in the Caribbean Region: A review of case-management and public health strategies. AIMS Public Health, 2021, 8(4): 665-681. doi: 10.3934/publichealth.2021053 |
[5] | Vasiliki Georgousopoulou, Panagiota Pervanidou, Pantelis Perdikaris, Efrosyni Vlachioti, Vaia Zagana, Georgios Kourtis, Ioanna Pavlopoulou, Vasiliki Matziou . Covid-19 pandemic? Mental health implications among nurses and Proposed interventions. AIMS Public Health, 2024, 11(1): 273-293. doi: 10.3934/publichealth.2024014 |
[6] | Akari Miyazaki, Naoko Kumada Deguchi, Tomoko Omiya . Difficulties and distress experienced by Japanese public health nurses specializing in quarantine services when dealing with COVID-19: A qualitative study in peri-urban municipality. AIMS Public Health, 2023, 10(2): 235-251. doi: 10.3934/publichealth.2023018 |
[7] | Abdulqadir J. Nashwan, Rejo G. Mathew, Reni Anil, Nabeel F. Allobaney, Sindhumole Krishnan Nair, Ahmed S. Mohamed, Ahmad A. Abujaber, Abbas Balouchi, Evangelos C. Fradelos . The safety, health, and well-being of healthcare workers during COVID-19: A scoping review. AIMS Public Health, 2023, 10(3): 593-609. doi: 10.3934/publichealth.2023042 |
[8] | Sameer Badri Al-Mhanna, Alexios Batrakoulis, Abdulrahman M. Sheikh, Abdulaziz A. Aldayel, Abdulwali Sabo, Mahaneem Mohamed, Hafeez Abiola Afolabi, Abdirizak Yusuf Ahmed, Sahra Isse Mohamed, Mehmet Gülü, Wan Syaheedah Wan Ghazali . Impact of COVID-19 lockdown on physical activity behavior among students in Somalia. AIMS Public Health, 2024, 11(2): 459-476. doi: 10.3934/publichealth.2024023 |
[9] | Nguyen Tuan Hung, Vu Thu Trang, Trinh Van Tung, Nguyen Xuan Long, Ha Thi Thu, Tran Song Giang, Tran Hoang Thi Diem Ngoc, Vu Thi Thanh Mai, Nguyen Kim Oanh, Nguyen Thi Phuong, Nguyen Hang Nguyet Van, Nguyen Hanh Dung, Pham Tien Nam . COVID-19-related music-video-watching among the Vietnamese population: lessons on health education. AIMS Public Health, 2021, 8(3): 428-438. doi: 10.3934/publichealth.2021033 |
[10] | Dimitris Zavras . A cross-sectional population-based study on the influence of the COVID-19 pandemic on incomes in Greece. AIMS Public Health, 2021, 8(3): 376-387. doi: 10.3934/publichealth.2021029 |
Cardiac arrest (CA) is a fatal acute event. The development of new CA early warning system based on time series of vital signs from electronic health records (EHR) has great potential to reduce CA damage. In this process, recursive architecture-based deep learning, as a powerful tool for time series data processing, enables automatically extract features from various monitoring clinical parameters and to further improve the performance for acute critical illness prediction. However, the unexplainable nature and excessive time caused by black box structure with poor parallelism are the limitations of its development, especially in the CA clinical application with strict requirement of emergency treatment and low hidden dangers. In this study, we present an explainable and efficient deep early warning system for CA prediction, which features are captured by an efficient temporal convolutional network (TCN) on EHR clinical parameters sequence and explained by deep Taylor decomposition (DTD) theoretical framework. To demonstrate the feasibility of our method and further evaluate its performance, prediction and explanation experiments were performed. Experimental results show that our method achieves superior CA prediction accuracy compared with standard national early warning score (NEWS), in terms of overall AUROC (0.850 Vs. 0.476) and F1-Score (0.750 Vs. 0.450). Furthermore, our method improves the interpretability and efficiency of deep learning-based CA early warning system. It provides the relevance of prediction results for each clinical parameter and about 1.7 times speed enhancement for system calculation compared with the long short-term memory network.
Subphthalocyanine is an aromatic bowl-shaped macrocyclic ligand that exclusively chelates a boron atom (BsubPc,
Recently, our group synthesized and characterized the first BsubPc containing polymer using a post-polymerization coupling reaction between bromo-BsubPc (Br-BsubPc, 1,
In our previous studies, the BsubPc molecule was incorporated pendent to the main chain of a copolymer [8,9,10]. The dispersity of the starting copolymers was kept low in each case by using nitroxide mediated polymerization (NMP), a known form of controlled radical polymerization (CRP) [14,15,16]. Originally, NMP was restricted to the homopolymerization of styrenics when using first-generation stable free radicals (SFRs) such as 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) [14,15,16]. However, the development of second-generation SFRs such as N-tert-butyl-N-[1-diethylphosphono-(2,2-dimethylpropyl)] nitroxide (SG1) [18] and 2,2,5-trimethyl-4-phenyl-3-azahexane-3-nitroxide (TIPNO) [19] have allowed the extension of NMP methodology to the homopolymerization of other monomers including acrylates [20,21], acrylamides [14,22] and methacrylates (when using a small amount of a comonomer [23,24,25]). Through the appropriate selection of initiator and/or SFR, NMP produces polymers with a high level of chain end predictability, which makes targeted chain-end reactions possible [14,15,16]. NMP methods have been previously used to functionalize both the ω- or the α-chain end of a polymer with a variety of functional groups. For example, uracil α-chain end has been functionalized using TEMPO- and SG1-based alkoxyamine unimolecular initiators having been purposefully synthesized to study polymer diffusion [26] and to modify the melt viscosity. [27] The use of commercially available SG1-based alkoxyamine, 2-methyl-2-[N-tert-butyl-N-(1-diethoxyphosphoryl-2,2-dimethylpropyl) aminoxy] propionic acid (BlocBuilder-MA [28,29,30],
In this paper we illustrate how the BsubPc chromophore can be selectively introduced to either the ω- or the α- chain end of a polymer produced by NMP utilizing BlocBuilder-MA as the initiating species. This approach will facilitate the functionalization of polymer chain ends in later studies for the synthesis of BsubPc containing polymers for use as the active material in OLEDs and OPVs.
4-Hydroxy-2,2,6,6-tetramethylpiperidine 1-oxyl (4-hydroxy-TEMPO, 97%) was obtained from Aldrich and used as received. Bromo boron subphtathalocyanine (Br-BsubPc) was synthesized according to our previously reported method [42]. Common solvents and reagents were obtained from Caledon Laboratories (Caledon, Ontario, Canada) or Sigma-Aldrich Co. (Ontario, Canada) and used as received. BlocBuilder-MA was graciously donated by Prof. Milan Marić (McGill University), who obtained it from Arkema Inc. (USA).
In all cases the homopolymers were synthesized under identical experimental setup and similar formulations (Table 1). For example, PS1 was synthesized by combining styrene (5.50 g, 52.9 mmol) and BlocBuilder-MA (0.10 g, 0.26) with a Teflon stir bar in a 50mL three neck round bottom glass flask prior to sealing with rubber septa. The mixture was then bubbled with nitrogen for 20 min and the mixture was heated to 90 °C while maintaining a light nitrogen purge. Once the stirring of the mixture became labored, due to the increase in viscosity, the reaction was allowed to cool to room temperature while maintaining the nitrogen purge (in this example approximately 20 h). Once cooled the mixture was precipitated in methanol, filtered and dried in a vacuum oven at 60 °C overnight. Yield: 1.92 g (35%), Mn = 7.7 kg·mol−1, Mw/Mn = 1.37.
Exp. IDa | [BlocBuilder]0 (mol·L-1) | [S]0 (mol L-1) | Mn Target (kg·mol-1) | Temp. (°C) | tpolym (h) | Mnb (kg·mol-1) | Mw/Mnb |
PS1 | 0.044 | 8.74 | 20.8 | 90 | 20.0 | 7.7 | 1.37 |
PS1Tc | – | – | – | 120 | 4.0 | 7.6 | 1.31 |
PS1-αBsubPc | – | – | – | 125 | 40 | 7.6 | 1.30 |
PS1T-αBsubPc | – | – | – | 125 | 40 | 7.1 | 1.31 |
PS1-ωBsubPc | – | – | – | 100 | 0.5 | 8.0 | 1.30 |
BA1 | 0.056 | 4.83 | 15.9 | 115 | 2.0 | 22.2 | 1.47 |
BA1-αBsubPc | – | – | – | 125 | 40 | 23.2 | 1.45 |
BA1-ωBsubPc | – | – | – | 100 | 0.5 | 23.2 | 1.45 |
a Experimental identification (Exp. ID) for the synthesis of poly(styrene) (PS) are given by PS1-Y and for the synthesis of n-butyl acryalte (BA) is given by BA1-Y, Y representing post polymerization reaction . All polymerizations were done in bulk. Y = αBsubPc is the coupling of (1) by reaction in chlorobenzene and Y = ωBsubPc is the coupling of (2) by reaction in toluene. | |||||||
b Molecular weight characterization was determined by GPC using PS standards. | |||||||
c The PST is PS after being treated to thiophenol in toluene for 4 h for the removal of the SG1 group. |
To an oven dried round bottom flask a mixture of Br-BsubPc (1.00 g, 2.11 mmol) and 4-hydroxy-TEMPO (0.54 g, 3.16 mmol) and toluene (≈ 4 mL) were added prior to sealing under nitrogen. The mixture was then heated to reflux for 20 hours. Once cooled to room temperature the organic phase was washed 3 times using 1 M KOH solution and 3 times using neutral water. Once dried, the crude product was purified by column chromatography using 50 vol% ethyl acetate to hexanes as the eluent. Once dried the resulting magenta powder was determined to be BsubPc-TEMPO. Yield = 0.40 g (33%). 1H NMR: 8.8 ppm (6H), 7.9 ppm (6H), 2.8 ppm (1H), 2.0 ppm (4H), 1.4 ppm (12H). 11B NMR: −14.1 ppm (1B) HR-MS (DART) calcd for C33H29BBrN6 ([M]+): m/z 566.2476, found 566.2479.
In the case of BsubPc coupling to the α-chain end of PS1, a 1:1.2 molar ratio of the PS1 to Br-BsubPc was dissolved in toluene then heated to 115 °C for 20 h under a nitrogen purge or blanket. Once cooled to room temperature the polymer was precipitated in methanol and collected by filtration. The crude polymer was then dried in a vacuum oven overnight at 60 °C. In the case of BsubPc coupling to the ω-chain end of the PS homopolymers a 1:1.2 molar ratio of the PS homopolymer to BsubPc-TEMPO was dissolved in toluene followed by a 20 min bubbling and purging with nitrogen. The mixture was then heated to 100 °C for 30 min. Once cooled to room temperature the nitrogen purge was removed and the polymer was precipitated in methanol and collected by filtration.
1H NMR spectroscopy that was run on a 400 MHz Varian Mercury spectrometer at 23 °C in deuterated chloroform (CDCl3, with Me4Si) obtained from Cambridge Isotopes Laboratory, Inc.. Molecular weight characterization and determination of BsubPc coupling was performed by gel permeation chromatography (GPC), using narrow molecular weight distribution poly(styrene) (PS) standards and two Waters Styragel 5 μm, HR 4E 7.8 × 300 mm column in series. The GPC was equipped with a Waters 2695 separation module, a Waters 2998 photodiode array (PDA detector) and a Waters 2414 refractive index detector (RI Detector) and HPLC grade THF was used as the mobile phase at a flow rate of 1.2 mL·min−1. Single crystal X-ray diffraction data was collected on a Bruker Kappa APEX-DUO diffractometer using a Copper ImuS (microsource) tube with multi-layer optics and were measured using a combination of f scans and w scans. The data was then processed using APEX2 and SAINT and the absorption correction was carried out using SADABS [43]. The structure was then solved and refined using SHELXTL [44] for full-matrix least-squares refinement that was based on F2.
We began by synthesizing poly(styrene) homopolymers (PS1) using BlocBuilder-MA as the initiating species (
Br-BsubPc contains a highly reactive Br-B bond which in the presence of a carboxylic acid group results in the facile coupling of the BsubPc chromophore to the carboxylic acid group [9]. When using BlocBuilder-MA as an NMP initiating species, the PS chains have a carboxylic acid group present on the α-chain end (
To place the BsubPc chromophore at the ω-chain end we took the strategy of removing or replacing the SFR fragment [38,39,40,41]. It is well known that the equilibrium constant of TEMPO is significantly lower than that of SG1 and that substitution of the SG1 fragment on a polymer (polymer that is reversibly terminated with SG1) for a TEMPO group is possible at relatively low temperatures (≈ 70 to 110 °C) [38,39,40,41]. Therefore, a BsubPc derivative-containing TEMPO (BsubPc-TEMPO, 2,
The synthesis of BsubPc-TEMPO was achieved by reaction of Br-BsubPc with 4-hydroxy-2,2,6,6-tetramethylpiperidin-1-oxyl (
In attempts to find an appropriate synthetic operating window for BsubPc-TEMPO, we decided to explore the synthesis of both reactive sites. Therefore we performed two independent syntheses, under the same reaction conditions (chlorobenzene, 112 °C), and measured the relative yield with time. TEMPO (not 4-hydroxy-TEMPO) and cyclohexanol were both independently reacted with Br-BsubPc in a 5:1 molar excess (Figure 4) to compare the relative kinetics. The kinetics was tracked by comparing the relative area under the curve of the HPLC chromatogram obtained using a mixture of 80/20 volume ratio of acetonitrile and DMF. We observed that the reaction of cyclohexanol (which we assumed has a similar reactivity to the hydroxyl group on 4-hydrox-TEMPO) was much faster than the reaction of Br-BsubPc with TEMPO (Figure 4). Therefore the reaction of 4-hydrox-TEMPO with Br-BsubPc was time controlled to a couple hours to prevent/significantly reduce the formation of the dimeric side product. We did investigate the possibility that the BsubPc coupled with TEMPO could be used as a unimolecular initiator for styrene polymerization but at 125 °C no polymerization was noted and the compound degraded. A single crystal of the cyclohexanol BsubPc derivative was also grown by evaporation from DCM and diffracted (CCD deposition # 968575). The 50% probability thermal ellipsoid plot can be found in Figure 4b. Detailed crystallographic information can be found in Table S7-11 in the supporting information.
BsubPc-TEMPO was allowed to react at 90 °C for 30 min with PS1 (terminated with SG1) to produce PS1-ωBsubPc. A temperature of 90 °C was sufficient to facilitate the replacement of the SG1 fragment for the BsubPc-TEMPO fragment as indicated by the GPC chromatogram obtained from the PDA detector set at 565 nm (Figure 2B). We determined that if the reaction was allowed to proceed for longer than 30 min the concentration of irreversible terminated chains increased resulting in a decrease in BsubPc-TEMPO terminated chains. For PS1 very little change in their apparent mass due to the addition of BsubPc-TEMPO was noted based on RI detection (Figure 2A). The reader should also note that our GPC is set up with a PDA detector first and in line with a RI detector. The length of tubing between the two is approximately 60 cm which accounts for the differences in retention time seen between chromatograms extracted from the PDA and RI detectors (Figure 2B).
To further explore the application of this technique and the ability of BsubPc reagents to couple to the chain ends, n-butyl acrylate was homopolymerized under similar conditions using BlocBuilder-MA (BA1) and again both chain ends were coupled with BsubPc (
BsubPc have been reported to have extremely high extinction coefficient (65 000-75 000 L mol−1 cm−1) [45] thus their detection even within polymers of very high molecular weight is or would be easy. These features would make BsubPc containing polymers ideal candidates for dyes-labeled polymers [46]. When considering Figure 2B and 5B it is apparent that the GPC traces corresponding to the polymers functionalized with BsubPc on the α-chain end (initiating side) and those with BsubPc on the ω-chain end (propagation side) are not identical and that the PDA traces differ from each other and the corresponding RI traces of the unreacted parent homopolymers. We can offer several hypotheses or explanations. Firstly is that the RI detector response is based on mass concentration while the PDA detector response is based on molar concentration basis. This difference means that when a BsubPc chromophore is coupled to a small chain, the response is more significant as observed by the PDA detector, then when coupled to a larger chain. This amplification in detector response for the smaller polymer chains could potentially be utilized to shed insight on the nature of the chain ends and ultimately it could be used to comment on the degree of irreversible termination which has occurred in the initial polymer synthesis. For example, Scott et al. have previously demonstrated that they could quantified the chain end livingness once they reversibly end-capped a series of polymers with a fluorescent-TEMPO SFR followed by the comparison of the fluorescence spectra and the RI trace obtained when running a GPC [47]. However in our case the same polymer sample is coupled with a BsubPc chromophore selectively on either end and therefore could potentially give insight on the nature of both the propagating and the initiating end of the polymer chains. For example, when comparing the scaled GPC chromatograms for PS1 reacted with Br-BsubPc (PS1-αBsubPc, Figure 2) to that reacted with BsubPc-TEMPO (PS1-ωBsubPc,
The difference in GPC traces acquired by the RI detector and the PDA detector (Figure 2 and 5) can give an illustration of the amount of irreversible termination imparted to the system when using BlocBuilder. It is, however, crucial to note that styrene has a tendency to auto-initiate [50,51] and therefore in this simple system there may be a portion of chains which are not initiated by BlocBuilder-MA but do contain the nitroxide fragment at their ω-ends, making this analysis that we present here non quantitative. Regardless, our experiments show that by derivatizing a polymer at each end using the same chromophore (such as BsubPc), which has a very high molar extinction coefficient, an amplification in detector response is observed for the smaller chains, which can give insight on the degree of irreversible termination of the sample.
This study illustrates a novel and selective method for functionalizing each end of a polymer chain produced by NMP with a BsubPc chromophore. By synthesizing polymers using BlocBuilder-MA as the initiating species and using two separate and chemically selective BsubPc reagents, Br-BsubPc and TEMPO-BsubPc we can selectively couple a BsubPc chromophore to either the initiating chain end, α-chain end, or the propagating chain end, w-chain end, respectively. Two generic model polymers were used to illustrate that this technique is versatile, poly(styrene) and poly(n-butyl acrylate). In both cases the coupling was observed using a GPC equipped with a PDA detector in line with an RI detector. In this paper we illustrate how the BsubPc chromophore can be selectively introduced to either the ω- or the α- chain end of a polymer produced by NMP utilizing BlocBuilder-MA as the initiating species. This method will facilitate the functionalization of polymer chain ends resulting in BsubPc polymers with BsubPc chromophores both pendent to the main chain [10] and at both chain ends. Future studies will utilize this technique to investigate the effect of BsubPc end -group coupling on OLED device performance.
We would like to thank Natural Science and Engineering Research Council (NSERC) of Canada Discovery Grant program for financial support. BL would like to thank the government of Canada for the Banting Post Doctoral Fellowship. We thank Alan Lough (University of Toronto) for the single crystal x-ray diffractions. We also thank Prof. Milan Maric from McGill University (Quebec, Canada) for graciously donating BlocBuilder-MA (originally obtained from Arkema).
Electronic SupplementaryInformation (ESI) available: X-ray crystalography data; proof of reproducibility of GPC chromatograms.
[1] |
S. Girotra, B. K. Nallamothu, J. A. Spertus, Y. Li, H. M. Krumholz, P. S. Chan, Trends in survival after in-hospital cardiac arrest, N. Engl. J. Med., 367 (2012), 1912-1920. https://doi.org/10.1016/j.jemermed.2013.02.007 doi: 10.1016/j.jemermed.2013.02.007
![]() |
[2] |
L. Mandigers, F. Termorshuizen, N. F. Keizer, D. Gommers, D. Reis Miranda, W. J. Rietdijk, et al., A nationwide overview of 1-year mortality in cardiac arrest patients admitted to intensive care units in the Netherlands between 2010 and 2016, Resuscitation, 147 (2020), 88-94. https://doi.org/10.1016/j.resuscitation.2019.12.029 doi: 10.1016/j.resuscitation.2019.12.029
![]() |
[3] |
J. Soar, J. P. Nolan, B. W. Böttiger, G. D. Perkins, C. Lott, N. I. Nikolaou, et al., European resuscitation council guidelines for resuscitation 2015: section 3. Adult advanced life support, Resuscitation, 95 (2015), 100-147. https://doi.org/10.1016/j.resuscitation.2015.07.016 doi: 10.1016/j.resuscitation.2015.07.016
![]() |
[4] | G. B. Smith, D. R. Prytherch, P. Meredith, P. E. Schmidt, P. I. Featherstone, The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death, Resuscitation, 84 (2013), 465-470. https://doi.org/10.1016/j.resuscitation.2012.12.016 |
[5] |
I. Nishijima, S. Oyadomari, S. Maedomari, R. Toma, C. Igei, S. Kobata, et al., Use of a modified early warning score system to reduce the rate of in-hospital cardiac arrest, J. Intensive Care, 4 (2016), 1-6. https://doi.org/10.1186/s40560-016-0134-7 doi: 10.1186/s40560-016-0134-7
![]() |
[6] |
F. E. Shamout, T. Zhu, P. Sharma, P. J. Watkinson, D. A. Clifton, Deep interpretable early warning system for the detection of clinical deterioration, IEEE J. Biomed. Heal. Inf., 24 (2019), 437-446. https://doi.org/10.1109/jbhi.2019.2937803 doi: 10.1109/jbhi.2019.2937803
![]() |
[7] |
J. Kwon, Y. Lee, Y. Lee, S. Lee, J. Park, An algorithm based on deep learning for predicting in‐hospital cardiac arrest, J. Am. Heart Assoc., 7 (2018), 1-11. https://doi.org/10.1161/jaha.118.008678 doi: 10.1161/jaha.118.008678
![]() |
[8] |
J. Kim, M. Chae, H. J. Chang, Y. A. Kim, E. Park, Predicting cardiac arrest and respiratory failure using feasible artificial intelligence with simple trajectories of patient data, J. Clin. Med., 8 (2019), 1336-1350. https://doi.org/10.3390/jcm8091336 doi: 10.3390/jcm8091336
![]() |
[9] |
T. Pham, T. Tran, D. Phung, S. Venkatesh, Predicting healthcare trajectories from medical records: A deep learning approach, J. Biomed. Inf., 69 (2017), 218-229. https://doi.org/10.1016/j.jbi.2017.04.001 doi: 10.1016/j.jbi.2017.04.001
![]() |
[10] |
S. M. Lauritsen, M. E. Kalør, E. L. Kongsgaard, K. M. Lauritsen, M. J. Jørgensen, J. Lange, et al., Early detection of sepsis utilizing deep learning on electronic health record event sequences, Artif. Intell. Med., 104 (2020), 101820. https://doi.org/10.1016/j.artmed.2020.101820 doi: 10.1016/j.artmed.2020.101820
![]() |
[11] | M. Aczon, D. Ledbetter, L. Ho, A. Gunny, A. Flynn, J. Williams, et al., Dynamic mortality risk predictions in pediatric critical care using recurrent neural networks, preprint, arXiv: 1701.06675. |
[12] |
H. Harutyunyan, H. Khachatrian, D. C. Kale, G. Ver Steeg, A. Galstyan, Multitask learning and benchmarking with clinical time series data, Sci. Data, 6 (2019), 1-18. https://doi.org/10.1038/s41597-019-0103-9 doi: 10.1038/s41597-019-0103-9
![]() |
[13] |
S. Hong, S. Lee, J. Lee, W. Cha, K. Kim, Prediction of cardiac arrest in the emergency department based on machine learning and sequential characteristics: model development and retrospective clinical validation study, JMIR Med. Inf., 8 (2020), 1-14. https://doi.org/10.2196/15932 doi: 10.2196/15932
![]() |
[14] |
Y. J. Lee, K Cho, O. Kwon, H. Park, Y. Lee, J. Kwon, et al., A multicentre validation study of the deep learning-based early warning score for predicting in-hospital cardiac arrest in patients admitted to general wards, Resuscitation, 163 (2021), 78-85. https://doi.org/10.1016/j.resuscitation.2021.04.013 doi: 10.1016/j.resuscitation.2021.04.013
![]() |
[15] |
S. J. Park, K. Cho, O. Kwon, H. Park, Y. Lee, W. H. Shim, et al., Development and validation of a deep-learning-based pediatric early warning system: A single-center study, Biomed. J., 45 (2022), 155-168. https://doi.org/10.1016/j.bj.2021.01.003 doi: 10.1016/j.bj.2021.01.003
![]() |
[16] | M. Moor, M. Horn, B. Rieck, D. Roqueiro, K. Borgwardt, Temporal convolutional networks and dynamic time warping can drastically improve the early prediction of sepsis, preprint, arXiv: 1902.01659. |
[17] | S. Bai, J. Z. Kolter, V. Koltun, An empirical evaluation of generic convolutional and recurrent networks for sequence modeling, preprint, arXiv: 1803.01271. |
[18] | Y. Chang, J. Rubin, G. Boverman, S. Vij, A. Rahman, A. Natarajan, et al., A multi-task imputation and classification neural architecture for early prediction of sepsis from multivariate clinical time series, in 2019 Computing in Cardiology (CinC), (2019), 1-4. https://doi.org/10.23919/CinC49843.2019.9005751 |
[19] |
N. Sato, E. Uchino, R. Kojima, S. Hiragi, M. Yanagita, Y. Okuno, Prediction and visualization of acute kidney injury in intensive care unit using one-dimensional convolutional neural networks based on routinely collected data, Comput. Methods Programs Biomed., 206 (2021), 106129. https://doi.org/10.1016/j.cmpb.2021.106129 doi: 10.1016/j.cmpb.2021.106129
![]() |
[20] | O. Almqvist, A comparative study between algorithms for time series forecasting on customer prediction: An investigation into the performance of ARIMA, RNN, LSTM, TCN and HMM, 2019. Available from: https://www.researchgate.net/publication/333731678 |
[21] |
G. Montavon, S. Lapuschkin, A. Binder, W. Samek, K. R. Müller, Explaining nonlinear classification decisions with deep taylor decomposition, Pattern Recognit., 65 (2017), 211-222. https://doi.org/10.1016/j.patcog.2016.11.008 doi: 10.1016/j.patcog.2016.11.008
![]() |
[22] |
C. Xiao, E. Choi, J. Sun, Opportunities and challenges in developing deep learning models using electronic health records data: a systematic review, J. Am. Med. Inf. Assoc., 25 (2018), 1419-1428. https://doi.org/10.1093/jamia/ocy068 doi: 10.1093/jamia/ocy068
![]() |
[23] |
B. Shickel, P. J. Tighe, A. Bihorac, P. Rashidi, Deep EHR: A survey of recent advances in deep learning techniques for electronic health record (EHR) analysis, IEEE J. Biomed. Heal. Inf., 22 (2017), 1589-1604. https://doi.org/10.1109/jbhi.2017.2767063 doi: 10.1109/jbhi.2017.2767063
![]() |
[24] |
W. Samek, G. Montavon, S. Lapuschkin, C. J. Anders, K. R. Müller, Explaining deep neural networks and beyond: A review of methods and applications, Proc. IEEE, 109 (2021), 247-278. https://doi.org/10.1109/jproc.2021.3060483 doi: 10.1109/jproc.2021.3060483
![]() |
[25] | H. Chefer, S. Gur, L. Wolf, Transformer interpretability beyond attention visualization, in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), (2021), 782-791. https://doi.org/10.1109/cvpr46437.2021.00084 |
[26] | W. J. Nam, S. Gur, J. Choi, L. Wolf, S. W. Lee, Relative attributing propagation: Interpreting the comparative contributions of individual units in deep neural networks, in Proceedings of the AAAI Conference on Artificial Intelligence, 34 (2020), 2501-2508. https://doi.org/10.1609/aaai.v34i03.5632 |
[27] | S. Gur, A. Ali, L. Wolf, Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization, preprint, arXiv: 2012.02166. |
[28] |
S. M. Lauritsen, M. Kristensen, M. V. Olsen, M. S. Larsen, K. M. Lauritsen, M. J. Jørgensen, et al., Explainable artificial intelligence model to predict acute critical illness from electronic health records, Nat. Commun., 11 (2020), 1-11. https://doi.org/10.1038/s41467-020-17431-x doi: 10.1038/s41467-020-17431-x
![]() |
1. | N.P.S. Chauhan, N.S. Hosmane, M. Mozafari, Boron-based polymers: opportunities and challenges, 2019, 14, 24685194, 100184, 10.1016/j.mtchem.2019.08.003 | |
2. | Qiankun Zhu, Xiaoming Li, Yonghua Xiao, Yan Xiong, Suiping Wang, Changli Xu, Jun Zhang, Xuewen Wu, Synthesis of Molecularly Imprinted Polymer via Visible Light Activated RAFT Polymerization in Aqueous Media at Room Temperature for Highly Selective Electrochemical Assay of Glucose, 2017, 218, 10221352, 1700141, 10.1002/macp.201700141 | |
3. | Daniele Vinciguerra, Stéphanie Denis, Julie Mougin, Merel Jacobs, Yohann Guillaneuf, Simona Mura, Patrick Couvreur, Julien Nicolas, A facile route to heterotelechelic polymer prodrug nanoparticles for imaging, drug delivery and combination therapy, 2018, 286, 01683659, 425, 10.1016/j.jconrel.2018.08.013 | |
4. | Kacper Wojtkiewicz, Alan Lough, Timothy P. Bender, Analysis of the solvent effects on the crystal growth of peripherally chlorinated boron subphthalocyanines, 2022, 24, 1466-8033, 2791, 10.1039/D1CE01320C | |
5. | Giulia Lavarda, Jorge Labella, M. Victoria Martínez-Díaz, M. Salomé Rodríguez-Morgade, Atsuhiro Osuka, Tomás Torres, Recent advances in subphthalocyanines and related subporphyrinoids, 2022, 51, 0306-0012, 9482, 10.1039/D2CS00280A |
Exp. IDa | [BlocBuilder]0 (mol·L-1) | [S]0 (mol L-1) | Mn Target (kg·mol-1) | Temp. (°C) | tpolym (h) | Mnb (kg·mol-1) | Mw/Mnb |
PS1 | 0.044 | 8.74 | 20.8 | 90 | 20.0 | 7.7 | 1.37 |
PS1Tc | – | – | – | 120 | 4.0 | 7.6 | 1.31 |
PS1-αBsubPc | – | – | – | 125 | 40 | 7.6 | 1.30 |
PS1T-αBsubPc | – | – | – | 125 | 40 | 7.1 | 1.31 |
PS1-ωBsubPc | – | – | – | 100 | 0.5 | 8.0 | 1.30 |
BA1 | 0.056 | 4.83 | 15.9 | 115 | 2.0 | 22.2 | 1.47 |
BA1-αBsubPc | – | – | – | 125 | 40 | 23.2 | 1.45 |
BA1-ωBsubPc | – | – | – | 100 | 0.5 | 23.2 | 1.45 |
a Experimental identification (Exp. ID) for the synthesis of poly(styrene) (PS) are given by PS1-Y and for the synthesis of n-butyl acryalte (BA) is given by BA1-Y, Y representing post polymerization reaction . All polymerizations were done in bulk. Y = αBsubPc is the coupling of (1) by reaction in chlorobenzene and Y = ωBsubPc is the coupling of (2) by reaction in toluene. | |||||||
b Molecular weight characterization was determined by GPC using PS standards. | |||||||
c The PST is PS after being treated to thiophenol in toluene for 4 h for the removal of the SG1 group. |
Exp. IDa | [BlocBuilder]0 (mol·L-1) | [S]0 (mol L-1) | Mn Target (kg·mol-1) | Temp. (°C) | tpolym (h) | Mnb (kg·mol-1) | Mw/Mnb |
PS1 | 0.044 | 8.74 | 20.8 | 90 | 20.0 | 7.7 | 1.37 |
PS1Tc | – | – | – | 120 | 4.0 | 7.6 | 1.31 |
PS1-αBsubPc | – | – | – | 125 | 40 | 7.6 | 1.30 |
PS1T-αBsubPc | – | – | – | 125 | 40 | 7.1 | 1.31 |
PS1-ωBsubPc | – | – | – | 100 | 0.5 | 8.0 | 1.30 |
BA1 | 0.056 | 4.83 | 15.9 | 115 | 2.0 | 22.2 | 1.47 |
BA1-αBsubPc | – | – | – | 125 | 40 | 23.2 | 1.45 |
BA1-ωBsubPc | – | – | – | 100 | 0.5 | 23.2 | 1.45 |
a Experimental identification (Exp. ID) for the synthesis of poly(styrene) (PS) are given by PS1-Y and for the synthesis of n-butyl acryalte (BA) is given by BA1-Y, Y representing post polymerization reaction . All polymerizations were done in bulk. Y = αBsubPc is the coupling of (1) by reaction in chlorobenzene and Y = ωBsubPc is the coupling of (2) by reaction in toluene. | |||||||
b Molecular weight characterization was determined by GPC using PS standards. | |||||||
c The PST is PS after being treated to thiophenol in toluene for 4 h for the removal of the SG1 group. |