A hybrid CNN-LSTM model with adaptive instance normalization for one shot singing voice conversion

Assila Yousuf; David Solomon George; Assila Yousuf; David Solomon George

doi:10.3934/electreng.2024013

AIMS Electronics and Electrical Engineering

2024, Volume 8, Issue 3: 292-310. doi: 10.3934/electreng.2024013

Previous Article Next Article

Research article

A hybrid CNN-LSTM model with adaptive instance normalization for one shot singing voice conversion

Assila Yousuf ^,,
David Solomon George

Department of Electronics and Communication Engineering, Rajiv Gandhi Institute of Technology, Kottayam, Kerala, 686501, India (Affiliated to APJ Abdul Kalam Technological University, Kerala)

Received: 04 March 2024 Revised: 22 May 2024 Accepted: 23 May 2024 Published: 03 June 2024

Singing voice conversion methods encounter challenges in achieving a delicate balance between synthesis quality and singer similarity. Traditional voice conversion techniques primarily emphasize singer similarity, often leading to robotic-sounding singing voices. Deep learning-based singing voice conversion techniques, however, focus on disentangling singer-dependent and singer-independent features. While this approach can enhance the quality of synthesized singing voices, many voice conversion systems still grapple with the issue of singer-dependent feature leakage into content embeddings. In the proposed singing voice conversion technique, an encoder decoder framework was implemented using a hybrid model of convolutional neural network (CNN) accompanied by long short term memory (LSTM). This paper investigated the use of activation guidance and adaptive instance normalization techniques for one shot singing voice conversion. The instance normalization (IN) layers within the auto-encoder effectively separated singer and content representations. During conversion, singer representations were transferred using adaptive instance normalization (AdaIN) layers. This singing voice system with the help of activation function prevented the transfer of singer information while conveying the singing content. Additionally, the fusion of LSTM with CNN can enhance voice conversion models by capturing both local and contextual features. The one-shot capability simplified the architecture, utilizing a single encoder and decoder. Impressively, the proposed hybrid CNN-LSTM model achieved remarkable performance without compromising either quality or similarity. The objective and subjective evaluation assessments showed that the proposed hybrid CNN-LSTM model outperformed the baseline architectures. Evaluation results showed a mean opinion score (MOS) of 2.93 for naturalness and 3.35 for melodic similarity. These hybrid CNN-LSTM techniques allowed it to perform high-quality voice conversion with minimal training data, making it a promising solution for various applications.

Keywords:

Citation: Assila Yousuf, David Solomon George. A hybrid CNN-LSTM model with adaptive instance normalization for one shot singing voice conversion[J]. AIMS Electronics and Electrical Engineering, 2024, 8(3): 292-310. doi: 10.3934/electreng.2024013

Related Papers:

[1]	Mohamad Arif Awang Nawi, Nor Farid Mohd Noor, Ramizu Shaari, Ameera Kamal Khaleel, Muhamamd Amirul Mat Lazin, Ibrahim Mohammed Sulaiman, Mustafa Mamat . The patterns of facial fractures in traumatic brain injury (TBI) patients using ordinal regression: a retrospective study of five years. AIMS Neuroscience, 2022, 9(3): 345-357. doi: 10.3934/Neuroscience.2022019
[2]	Naeimeh Akbari-Gharalari, Maryam Ghahremani-Nasab, Roya Naderi, Leila Chodari, Farshad Nezhadshahmohammad . The potential of exosomal biomarkers: Revolutionizing Parkinson's disease: How do they influence pathogenesis, diagnosis, and therapeutic strategies?. AIMS Neuroscience, 2024, 11(3): 374-397. doi: 10.3934/Neuroscience.2024023
[3]	Carolyn E. Jones, Miranda M. Lim . Phasic Sleep Events Shape Cognitive Function after Traumatic Brain Injury: Implications for the Study of Sleep in Neurodevelopmental Disorders. AIMS Neuroscience, 2016, 3(2): 232-236. doi: 10.3934/Neuroscience.2016.2.232
[4]	Robin Quilliam, Samantha Quilliam, Morag Turnbull, Shelagh Parkinson, Godwin Oligbu . Catatonia as a presentation of autism in a child: a case report. AIMS Neuroscience, 2020, 7(3): 327-332. doi: 10.3934/Neuroscience.2020019
[5]	Moazameh Golshani, Mohsen Basiri, Mohammad Shabani, Iraj Aghaei, Majid Asadi-Shekaari . Effects of erythropoietin on bile duct ligation-induced neuro-inflammation in male rats. AIMS Neuroscience, 2019, 6(2): 43-53. doi: 10.3934/Neuroscience.2019.2.43
[6]	Kimberly A Cote, Catherine E Milner, Tamara A Speth . Altered Sleep Mechanisms following Traumatic Brain Injury and Relation to Waking Function. AIMS Neuroscience, 2015, 2(4): 203-228. doi: 10.3934/Neuroscience.2015.4.203
[7]	Dinesh Chouksey, Pankaj Rathi, Ajoy Sodani, Rahul Jain, Hashash Singh Ishar . Postural orthostatic tachycardia syndrome in patients of orthostatic intolerance symptoms: an ambispective study. AIMS Neuroscience, 2021, 8(1): 74-85. doi: 10.3934/Neuroscience.2021004
[8]	Che Aishah Nazariah Ismail, Rahimah Zakaria, Khairunnuur Fairuz Azman, Nazlahshaniza Shafin, Noor Azlina Abu Bakar . Brain-derived neurotrophic factor (BDNF) in chronic pain research: A decade of bibliometric analysis and network visualization. AIMS Neuroscience, 2024, 11(1): 1-24. doi: 10.3934/Neuroscience.2024001
[9]	Robert A. Moss . Psychotherapy in pain management: New viewpoints and treatment targets based on a brain theory. AIMS Neuroscience, 2020, 7(3): 194-207. doi: 10.3934/Neuroscience.2020013
[10]	Galina V. Portnova, Michael S. Atanov . Nonlinear EEG parameters of emotional perception in patients with moderate traumatic brain injury, coma, stroke and schizophrenia. AIMS Neuroscience, 2018, 5(4): 221-235. doi: 10.3934/Neuroscience.2018.4.221

Abstract

1. Introduction

Chronic traumatic encephalopathy (CTE) is a neurodegenerative disease secondary to repetitive mild traumatic brain injury (mTBI), including concussions and sub-concussive impacts, resulting in long-term issues with cognition, behavior and mood [1]–[7]. CTE was initially recognized in boxers who developed symptoms like ataxia, memory loss and personality change, and it was coined as the “punch drunk” syndrome or “dementia pugilistica” [1],[3],[4],[8]–[10]. Over time, it became evident that CTE also affected military personnel, domestic violence victims and those participating in contact sports like football, ice hockey, professional wrestling, rugby, soccer and boxing [1]–[4],[8].

Neurodegeneration and symptoms in CTE progress even in the absence of further traumatic insults [2],[6],[11],[12]. mTBI is thought to trigger an inflammatory cascade and lead to blood brain barrier permeability, axonal injury and micro-hemorrhages [13]–[16]. As a result, there is deposition of pathogenic proteins, including the pathogenic cis-isoform of p-tau, which, through the process termed cistauosis, catalyzes conversion of normal into pathogenic tau [2],[17]–[21]. As such, CTE develops in pathological stages with worsening depositions of p-tau, neurofibrillary tangles and brain atrophy in similar but distinct fashions as other neurodegenerative diseases like Alzheimer's disease [2].

The current gold standard diagnosis for CTE is post-mortem pathological examination. Trauma encephalopathy syndrome was proposed to help diagnose patients with CTE. This criterion consists of a history of repetitive brain injury, persistent symptoms over a year and an absence of comorbidities that may also account for the symptoms. Also present should be a cognitive, behavioral or mood impairment in the presence of progressive decline over more than a year, impulsivity or headaches [2],[22]–[24]. However, patients with CTE present differently and there is no consensus on a single, best set of clinical or research diagnostic criteria [2],[5],[6],[25]. Therefore, there is increasing investigation of adjunctive non-invasive diagnostic modalities. In this paper, we review the recent advances in the use of neuroimaging and fluid biomarkers for early CTE detection.

1.1. Diagnostic Imaging

1.1.1. Magnetic Resonance Imaging

Magnetic resonance imaging (MRI) produces images by analyzing tissue characteristics using magnetic fields and radio waves. It is the current imaging modality of choice due to its improved soft tissue differentiation, ability to detect diffuse axonal injury and lack of ionizing radiation compared to computed tomography (CT). The gross, macroscopic structural changes with CTE include cerebral atrophy that is most severe in the frontotemporal lobes, vermis, thalamus, mamillary bodies and hypothalamus. There is also ventricular enlargement, thinning of the corpus callosum and depigmentation of the substantia nigra and locus coeruleus. Though it is not a consistent feature of CTE, neuropathologic change (CTE-NC), i.e., the presence of cavum septum pellucidum in imaging, is associated with CTE. Microhemorrhages representing diffuse axonal injury may also be present [26]–[30]. These structural findings are not specific to CTE, however [30]. Therefore, there has been increasing research on the use of alternative, more advanced imaging methods as tools to identify and understand the progression of CTE in vivo.

2. Investigational imaging modalities

2.1. Quantitative Structural Analyses

Methods of quantitative brain volume analyses, like assessing cortical thickness in other neurodegenerative diseases, have provided useful for diagnosis and prognosis [31]–[34]. A study showed that hippocampal volume in football athletes was inversely correlated with the presence of concussions and amount of football played [35]. These findings, however, may not be specific to CTE considering the volume loss seen in other neurodegenerative diseases [36]. Furthermore, the inclusion criteria consisted of sport participation within the last year in those aged between 18–26. Therefore, it did not necessarily include or correlate with the presence of neuropsychiatric symptoms associated with CTE [16].

2.2. Diffusion Tensor Imaging

Diffusion tensor imaging (DTI) is an MRI technique that examines the longitudinal diffusion of water through axons to evaluate the orientation and integrity of white matter tracts. A fractional anisotropy (FA) close to 1 means that diffusion occurs along one axis and is otherwise restricted. Axial diffusivity (AD) and radial diffusivity (RD) are similar measures that reflect the magnitude of diffusion running parallel and perpendicular to white matter tracts, respectively [37],[38]. As such, decreased FA, decreased AD and increased RA would be expected in CTE due to decreased white matter integrity. These findings have been demonstrated in mTBI, and even had prognostic value [39]–[47]. A post-mortem tissue DTI analysis of patients with confirmed CTE-NC by Holleran et al. demonstrated associations between decreased FA and reduced white matter integrity [38]. A DTI analysis by Herweh et al. that evaluated male amateur boxers demonstrated associations between decreased FA and neuropsychological outcomes [48]. A study by Kraus et al. showed that study subjects who were included on the basis of having a history of mTBI (22 subjects) or moderate to severe TBI (17 subjects) had decreased FA and RD. The study also suggested that DTI can help to determine the relationship between TBI and cognitive differences and distinguish the spectrum and severity of TBI [49]. Another study showed decreased FA and no changes in the RD and FA in patients who were football players with sub-concussive impacts, with return to baseline after they abstained from play. Further studies are needed to assess the utility of RD and AD in patients specifically with CTE [50].

2.3. Functional MRI

Functional MRI (fMRI) is also known as blood oxygen level-dependent MRI. Neuronal activation in specific brain areas results in an increased oxyhemoglobin-to-deoxyhemoglobin ratio secondary to increased local blood flow, resulting in changes in magnetic susceptibility that are detected by fMRI when a specific task is performed [51]. This method is heavily used in behavioral and physiologic research, as it correlates well with neuronal activity. A theoretical limitation is that the results may be confounded in patients with CTE who already have reduced and altered cerebral blood flow. This modality is yet to be investigated in CTE. A few studies have, however, demonstrated altered brain activation patterns in the fMRI results of living patients with acute and repetitive mTBI [52]–[62]. The correlation between fMRI and altered brain activation patterns in those with mTBI may overwhelm the theoretical limitation. Furthermore, arterial spin imaging MRI, a type of fMRI, has shown to represent aberrant cerebral blood flow in those with mTBI [63].

2.4. Magnetic Resonance Spectroscopy

Magnetic resonance spectroscopy measures concentrations of metabolites within the brain based on the chemical shift of their protons, which is determined by the proton's chemical environment. This modality is useful for CTE when considering the pathological changes, including neuroinflammation, in the acute and chronic stages of the disease. In fact, a study investigating male USA National Football League (NFL) players between 40–69 with self-reported neuropsychiatric symptoms were found to have decreased cellular energy metabolism, as evidenced by lower creatinine in the parietal white matter. Neuro-inflammatory metabolites like glutamate, glutathione and myo-inositol also correlated with their behavioral and mood symptoms [64]. Several other studies have demonstrated metabolite abnormalities in patients with a history of repeated head impacts, including decreases in NAA, NAA/Cho and NAA/Cr, as well as increases in Cho, ml, glutamine, choline, fucose and phenylalanine [65]–[70].

2.5. Susceptibility Weighted Imaging

Susceptibility weighted imaging (SWI) takes advantage of different responses, or susceptibilities, to molecules within a magnetic field. These susceptibilities are measured as phase shifts and superimposed on an MRI, highlighting local susceptibility changes. In the setting of TBI, it can be used to reveal hemorrhagic contusions or diffuse axonal injury. SWI abnormalities, including microhemorrhages, have been demonstrated in contact sport participants, active duty military members and those with concussive-like symptoms and a history of repetitive mTBI [71]–[74]. Considering that it has shown utility in predicting neuropsychiatric outcomes for those with acute mTBI, its use should be considered in predicting the likelihood of the development of CTE [75],[76]. Neurodegenerative disorders often have characteristic features and positions of cerebral microbleeds [77],[78]. Research investigating the distribution and features of cerebral microbleeds in CTE to make a specific diagnosis would be beneficial.

2.6. Position Emission Tomography

Position emission tomography (PET) CT, which employs the use of radioisotopic biomarkers, has been garnering interest for elucidating elevated tau, beta-amyloid, neurofibrillary tangles and other neuroinflammatory proteins [79]. For example, FDDNP binds to the neurofibrillary tangles and proteins that are associated with CTE. As such, it has been employed in the diagnosis of CTE [79],[80]. However, FDDNP has also been shown to bind to beta-amyloid and hyperphosphorylated tau and is therefore limited in specificity when discriminating against other neurological degenerative diseases, such as Alzheimer's disease [79].

The development of biomarkers specific for the hyperphosphorylated tau proteins associated with CTE, such as [¹⁸F]AV-1451 (flortaucipir), are of interest and have been recently studied. [¹⁸F]AV-1451 binds to the paired helical tau deposition associated with Alzheimer's disease, and studies are being conducted to investigate its utility for visualizing tau deposition patterns that are associated with CTE, such as those in the medial temporal lobe, brain stem and diencephalon [81]. One such study involved [¹⁸F]AV-1451-PET scans from 26 former NFL players aged between 40–69 with reported neuropsychiatric symptoms; the researchers observed a statistically significant increase in the mean standard reuptake of [¹⁸F]AV-1451 in the bilateral superior frontal, bilateral medial temporal and left parietal regions as compared to the controls [82]. Another study followed a retired NFL player who underwent an MRI and [¹⁸F]AV-1451 PET scan, which revealed uptake in the bilateral medial temporal lobes and parietal regions 4 years before a post-mortem diagnosis of stage-4 CTE [83]. These studies, which are confined to small sample sizes and the single case report, illustrate the need for further investigation to validate [¹⁸F]AV-1451 as an optimal radiotracer for in vivo PET scans to diagnose CTE. Other developing PET tracers of interest that bind to tau proteins include [11C]PBB3[84], THK-5105[85], THK-5117[86], THK-5351[87] and T807[88]. [¹⁸F]florbetapir and [¹¹C]PiB PET measure amyloid beta plaques [89],[90]. The characteristics of selected biomarkers are reviewed in Table 1. Studies have also investigated the use of [¹¹C]flumazenil and [¹⁸F]flumazenil, which bind to the GABA_A system in patients with a history of repeated head injury [91]. Lastly, the translocator protein (TSPO) and copper have also been targeted with several radiotracers to assess inflammation [91]. Several studies have also investigated the use of [¹⁸F]FDG for patients with a history of mTBI, generally showing decreased brain glucose metabolism in the cerebellum, vermis, pons, temporal lobe, prefrontal cortex and limbic system [91]. Though cortical uptake regions vary, studies have generally consistently demonstrated uptake in the temporal lobes, limbic system, midbrain, hippocampi and amygdalae [2],[79],[91]. The need for further research into PET biomarkers for hyperphosphorylated tau proteins specifically associated with CTE and tasked with reducing off-target binding is warranted.

Table 1. PET CT biomarkers.

Biomarker	Specificity
FDDNP	Affinity for intracellular neurofibrillary tangles but has been found to be “non-selective” due to its binding with extracellular β-amyloid and tau, which is a feature of Alzheimer's disease and not necessarily CTE [17],[79],[91]
[¹⁸F]AV-1451 (flortaucipir)	Affinity for hyperphosphorylated tau proteins. However, there is a possibility of false negatives, as there is also high binding affinity for paired helical tau filaments in Alzheimer's disease and not CTE [91],[92].
[¹¹C]PBB3	Affinity for Alzheimer's disease tau pathology but has mixed reviews over its ability to identify paired helical tau filaments in CTE [93],[94]
THK-5105	High binding affinity to tau protein aggregates and tau-rich Alzheimer disease, but it has reported to have a high background signal in PET images, which could affect its utility. Also, it has not been investigated for tau proteins seen in CTE-related pathology [86],[95].
THK-5117	Affinity for Alzheimer's disease-related tau protein in the medial temporal lobe in port-mortem patients, but not yet investigated for the CTE-related tau deposits [96].
THK-5351	Showed affinity with increased t-tau levels in the parahippocampal gyrus, but not investigated for CTE-associated tau patterns to date [97].

| Show Table

DownLoad: CSV

3. Fluid biomarkers

3.1. p-Tau

CTE is characterized by an accumulation of differentiated cis p-tau proteins in the vasculature of sulcal depths with large groups of astrocytic tangles, neurofibrillary tangles and neurites [17]. These abnormal, non-functional p-tau clusters develop within axons [17],[98]. High values of p-tau are also found in single-event TBI and other neurodegenerative diseases, and therefore cannot serve as the sole means of diagnosis of CTE [99],[100]. However, it can serve as a biomarker when considered in the overall clinical setting with accompanying imaging findings from developing diagnostic tools, like PET.

3.2. TREM2

A promising biomarker that is not as widely employed for tau is the inflammatory marker triggering receptor expressed on myeloid cells 2 (TREM2), a triggering receptor found in several myeloid lineage cells such as peripheral macrophages, dendritic cells and microglial cells in the central nervous system (CNS) [101]. TREM2 is also expressed in the microglia of the brain, regulating microglial activation and playing a multi-faceted role in its immune response [102],[103]. Animal studies have shown that TREM2 is upregulated in the early stages following injury, making it a potential biomarker for TBI and other head injuries [104]. When microglia in the brain are activated, following injury, cleavage of TREM2 by proteases follows. These proteases release soluble TREM2 (sTREM2), indicating that injury has occurred. A study highlighted the relationship between sTREM2 levels in cerebrospinal fluid (CSF) and t-tau concentrations in 68 former NFL players aged 40–69 with self-reported neuropsychiatric symptoms (compared to healthy controls), ultimately finding a positive correlation between the two [102]. The presence of TREM2 variants increases the likelihood of developing Alzheimer's disease by 2–4 times [105],[106]. Therefore, elevated levels of sTREM2 are non-specific. Because upregulation of TREM2 begins early and persists over time [104], it could prove to be a key inflammatory marker used for the diagnosis of CTE in the appropriate clinical context.

3.3. CCL11

A biomarker with potential for pre-mortem CTE diagnosis is the chemokine C-C motif chemokine ligand 11 (CCL11). Chemokines are proteins that play a central role and facilitate biochemical and cellular events in the immune response. They upregulate leukocytes and act as secondary pro-inflammatory mediators [107],[108]. CCL11 is a chemoattractant of eosinophils in the peripheral immune system and has recently been shown to also penetrate the blood brain barrier [107]. A study showed that the brain of mice secreted CCL11 as a response to the inflammation of astrocytes, pericytes and microglia [109]. A key study showed an increase in the plasma blood levels of CCL11 to correlate with a decrease in learning, memory and neurogenesis in the brains of mice [110]. It has been proposed that the main reason for CCL11's involvement in neurological decline is its ability to increase the microglial production of reactive oxygen species and promote excitotoxic neuronal death [111]. A study indicated that CCL11 is released in the CSF from the choroid plexus in the brain, suggesting direct effects on the brain [112]. This is also associated with an increase in the ratio of cytokine interleukin (IL)-4 and interferon (IFN)-γ in the choroid plexus and CSF. Prior research has shown the physiological importance of CCL11 to neurological function, but it may be a useful biomarker for CTE considering its ability to distinguish it among other neurodegenerative diseases. A collection of studies showed that plasma CCL11 increased in patients with Alzheimer's disease and Huntington's disease, while it decreased in amyotrophic lateral sclerosis and secondary progressive multiple sclerosis [113]–[115]. Using ELISA, a post-mortem study of 23 former football players with neuropathologically diagnosed CTE and 50 subjects with neuropathologically diagnosed Alzheimer's disease showed a statistically significant increase in the CCL11 levels in the dorsolateral frontal cortex of CTE subjects compared to the Alzheimer's disease and control subjects [116]. Another study with subjects aged 25–33 showed significant increases in CCL11 in the CSF of retired football players relative to swimmers with no TBI history and a sedentary control group. Analysis of the IL-4-to-IFN-γ ratio also showed significant increase in the football players compared to the others in the study. Lastly, CCL11 levels showed a strong positive correlation with the number of years of football played [107]. There is a lot of promise that CCL11 can provide CTE diagnosis for patients pre-mortem. More comprehensive research can be done to analyze the relationship of CCL11 levels with the IL-4-to-IFN-γ ratio found in the CSF of different types of neurodegenerative diseases in pre- and post-mortem brains to build a more accurate predictive model for diagnosis.

3.4. NFL

Another biomarker widely employed in TBI is neurofilament-L (NfL). NfL comes from the intermediate filament protein family and is part of the neuronal cytoskeleton [117]. It can serve as an indicator of CNS axonal damage. NfL is released into the CSF [117]. A meta-analysis with a sample size of 1118 patients showed that NfL CSF, serum and plasma levels were significantly higher in patients with TBI compared to the control patients without prior TBI [117]. Another study showed that patients with Alzheimer's disease, Guillain-Barré-syndrome and amyotrophic lateral sclerosis had increased levels of serum NfL compared to a control without CNS damage [118]. These studies suggest that NfL could be used to also conduct future research toward CTE diagnosis.

3.5. Glial fibrillary acidic protein

Glial fibrillary acidic protein (GFAP) is a cytoskeletal monomeric filament protein located in the astroglial cells of gray and white matter [119]. A study showed that levels of GFAP, tau and NfL were all higher in the group of 277 patients suspected with mTBI, with GFAP yielding high discriminatory power in differentiating these patients from the 49 healthy controls, with an area under the curve (AUC) of 0.93 [120]. In the same study, GFAP similarly served as a strong predictor of mTBI when examining MRI abnormalities, with an AUC of 0.83 [120]. Another study of GFAP's ability to predict CT abnormalities showed an AUC of 0.88 when examining 215 patients (83% with mTBI; mean age 42.5 ± 18.0) [121]. Further analysis could be done using pre-mortem CTE subjects and comparing GFAP levels in those patients to a control to assess whether this translates to specifically diagnosing CTE over other diseases as a result of a TBI. The characteristics of the biomarkers are summarized in Table 2.

Table 2. Blood biomarkers for the diagnosis of CTE.

Biomarker	Description	Significance	Advantages	Limitations
p-tau	Hyperphosphorylated tau protein found in the cortical vasculature within the sulcal depths [17].	-Repetitive head injury causes the conversion of typical tau protein to p-tau [2]. -P-tau is indicative of axonal functional decline and will deposit in predictable patterns and high concentrations following brain injury [17],[122],[123].	-Consistent and sensitive results considering large concentrations following injury [123]. -A blood sample is less invasive than lumbar puncture [124].	Other neurodegenerative diseases also express high concentrations of p-tau [123].
TREM2	Triggering receptor found in myeloid lineage cells that regulates CNS microglial activation [101].	-Cleavage of TREM2 by proteases following head injury produces sTREM2 -Increased sTREM2 levels in the CSF are indicative of increased protease activity, likely due to trauma [102].	Early upregulation of TREM2 following head injury may eventually lead to its diagnostic use in potential CTE cases in vivo [102],[104].	-Invasive sample collection with lumbar puncture [102]. -TREM2 variants can impair the function of receptors due to poor signaling, ultimately leading to decreased sTREM levels [125].
CCL11	-Chemokine that serves as a mediator in inflammatory cascades [107],[108]. -Penetrates the blood brain barrier [107]. -Secreted into CSF by choroid plexus in the brain [112].	-Increases microglial production of reactive oxygen species and promotes excitotoxic neuronal death [111]. -Reflective of neuroinflammatory processes [111].	-Potential ability to differentiate between CTE and other neurodegenerative diseases [116]. -Possible correlation with number of repeated head impacts [107].	-Its main role in the CNS is unclear, as it is a chemoattractant of eosinophils in the peripheral immune system [112].
NfL	Part of the Intermediate filament protein family and of the neuronal cytoskeleton [117].	-Can be measured, as axonal damage induces its release into the CSF [117].	-Released in a delayed fashion and may be correlated with cognitive decline in patients with chronic TBI [126]. -Specific to CTE [116].	-Conflicting data regarding its validity in accurately diagnosing CTE [117].
GFAP	Cytoskeletal monomeric filament protein in the brain's astroglial cells [119].	-Reflective of astroglial injury and released acutely following TBI [116].	-Better predictor of mTBI than NfL and p-tau [120].	-Utility in CTE specifically unknown [120],[121].

| Show Table

DownLoad: CSV

4. Conclusion

In conclusion, the current clinical diagnosis of CTE relies on clinical symptomatology and structural imaging findings. While there is extensive research on imaging and fluid biomarkers in relation to TBI, there is comparatively limited research on CTE. Particularly, the referenced studies are often small and investigate mTBI, TBI or repetitive head injury and do not study CTE directly. Investigating CTE directly is especially difficult considering the lack of consensus on pre-mortem diagnostic criteria. The studies also frequently include patients based on a history of sport participation alone, self-reported history of TBI or self-reported neuropsychiatric symptoms. The studied biomarkers are also often elevated in other neurodegenerative disorders, and there is relatively limited research on the use of biomarkers to differentiate CTE from other neurodegenerative disorders. Lastly, many of the fluid biomarkers are also elevated following a single TBI event, and there is not an imaging or fluid biomarker approved solely for CTE. We look forward to further research on the early promising imaging modalities and fluid biomarkers to potentially assist in the diagnosis of CTE and in differentiation of it from other neurodegenerative diseases.

References

[1]	Helander E, Virtanen T, Nurminen J, Gabbouj M (2010) Voice conversion using partial least squares regression. IEEE/ACM Transactions on Audio, Speech and Language Processing 18: 912–921. https://doi.org/10.1109/TASL.2011.2165944 doi: 10.1109/TASL.2011.2165944
[2]	Saito Y, Takamichi S, Saruwatari H (2017) Voice conversion using input-to-output highway networks. IEICE T Inf Syst 100: 1925–1928. https://doi.org/10.1587/transinf.2017EDL8034 doi: 10.1587/transinf.2017EDL8034
[3]	Yeh CC, Hsu PC, Chou JC, Lee HY, Lee LS (2018) Rhythm Flexible Voice Conversion Without Parallel Data Using Cycle-GAN Over Phoneme Posteriorgram Sequences. IEEE Spoken Language Technology Workshop (SLT) 274–281. https://doi.org/10.1109/SLT.2018.8639647 doi: 10.1109/SLT.2018.8639647
[4]	Sun L, Wang H, Kang S, Li K, Meng HM (2016) Personalized Cross-Lingual TTS Using Phonetic Posteriorgrams. Interspeech 322–326. https://doi.org/10.21437/Interspeech.2016-1043 doi: 10.21437/Interspeech.2016-1043
[5]	Tian X, Chng ES, Li H (2019) A Speaker-Dependent WaveNet for Voice Conversion with Non-Parallel Data. Interspeech 201–205. https://doi.org/10.21437/Interspeech.2019-1514 doi: 10.21437/Interspeech.2019-1514
[6]	Takahashi N, Singh MK, Mitsufuji Y (2023) Robust One-Shot Singing Voice Conversion. arXiv: 2210.11096v2. https://doi.org/10.48550/arXiv.2210.11096
[7]	Hono Y, Hashimoto K, Oura K, Nankaku Y, Tokuda K (2019) Singing Voice Synthesis Based on Generative Adversarial Networks. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 6955–6959. https://doi.org/10.1109/ICASSP.2019.8683154 doi: 10.1109/ICASSP.2019.8683154
[8]	Sun L, Kang S, Li K, Meng H (2015) Voice conversion using deep bidirectional long short-term memory based recurrent neural networks. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 4869–4873. https://doi.org/10.1109/ICASSP.2015.7178896 doi: 10.1109/ICASSP.2015.7178896
[9]	Kaneko T, Kameoka H, Hiramatsu K, Kashino K (2017) Sequence-to-Sequence Voice Conversion with Similarity Metric Learned Using Generative Adversarial Networks. Interspeech 2017: 1283–1287. http://dx.doi.org/10.21437/Interspeech.2017-970 doi: 10.21437/Interspeech.2017-970
[10]	Freixes M, Alías F, Carrie JC (2019) A unit selection text-to-speech-and-singing synthesis framework from neutral speech: proof of concept. EURASIP Journal on Audio, Speech, and Music Processing 2019: 1–14. https://doi.org/10.1186/s13636-019-0163-y doi: 10.1186/s13636-019-0163-y
[11]	Hono Y, Hashimoto K, Oura K, Nankaku Y, Tokuda K (2021) Sinsy: a deep neural network-based singing voice synthesis system. IEEE/ACM T Audio Spe 29: 2803–2815. https://doi.org/10.1109/TASLP.2021.3104165 doi: 10.1109/TASLP.2021.3104165
[12]	Sisman B, Vijayan K, Dong M, Li H (2019) SINGAN: Singing Voice Conversion with Generative Adversarial Networks. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) 112–118. https://doi.org/10.1109/APSIPAASC47483.2019.9023162 doi: 10.1109/APSIPAASC47483.2019.9023162
[13]	Sisman B, Li H (2020) Generative adversarial networks for singing voice conversion with and without parallel data. Odyssey 238–244. https://doi.org/10.21437/Odyssey.2020-34 doi: 10.21437/Odyssey.2020-34
[14]	Zhao W, Wang W, Sun Y, Tang T (2019) Singing voice conversion based on wd-gan algorithm. IEEE 4th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC) 950–954. https://doi.org/10.1109/IAEAC47372.2019.8997824 doi: 10.1109/IAEAC47372.2019.8997824
[15]	Fang F, Yamagishi J, Echizen I, Lorenzo-Trueba J (2018) High-Quality Nonparallel Voice Conversion Based on Cycle-Consistent Adversarial Network. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 5279–5283. https://doi.org/10.1109/ICASSP.2018.8462342 doi: 10.1109/ICASSP.2018.8462342
[16]	Kameoka H, Kaneko T, Tanaka K, Hojo N (2018) StarGAN-VC: non-parallel many-to-many Voice Conversion Using Star Generative Adversarial Networks. IEEE Spoken Language Technology Workshop (SLT) 266–273. https://doi.org/10.1109/SLT.2018.8639535 doi: 10.1109/SLT.2018.8639535
[17]	Chen Y, Xia R, Yang K, Zou K (2023) MICU: Image Super-resolution via Multi-level Information Compensation and U-net. Expert Syst Appl 245: 123111. https://doi.org/10.1016/j.eswa.2023.123111 doi: 10.1016/j.eswa.2023.123111
[18]	Chen Y, Xia R, Yang K, Zou K (2023) MFMAM: Image Inpainting via Multi-Scale Feature Module with Attention Module. Comput Vis Image Und 238: 103883. https://doi.org/10.1016/j.cviu.2023.103883 doi: 10.1016/j.cviu.2023.103883
[19]	Chen Y, Xia R, Yang K, Zou K (2023) GCAM: Lightweight Image Inpainting via Group Convolution and Attention Mechanism. Int J Mach Learn Cyb 15: 1815–1825. https://doi.org/10.1007/s13042-023-01999-z doi: 10.1007/s13042-023-01999-z
[20]	Chen Y, Xia R, Yang K, Zou K (2024) DNNAM: Image Inpainting Algorithm via Deep Neural Networks and Attention Mechanism. Appl Soft Comput 111392. https://doi.org/10.1016/j.asoc.2024.111392 doi: 10.1016/j.asoc.2024.111392
[21]	Chen Y, Xia R, Yang K, Zou K (2023) DARGS: Image Inpainting Algorithm via Deep Attention Residuals Group and Semantics. J King Saud Univ-Comput 35: 101567. https://doi.org/10.1016/j.jksuci.2023.101567 doi: 10.1016/j.jksuci.2023.101567
[22]	Chen L, Zhang X, Li Y, Sun M, Chen W (2024) A Noise-Robust Voice Conversion Method with Controllable Background Sounds. Complex Intell Syst 1–14. https://doi.org/10.1007/s40747-024-01375-6 doi: 10.1007/s40747-024-01375-6
[23]	Walczyna T, Piotrowski Z (2023) Overview of Voice Conversion Methods Based on Deep Learning. Applied sciences 13: 3100. https://doi.org/10.3390/app13053100 doi: 10.3390/app13053100
[24]	Liu EM, Yeh JW, Lu JH, Liu YW (2023) Speaker Embedding Space Cosine Similarity Comparisons of Singing Voice Conversion. The Journal of the Acoustical Society of America (JASA) 154: A244–A244. https://doi.org/10.1121/10.0023424 doi: 10.1121/10.0023424
[25]	Hsu CC, Hwang HT, Wu YC, Tsao Y, Wang HM (2016) Voice conversion from non-parallel corpora using variational auto-encoder. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA) 1–6. https://doi.org/10.1109/APSIPA.2016.7820786 doi: 10.1109/APSIPA.2016.7820786
[26]	Tobing PL, Wu YC, Hayashi T, Kobayashi K, Toda T (2019) Non-Parallel Voice Conversion with Cyclic Variational Autoencoder, Interspeech 674–678. https://doi.org/10.21437/Interspeech.2019-2307 doi: 10.21437/Interspeech.2019-2307
[27]	Yook D, Leem SG, Lee K, Yoo IC (2020) Many- to-many voice conversion using cycle-consistent variational autoencoder with multiple decoders. Odyssey 215–221. https://doi.org/10.21437/Odyssey.2020-31 doi: 10.21437/Odyssey.2020-31
[28]	Hsu CC, Hwang HT, Wu YC, Tsao Y, Wang HM (2017) Voice conversion from unaligned corpora using variational autoencoding wasserstein generative adversarial networks. arXiv preprint arXiv: 1704.00849. https://doi.org/10.48550/arXiv.1704.0084
[29]	Huang WC, Violeta LP, Liu S, Shi J, Toda T (2023) The Singing Voice Conversion Challenge 2023. 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) 1–8. https://doi.org/10.1109/ASRU57964.2023.10389671 doi: 10.1109/ASRU57964.2023.10389671
[30]	Chen Q, Tan M, Qi Y, Zhou J, Li Y, Wu Q (2022) V2C: Visual Voice Cloning. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 21242–21251.
[31]	Qian K, Zhang Y, Chang S, Yang X, Hasegawa-Johnson M (2019) Autovc: Zero-shot voice style transfer with only autoencoder loss. International Conference on Machine Learning 5210–5219.
[32]	Patel M, Purohit M, Parmar M, Shah NJ, Patil HA (2020) Adagan: Adaptive gan for many-to-many non-parallel voice conversion.
[33]	Liu F, Wang H, Peng R, Zheng C, Li X (2021) U2-VC: one-shot voice conversion using two-level nested U-structure. EURASIP Journal on Audio, Speech, and Music Processing 2021: 1–15. https://doi.org/10.1186/s13636-021-00226-3 doi: 10.1186/s13636-021-00226-3
[34]	Liu F, Wang H, Ke Y, Zheng C (2022) One-shot voice conversion using a combination of U2-Net and vector quantization. Appl Acoust 199: 109014. https://doi.org/10.1016/j.apacoust.2022.109014 doi: 10.1016/j.apacoust.2022.109014
[35]	Wu DY, Lee HY (2020) One-shot voice conversion by vector quantization. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 7734–7738. https://doi.org/10.1109/ICASSP40776.2020.9053854 doi: 10.1109/ICASSP40776.2020.9053854
[36]	Chou JC, Lee HY (2019) One-Shot Voice Conversion by Separating Speaker and Content Representations with Instance Normalization. Interspeech 664–668. https://doi.org/10.21437/Interspeech.2019-2663 doi: 10.21437/Interspeech.2019-2663
[37]	Huang X, Belongie S (2017) Arbitrary style transfer in real-time with adaptive instance normalization. IEEE International Conference on Computer Vision (ICCV) 1501–1510. https://doi.org/10.1109/ICCV.2017.167 doi: 10.1109/ICCV.2017.167
[38]	Lian J, Lin P, Dai Y, Li G (2022) Arbitrary Voice Conversion via Adversarial Learning and Cycle Consistency Loss. International Conference on Intelligent Computing 569–578. https://doi.org/10.1007/978-3-031-13829-4_49 doi: 10.1007/978-3-031-13829-4_49
[39]	Gu Y, Zhao X, Yi X, Xiao J (2022) Voice Conversion Using learnable Similarity-Guided Masked Autoencoder. International Workshop on Digital watermarking 13825: 53–67. https://doi.org/10.1007/978-3-031-25115-3_4 doi: 10.1007/978-3-031-25115-3_4
[40]	Chen YH, Wu DY, Wu TH, Lee HY (2021) AGAIN-VC: A one-shot voice conversion using activation guidance and adaptive instance normalization. IEEE International Conference on Acoustics, Speech, and Signal Processing 5954–5958. https://doi.org/10.1109/ICASSP39728.2021.9414257 doi: 10.1109/ICASSP39728.2021.9414257
[41]	Ulyanov D, Lebedev V, Vedaldi A, Lempitsky VS (2016) Texture networks: Feed-forward synthesis of textures and stylized images. Proceedings of the 33nd International Conference on Machine Learning 1349–1357.
[42]	Ioffe S, Szegedy C (2015) Batch normalization: Accelerating deep network training by reducing internal covariate shift. Proceedings of the 32nd International Conference on International Conference on Machine Learning 37: 448–456.
[43]	Li Y, Wang N, Shi J, Liu J, Hou X (2016) Revisiting batch normalization for practical domain adaptation. arXiv preprint arXiv: 1603.04779.
[44]	Ulyanov D, Vedaldi A, Lempitsky V (2017) Improved Texture Networks: Maximizing Quality and Diversity in Feed-Forward Stylization and Texture Synthesis. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 4105–4113. https://doi.org/10.1109/CVPR.2017.437 doi: 10.1109/CVPR.2017.437
[45]	Liu J, Han W, Ruan H, Chen X, Jiang D, Li H (2018) Learning Salient Features for Speech Emotion Recognition Using CNN. First Asian Conference on Affective Computing and Intelligent Interaction (ACII Asia) 1–5. https://doi.org/10.1109/ACIIAsia.2018.8470393 doi: 10.1109/ACIIAsia.2018.8470393
[46]	Lim W, Jang D, Lee T (2016) Speech emotion recognition using convolutional and Recurrent Neural Networks. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA) 1–4. https://doi.org/10.1109/APSIPA.2016.7820699 doi: 10.1109/APSIPA.2016.7820699
[47]	Hajarolasvadi N, Demirel H (2019) 3D CNN-Based Speech Emotion Recognition Using K-Means Clustering and Spectrograms. Entropy (Basel) 21: 479. https://doi.org/10.3390/e21050479 doi: 10.3390/e21050479
[48]	Graves A (2012) Long Short-Term Memory Supervised Sequence Labelling with Recurrent Neural Networks. Studies in Computational Intelligence 385: 37–45. https://doi.org/10.1007/978-3-642-24797-2 doi: 10.1007/978-3-642-24797-2
[49]	Chung J, Gulcehre C, Cho K, Bengio Y (2014) Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv: 1412.3555.
[50]	Kumar K, Kumar R, de Boissiere T, Gestin L, Teoh WZ, Sotelo J, et al. (2019) Melgan: Generative adversarial networks for conditional waveform synthesis. Advances in Neural Information Processing Systems 14910–14921.
[51]	Kong J, Kim J, Bae J (2020) HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis. Proceedings of the 34th International Conference on Neural Information Processing Systems 33: 17022–17033.
[52]	Duan Z, Fang H, Li B, Sim KC, Wang Y (2013) The NUS sung and spoken lyrics corpus: A quantitative comparison of singing and speech. Asia-Pacific Signal and Information Processing Association Annual Summit and Conference 1–9. https://doi.org/10.1109/APSIPA.2013.6694316 doi: 10.1109/APSIPA.2013.6694316
[53]	Kubichek R (1993) Mel-cepstral distance measure for objective speech quality assessment. Proceedings of IEEE Pacific Rim Conference on Communications Computers and Signal Processing 1: 125–128. https://doi.org/10.1109/PACRIM.1993.407206 doi: 10.1109/PACRIM.1993.407206
[54]	Kobayashi K, Toda T, Nakamura S (2018) Intra-gender statistical singing voice conversion with direct waveform modification using log spectral differential. Speech Commun 99: 211–220. https://doi.org/10.1016/j.specom.2018.03.011 doi: 10.1016/j.specom.2018.03.011
[55]	Toda T, Tokuda K (2007) A speech parameter generation algorithm considering global variance for hmm-based speech synthesis. IEICE T Inf Syst 90: 816–824. https://doi.org/10.1093/ietisy/e90-d.5.816 doi: 10.1093/ietisy/e90-d.5.816

This article has been cited by:

1.	Yue Huang, Qiao-Xin Li, Ling-Xiao Cao, Gang Wang, Daniel Kam Yin Chan, Conceicao Bettencourt, Adrienne E. Milward, Editorial: Human brain banking – Bridging brain health and precision neurology, 2023, 14, 1664-2295, 10.3389/fneur.2023.1322200
2.	Attila Biró, Antonio Ignacio Cuesta-Vargas, László Szilágyi, Precognition of mental health and neurogenerative disorders using AI-parsed text and sentiment analysis, 2023, 15, 2066-7760, 359, 10.2478/ausi-2023-0022
3.	William Blanks, Marcus Hanshaw, Daniela A. Perez-Chadid, Brandon Lucke-Wold, Emerging frontiers in Chronic Traumatic Encephalopathy: early diagnosis and implications for neurotherapeutic interventions, 2024, 24, 1473-7175, 953, 10.1080/14737175.2024.2385952
4.	Brendan P. Campbell, Katherine W. Turk, Andrew E. Budson, The major challenges with pharmacologic management of chronic traumatic encephalopathy, 2024, 24, 1473-7175, 1035, 10.1080/14737175.2024.2387264
5.	K.V. Duve, S.I. Shkrobot, Z.V. Salii, Chronic traumatic encephalopathy: predictors of the development of cognitive disorders and functional disability, 2024, 19, 2307-1419, 233, 10.22141/2224-0713.19.7.2023.1027
6.	Peter Preben Eggertsen, Johan Palmfeldt, Asger Roer Pedersen, Oana-Veronica Frederiksen, Rikke Katrine Jentoft Olsen, Jørgen Feldbæk Nielsen, Serum neurofilament light chain, inflammatory markers, and kynurenine metabolites in patients with persistent post-concussion symptoms: A cohort study, 2024, 460, 0022510X, 123016, 10.1016/j.jns.2024.123016
7.	Carlo Cervellati, Dario Pedrini, Pietro Pirro, Paola Guindani, Carlo Renzini, Gloria Brombo, Giovanni Zuliani, Qingdong Guan, Neutrophil–Lymphocytes Ratio as Potential Early Marker for Alzheimer’s Disease, 2024, 2024, 0962-9351, 10.1155/2024/6640130
8.	Susanna Friberg, Caroline Lindblad, Frederick A. Zeiler, Henrik Zetterberg, Tobias Granberg, Per Svenningsson, Fredrik Piehl, Eric P. Thelin, Fluid biomarkers of chronic traumatic brain injury, 2024, 20, 1759-4758, 671, 10.1038/s41582-024-01024-z
9.	Michal J. Halicki, Karen Hind, Paul L. Chazot, Blood-Based Biomarkers in the Diagnosis of Chronic Traumatic Encephalopathy: Research to Date and Future Directions, 2023, 24, 1422-0067, 12556, 10.3390/ijms241612556
10.	Lan-Wan Wang, Kuan-Hung Cho, Pi-Yu Chao, Li-Wei Kuo, Chia-Wen Chiang, Chien-Ming Chao, Mao-Tsun Lin, Ching-Ping Chang, Hung-Jung Lin, Chung-Ching Chio, White and gray matter integrity evaluated by MRI-DTI can serve as noninvasive and reliable indicators of structural and functional alterations in chronic neurotrauma, 2024, 14, 2045-2322, 10.1038/s41598-024-57706-7
11.	Natalia Surzenko, Johana Bastidas, Robert W. Reid, Julien Curaba, Wei Zhang, Hamed Bostan, Mickey Wilson, Ashley Dominique, Julia Roberson, Glicerio Ignacio, Slavko Komarnytsky, Alexa Sanders, Kevin Lambirth, Cory R. Brouwer, Bassem F. El‐Khodor, Functional recovery following traumatic brain injury in rats is enhanced by oral supplementation with bovine thymus extract, 2024, 38, 0892-6638, 10.1096/fj.202301859R
12.	2024, 9783437224874, 10, 10.1016/B978-3-437-22487-4.16001-0
13.	Sikandar Khan, Lora Talley, Beyond the Hit: The Hidden Costs of Repetitive Head Trauma, 2025, 20, 2633-1055, 10.1177/26331055251316315
14.	Annalise E. Miner, Jenna R. Groh, Chad Farris, Sarina Hattiangadi, Anna Cui, Adam M. Brickman, Mohamad Alshikho, Gil D. Rabinovici, Howie J. Rosen, Yann Cobigo, Breton Asken, Christopher J. Nowinski, Samantha Bureau, Fereydoun Shahrokhi, Yorghos Tripodis, Monica Ly, Caroline Altaras, Steven Lenio, Robert A. Stern, Grace Rosen, Hunter Kelley, Bertrand Russell Huber, Thor D. Stein, Jesse Mez, Ann C. McKee, Michael L. Alosco, Does white matter and vascular injury from repetitive head impacts lead to a novel pattern on T2 FLAIR MRI? A hypothesis proposal and call for research, 2025, 21, 1552-5260, 10.1002/alz.70085

Reader Comments

Your name:*

Email:*
© 2024 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)