Ph. D. Project
Title:
Heterogeneous data processing (EEG/ERG/clinical data) for the diagnosis of bipolar disorder
Dates:
2024/10/01 - 2027/09/30
Supervisor(s):
Other supervisor(s):
PUPH Thomas Schwitzer (thomas.schwitzer@univ-lorraine.fr)
Description:
Bipolar disorder is a chronic mental illness responsible for mood disturbances, which often requires long-term diagnosis as it shares many pathophysiological and behavioural symptoms with unipolar disorder and major depression. The non-detection rate of the disease is estimated at between 30% and 69% in Europe and the United States, pleading for the urgent need to find reliable and objective indicators of the disease.
Retinal and cortical electrophysiological measurements are relevant for the analysis of mental states explored in mental illness [Tursini2023]. Retinal function is assessed by electrophysiological techniques known as electroretinograms (ERGs). Herbert et al [Hébert2020] established that depressed patients had retinal responses that differed from controls, and that retinal responses differed between depressed patients without medication and depressed patients treated with pharmacotherapy. Analysis of electroencephalogram (EEG) data has also made it possible to extract objective biomarkers for mental disorders, and for bipolar disorder in particular [Tursini2023].
The first aim of this thesis is to use simultaneous EEG/ERG measurements to explore dysfunctions in visual systems from the retina to cortical responses in patients with bipolar disorder. Our analyses are based on flash (ffERG) or pattern (PERG) visual stimulation protocols leading to reproducible, well-controlled responses, and enabling the targeting of the cerebral network associated with vision. From a fundamental point of view, we want to identify a functional model for the propagation of information along the ventral visual pathway, and quantify the alteration in this propagation in the presence of bipolar disorders. This study will lead to the construction of connectivity graphs reflecting the underlying neural network, making it possible to assess the modification of this network in bipolar disorder and to quantify the effect of a therapeutic treatment on this network.
We also plan to search for reliable indicators of mental illness to bring objective tools for the diagnosis of these illnesses. We have already carried out a preliminary study in the case of a recruitment campaign for bipolar subjects [BIMAR], and robust time/frequency biomarkers have been identified that make it possible to discriminate these patients from a control population [Ren2023], but also showing an ability to quantify the evolution of the illness during treatment [Schwitzer2022]. The inclusion of quantitative measures derived from the connectivity graphs previously analysed will also be considered.
The diagnosis of mental illnesses is currently established by the clinician on the basis of clinical observations and qualitative data from neuropsychological surveys, giving rise to scores graded on scales of values. The third objective is to develop methods for combining quantitative measurements extracted from EEG/ERG signals with qualitative data of a discrete and ordinal nature, by modelling the relationships between these variables and their influence on the target (the label) to be estimated. Multivariate probabilistic approaches will be considered to model these heterogeneous data [DeLeon2011, Marbac2017, Yilmaz2021].
References:
[Tursini2023] Katelyne Tursini, et al. "Subsequent and simultaneous electrophysiological investigation of the retina and the visual cortex in neurodegenerative and psychiatric diseases: what are the forecasts for the medicine of tomorrow?." Frontiers in Psychiatry 14 (2023): 1167654.
[Hébert2020] Marc Hébert et al. "The electroretinogram may differentiate schizophrenia from bipolar disorder." Biological psychiatry 87.3 (2020): 263-270.
[BIMAR] Projet PHRC BIMAR, "Étude des Troubles Bipolaires et Marqueurs électrophysiologiques Rétiniens" portée par le Pr. Thomas Schwitzer
[Ren2023] Xiaoxi Ren, Steven Le Cam, Ruggero G. Bettinardi, Katelyne Tursini, Thomas Schwitzer, Valérie Louis Dorr, "Discrimination entre sujets atteints de troubles bipolaires et sujet contrôles à l'aide de biomarqueurs ERG temps-fréquence", Journée Neurosciences Psychiatrie Neurologie (JNPN), juin 2023.
[Schwitzer2022] Thomas Schwitzer, Steven Le Cam et al. "Retinal electroretinogram features can detect depression state and treatment response in adults: A machine learning approach." Journal of Affective Disorders 306 (2022): 208-214.
[DeLeon2011] A. R. De Leon, A. Soo, and T. Williamson. "Classification with discrete and continuous variables via general mixed-data models." Journal of Applied Statistics 38.5 (2011): 1021-1032.
[Marbac2017] Matthieu Marbac, Christophe Biernacki, and Vincent Vandewalle. "Model-based clustering of Gaussian copulas for mixed data." Communications in Statistics-Theory and Methods 46.23 (2017): 11635-11656.
[Yilmaz2021] Yasin Yilmaz, Mehmet Aktukmak, and Alfred O. Hero. "Multimodal Data Fusion in High-Dimensional Heterogeneous Datasets Via Generative Models." IEEE Transactions on Signal Processing 69 (2021): 5175-5188.
Retinal and cortical electrophysiological measurements are relevant for the analysis of mental states explored in mental illness [Tursini2023]. Retinal function is assessed by electrophysiological techniques known as electroretinograms (ERGs). Herbert et al [Hébert2020] established that depressed patients had retinal responses that differed from controls, and that retinal responses differed between depressed patients without medication and depressed patients treated with pharmacotherapy. Analysis of electroencephalogram (EEG) data has also made it possible to extract objective biomarkers for mental disorders, and for bipolar disorder in particular [Tursini2023].
The first aim of this thesis is to use simultaneous EEG/ERG measurements to explore dysfunctions in visual systems from the retina to cortical responses in patients with bipolar disorder. Our analyses are based on flash (ffERG) or pattern (PERG) visual stimulation protocols leading to reproducible, well-controlled responses, and enabling the targeting of the cerebral network associated with vision. From a fundamental point of view, we want to identify a functional model for the propagation of information along the ventral visual pathway, and quantify the alteration in this propagation in the presence of bipolar disorders. This study will lead to the construction of connectivity graphs reflecting the underlying neural network, making it possible to assess the modification of this network in bipolar disorder and to quantify the effect of a therapeutic treatment on this network.
We also plan to search for reliable indicators of mental illness to bring objective tools for the diagnosis of these illnesses. We have already carried out a preliminary study in the case of a recruitment campaign for bipolar subjects [BIMAR], and robust time/frequency biomarkers have been identified that make it possible to discriminate these patients from a control population [Ren2023], but also showing an ability to quantify the evolution of the illness during treatment [Schwitzer2022]. The inclusion of quantitative measures derived from the connectivity graphs previously analysed will also be considered.
The diagnosis of mental illnesses is currently established by the clinician on the basis of clinical observations and qualitative data from neuropsychological surveys, giving rise to scores graded on scales of values. The third objective is to develop methods for combining quantitative measurements extracted from EEG/ERG signals with qualitative data of a discrete and ordinal nature, by modelling the relationships between these variables and their influence on the target (the label) to be estimated. Multivariate probabilistic approaches will be considered to model these heterogeneous data [DeLeon2011, Marbac2017, Yilmaz2021].
References:
[Tursini2023] Katelyne Tursini, et al. "Subsequent and simultaneous electrophysiological investigation of the retina and the visual cortex in neurodegenerative and psychiatric diseases: what are the forecasts for the medicine of tomorrow?." Frontiers in Psychiatry 14 (2023): 1167654.
[Hébert2020] Marc Hébert et al. "The electroretinogram may differentiate schizophrenia from bipolar disorder." Biological psychiatry 87.3 (2020): 263-270.
[BIMAR] Projet PHRC BIMAR, "Étude des Troubles Bipolaires et Marqueurs électrophysiologiques Rétiniens" portée par le Pr. Thomas Schwitzer
[Ren2023] Xiaoxi Ren, Steven Le Cam, Ruggero G. Bettinardi, Katelyne Tursini, Thomas Schwitzer, Valérie Louis Dorr, "Discrimination entre sujets atteints de troubles bipolaires et sujet contrôles à l'aide de biomarqueurs ERG temps-fréquence", Journée Neurosciences Psychiatrie Neurologie (JNPN), juin 2023.
[Schwitzer2022] Thomas Schwitzer, Steven Le Cam et al. "Retinal electroretinogram features can detect depression state and treatment response in adults: A machine learning approach." Journal of Affective Disorders 306 (2022): 208-214.
[DeLeon2011] A. R. De Leon, A. Soo, and T. Williamson. "Classification with discrete and continuous variables via general mixed-data models." Journal of Applied Statistics 38.5 (2011): 1021-1032.
[Marbac2017] Matthieu Marbac, Christophe Biernacki, and Vincent Vandewalle. "Model-based clustering of Gaussian copulas for mixed data." Communications in Statistics-Theory and Methods 46.23 (2017): 11635-11656.
[Yilmaz2021] Yasin Yilmaz, Mehmet Aktukmak, and Alfred O. Hero. "Multimodal Data Fusion in High-Dimensional Heterogeneous Datasets Via Generative Models." IEEE Transactions on Signal Processing 69 (2021): 5175-5188.
Keywords:
ERG/EEG, Bipolar disorder, Connectivity analysis, Multimodal/heterogeneous data fusion
Department(s):
| Biology, Signals and Systems in Cancer and Neuroscience |