SDSimon Duchesne

Simon Duchesne

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Simon Duchesne has a total of 535 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Incidence of Dementia and Subsequent Outcomes in Patients With Heart Failure

    by Paul Tieu, Raed A. Joundi, Finlay A. McAlister, Moira K. Kapral, Nadia Khan, Richard H. Swartz, Clare L. Atzema, Peter C. Austin, Philip A. Barber, Filio Billia, et al.

  • Open Access

    A scoping review of computational models on human glucose cerebral metabolism

    by Parissa Fereydouni-Forouzandeh, Andréanne Michaud, Nicolas Doyon, Simon Duchesne

    Understanding cerebral metabolism is an essential undertaking to address brain health and tackle ambiguities in associated neurodegenerative diseases. Despite the abundance of detailed observational findings on specific functionalities of metabolism, there remains a gap in holistic understanding that could provide a suitable predictive framework, and hence be put to clinical use. Computational models can support such wider goals. Accordingly, we performed a scoping review of computational models on human cerebral glucose metabolism on the PubMed database. We selected 15 models on which we conducted a qualitative assessment of Verification based on reproducibility and internal validity; Validation for computational and biological relevance, sensitivity analysis and external validity; and Evaluation , where we inquired whether the model increased domain knowledge, with the capacity to be extended for improved generalizability. We raised critical shortcomings, including overall poor reproducibility, lack of internal and external validity, and certain unclear metabolic assumptions that compromised biological relevance. We proposed that for a computational model to improve its applicability such as predicting metabolic state through the human lifespan, Verification, Validation and Evaluation (VV&E) issues raised should be addressed, which would facilitate the involvement of disciplines beyond mathematics.

  • Open Access

    Resting-state functional MRI derivatives: A dataset derived from the The Comprehensive Assessment of Neurodegeneration and Dementia Study

    by Natasha Clarke, Hao-Ting Wang, Désirée Lussier, Arnaud Boré, Loic Tetrel, Camille Beaudoin, Samir Das, Randi Pilon, Alan C. Evans, Howard Chertkow, et al.

    Resting-state functional connectivity (RSFC) holds promise for the detection and characterisation of dementia. The Comprehensive Assessment of Neurodegeneration and Dementia (COMPASS-ND) Study, by the Canadian Consortium on Neurodegeneration in Aging (CCNA), provides a unique resource to study deeply phenotyped neurodegenerative conditions. We present RSFC derivatives for 784 participants (data release 7 of the cohort) who were either cognitively unimpaired or diagnosed primarily with Alzheimer's disease (AD), mixed dementia (AD with a vascular component), mild cognitive impairment (MCI), vascular MCI, frontotemporal dementia, Parkinson's disease with or without MCI or dementia, Lewy body disease or subjective cognitive impairment. Functional MRI scans were preprocessed using fMRIPrep, and time-series and whole-brain connectomes generated using three atlases at multiple resolutions, denoised using seven different techniques. High-motion artifacts were managed using a liberal quality control threshold appropriate for an older clinical population, resulting in data from 680 participants. These derivatives are made available to the research community to accelerate research on RSFC biomarkers of neurodegenerative disease, reducing duplication of effort, saving computational resources, and improving standardisation across studies.

  • Open Access

    Predicting Episodes of Hypovigilance in Intensive Care Units Using Routine Physiological Parameters and Artificial Intelligence: Derivation Study

    by Raphaëlle Giguère, Victor Niaussat, Monia Noël-Hunter, William Witteman, Tanya S Paul, Alexandre Marois, Philippe Després, Simon Duchesne, Patrick M Archambault

    Background: Delirium is prevalent in intensive care units (ICUs), often leading to adverse outcomes. Hypoactive delirium is particularly difficult to detect. Despite the development of new tools, the timely identification of hypoactive delirium remains clinically challenging due to its dynamic nature, lack of human resources, lack of reliable monitoring tools, and subtle clinical signs including hypovigilance. Machine learning models could support the identification of hypoactive delirium episodes by better detecting episodes of hypovigilance. Objective: Develop an artificial intelligence prediction model capable of detecting hypovigilance events using routinely collected physiological data in the ICU. Methods: This derivation study was conducted using data from a prospective observational cohort of eligible patients admitted to the ICU in Lévis, Québec, Canada. We included patients admitted to the ICU between October 2021 and June 2022 who were aged ≥18 years and had an anticipated ICU stay of ≥48 hours. ICU nurses identified hypovigilant states every hour using the Richmond Agitation and Sedation Scale (RASS) or the Ramsay Sedation Scale (RSS). Routine vital signs (heart rate, respiratory rate, blood pressure, and oxygen saturation), as well as other physiological and clinical variables (premature ventricular contractions, intubation, use of sedative medication, and temperature), were automatically collected and stored using a CARESCAPE Gateway (General Electric) or manually collected (for sociodemographic characteristics and medication) through chart review. Time series were generated around hypovigilance episodes for analysis. Random Forest, XGBoost, and Light Gradient Boosting Machine classifiers were then used to detect hypovigilant episodes based on time series analysis. Hyperparameter optimization was performed using a random search in a 10-fold group-based cross-validation setup. To interpret the predictions of the best-performing models, we conducted a Shapley Additive Explanations (SHAP) analysis. We report the results of this study using the TRIPOD+AI (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis for machine learning models) guidelines, and potential biases were assessed using PROBAST (Prediction model Risk Of Bias ASsessment Tool). Results: Out of 136 potentially eligible participants, data from 30 patients (mean age 69 y, 63% male) were collected for analysis. Among all participants, 30% were admitted to the ICU for surgical reasons. Following data preprocessing, the study included 1493 hypovigilance episodes and 764 nonhypovigilant episodes. Among the 3 models evaluated, Light Gradient Boosting Machine demonstrated the best performance. It achieved an average accuracy of 68% to detect hypovigilant episodes, with a precision of 76%, a recall of 74%, an area under the curve (AUC) of 60%, and an F1-score of 69%. SHAP analysis revealed that intubation status, respiratory rate, and noninvasive systolic blood pressure were the primary drivers of the model's predictions. Conclusions: All classifiers produced precision and recall values that show potential for further development, with slightly different yet comparable performances in classifying hypovigilant episodes. Machine learning algorithms designed to detect hypovigilance have the potential to support early detection of hypoactive delirium in patients in the ICU.


  • Open Access

    A brain health framework with application to the study of neurodegeneration

    by Simon Duchesne, Olivier Potvin, Carol Hudon, Christian Bocti

    An all-encompassing framework seems necessary for understanding brain health in the context of neurodegeneration, particularly due to Alzheimer's disease (AD). We argue that current views, dominated by the amyloid-beta hypothesis, oversimplify the complexity of brain degeneration. We propose a multi-scale, multi-entity approach, treating the brain as a complex system composed of interacting subsystems across various scales, from the nanoscale (genes and proteins) to the macroscale (cognition) and beyond. By redefining brain health through mathematical complexity, we emphasize the importance of integrating diverse biological, environmental, and lifestyle factors to account for the heterogeneity in AD presentations. This framework suggests that failures at different levels of brain function can lead to cascading effects that influence neurodegenerative trajectories. Additionally, it calls for a shift toward personalized treatment approaches that target multiple yet specific pathological mechanisms affecting an individual. We propose computational models as essential tools for simulating and testing these complex interactions. Through this framework, we aim to provide a deeper understanding of neurodegeneration, and advocate for more comprehensive and multi-factorial approaches in both research and clinical practice to advance the treatment of brain health disorders such as AD.


  • Open Access

    Use of lecanemab and donanemab in the Canadian healthcare system: Evidence, challenges, and areas for future research

    by Eric E. Smith, Natalie A. Phillips, Howard H. Feldman, Michael Borrie, Aravind Ganesh, Alexandre Henri-Bhargava, Philippe Desmarais, Andrew Frank, AmanPreet Badhwar, Laura Barlow, et al.


  • Open Access

    Predicting cognitive function from multimodal mesoscale neuroimaging data

    by Simon Duchesne, Olivier Potvin, Louis Dieumegarde

    Background: Our objective was to assess individual and joint relationships between various mesoscale indicators of brain health (e.g., neuronal, metabolic, and vascular integrity) and cognitive function.Method: We used 7224 timepoint measures originating from 1825 participants of the ADNI study who were cognitively healthy or with mild cognitive impairment at initial assessment. Cognitive function outcomes were test results on the Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS‐Cog), the Rey auditory Verbal Learning Test immediate recall (RAVLTi), the Rey auditory Verbal Learning Test delayed recall (RAVLTd), the Trail Making Test A (Trail a), and the Trail Making Test B (Trail b) tasks. Four families of markers were used. First were three families of mesoscale neuroimaging assessed by MRI for 88 brain regions: 1. Neuronal integrity was estimated by FreeSurfer 6.0 produced volumes extracted from T1‐weighted (T1w) images; 2. Neurite integrity was estimated through T1w/T2‐weighted MRI ratios; 3. Cerebral metabolism integrity was measured by synthetic fluorodeoxyglucose‐positron emission tomography, generated by a model based on a latent‐space regularized generative adversarial net from T1w MRIs. Secondly, 4. Systemic markers of vascular health‐related factors were included in the model (history of cardiovascular diseases, alcohol abuse, smoking history, systolic and diastolic blood pressure, history of hypertension, type 2 diabetes, Hachinski score). Cross‐sectional and future cognitive outcomes were all predicted in a single machine learning model using all combinations of predictor families at baseline, time interval between outcomes and baseline as long as sex and education. Three types of regression models were tested (Ridge, Partial Least Squares and Support Vector).Result: Individually, neuronal and metabolism integrity were the best predictors of cognitive function (Fig1; R2: .15‐.35 and MAE in Z score: .08‐.23). The combination of two or more mesoscale brain health markers improved predictions, with the best combination being neuronal integrity with metabolism (Figs2‐3; mean R2: .28; mean MAE: .12). However, increasing the number of markers led to similar results.Conclusion: While the prediction of cognition improved with a combination of brain health mesoscale markers, a plateau seems to be reached after combining two families.

  • Open Access

    Predicting cognitive function from multimodal mesoscale neuroimaging data

    by Simon Duchesne, Olivier Potvin, Louis Dieumegarde

    Background: Our objective was to assess individual and joint relationships between various mesoscale indicators of brain health (e.g., neuronal, metabolic, and vascular integrity) and cognitive function.Method: We used 7224 timepoint measures originating from 1825 participants of the ADNI study who were cognitively healthy or with mild cognitive impairment at initial assessment. Cognitive function outcomes were test results on the Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS‐Cog), the Rey auditory Verbal Learning Test immediate recall (RAVLTi), the Rey auditory Verbal Learning Test delayed recall (RAVLTd), the Trail Making Test A (Trail a), and the Trail Making Test B (Trail b) tasks. Four families of markers were used. First were three families of mesoscale neuroimaging assessed by MRI for 88 brain regions: 1. Neuronal integrity was estimated by FreeSurfer 6.0 produced volumes extracted from T1‐weighted (T1w) images; 2. Neurite integrity was estimated through T1w/T2‐weighted MRI ratios; 3. Cerebral metabolism integrity was measured by synthetic fluorodeoxyglucose‐positron emission tomography, generated by a model based on a latent‐space regularized generative adversarial net from T1w MRIs. Secondly, 4. Systemic markers of vascular health‐related factors were included in the model (history of cardiovascular diseases, alcohol abuse, smoking history, systolic and diastolic blood pressure, history of hypertension, type 2 diabetes, Hachinski score).Cross‐sectional and future cognitive outcomes were all predicted in a single machine learning model using all combinations of predictor families at baseline, time interval between outcomes and baseline as long as sex and education. Three types of regression models were tested (Ridge, Partial Least Squares and Support Vector).Result: Individually, neuronal and metabolism integrity were the best predictors of cognitive function (Figure 1; R2: .15‐.35 and MAE in Z score: .08‐.23). The combination of two or more mesoscale brain health markers improved predictions, with the best combination being neuronal integrity with metabolism (Figures 2‐3; mean R2: .28; mean MAE: .12). However, increasing the number of markers led to similar results.Conclusion: While the prediction of cognition improved with a combination of brain health mesoscale markers, a plateau seems to be reached after combining two families.

  • Open Access

    A scoping review of the wide array of Alzheimer’s disease theories and hypotheses supports a multifactorial etiology

    by Simon Duchesne, Louis‐Simon Rousseau, Olivier Potvin, Florence Belzile, Laurie‐Ann Welch, Béatrice Cournoyer, Marianne Arseneau, Véronick Lapierre, Sara‐Maude Poulin, Carol Hudon

    Background: There is a common agreement that Alzheimer’s disease (AD) is inherently complex; otherwise, a general disagreement remains on its etiological underpinning, with numerous alternative hypotheses having been proposed. Our objective was to perform a scoping review of original manuscripts describing hypotheses and theories of AD published in past decades.Methods: We reviewed 127 original manuscripts that fulfilled our inclusion criteria out of more than 13,807 references extracted from open databases (from inception to 14 Sept 2023). Each entry was characterized as having a single or multifactorial focus and assigned to one of 15 theoretical groupings. Impact was tracked using open citation tools.Results: We found that three quarter of studies proposed a hypothesis characterized as being single‐focus (97/127), with the most important theoretical groupings being the Amyloid group, followed by Metabolism and Mitochondrial dysfunction, then Infection and Cerebrovascular. Three stages can be discerned in terms of hypotheses generation. The first phase (∼1980‐1995) included the establishment of the main thrusts that have endured to this day (Amyloid, Glial, Infection, Inflammation, Metabolism, Oxidative stress, and Proteinopathies hypotheses; multifactoriality; and neurotoxicity). In the second phase (1995‐2005), the importance of the Cerebrovasculature, Mitochondrial dysfunction, and Neurotransmitters were recognized. Lately (2005‐2023), evidence towards Genetics (outside of the autosomal dominant form of AD), and especially Gut/Brain interactions came to the fore. Impact, as measured by total citations count, followed a similar pattern. Out of a total of 19,996 citations for the 127 papers, the most cited single‐focused grouping was Amyloid (52%), followed by Mitochondrial dysfunction (6%, reaching 10.0% when adding Metabolism articles), then Cerebrovascular (5%) and Oxidative Stress (4%).Conclusion: When viewed together, these multi‐faceted reports reinforce the notion that AD affects multiple sub‐cellular, cellular, anatomical, and physiological systems at the same time but at varying degree between individuals. Future work should integrate these competing views into a comprehensive, systemic framework able to capture individual and combinatorial factor evolutions and interactions, to manage its inherent complexity.


  • Open Access

    A nano‐ to microscale, complex and multifactorial computational model of Alzheimer’s disease

    by Simon Duchesne, Éléonore Chamberland, Seyedadel Moravveji, Nicolas Doyon

    Background: While individual etiological hypotheses for AD are researched, few large‐scale theoretical integrative efforts linking entities involved in these dysfunctions have been attempted. Experimentally, assessing such a global theory is logistically near impossible to achieve as the number of variables is substantial. Alternatively, computational neuroscience allows for the joint study of multiple entities at this scale, the generation of predictions, and their validation with real data. In this work, we present a theoretical computational model describing the progression of AD through an individual’s lifespan covering more than 50 different entities.Method: Our computational model is a system of 19 ordinary differential equations (cf. Figure 1). The equations describe the evolution of proteins at the nanoscale (Aβ monomers, oligomers and plaques; tau filaments and tangles, anti‐inflammatory cytokines, insulin) and cell populations at the microscale (neurons, astrocytes, macrophages, microglia). Equations were parametrized according to sex and APOE status, and initial conditions chosen from the literature. Our key outcomes of interest were the accumulation of recognized pathological markers of AD (Aβ monomers or plaques, tau filaments or tangles) and neuronal death. The model was solved in daily increments over a 50‐year lifespan.Results: The evolution of every variable of our model is shown in Figure 2. For the different forms of amyloid, the curves for APOE4‐negative men and women mostly overlap. This is also observed for APOE4‐positive individuals. However, it is important to note that despite the overlap, there are noticeable variations in the curves between different groups. Neuronal loss occurs earlier in individuals with an APOE4 allele regardless of sex. We obtain losses of 10.9% and 11.7%, for APOE4‐ and APOE+ women respectively, and of 10.9% and 12.1% for APOE4‐ and APOE4+ men respectively, which are commensurate with the literature. We also observe a transition to a proinflammatory state occurring around 50 years of age, with women who are APOE4+ experience this transition earlier, followed by APOE4+ men, APOE4‐ women, and finally APOE4‐ men.Conclusion: Computational models represent an essential initial step toward constructing a complex, predictive framework, and hold significant potential for identifying effective therapeutic targets in the fight against AD.