📘 Publications

Here are some key projects and publications I’ve worked on, with brief explanations and links to the full articles.

Thesis

- PhD thesis: MRI-Based Texture Analysis for Characterizing Cerebral Degeneration Progression in Amyotrophic Lateral Sclerosis

University of Alberta, 2024

Summary: Amyotrophic lateral sclerosis (ALS) is a highly heterogeneous, progressive neurodegenerative disease with a median survival of three years from the onset of symptoms. As a multi-system disease, ALS is not limited to the motor system and includes degeneration of other cerebral regions, including frontotemporal areas. The complex heterogeneity of the disease is linked to a multitude of contributing factors, such as polygenic profiles, the site of initial symptoms, the rates of disease progression, and the progressive spread of pathological processes. To achieve a deeper understanding of the disease and take a step forward toward personalized medicine, magnetic resonance imaging (MRI) provides the potential to capture the cerebral neurodegenerative process, in vivo. The overall objective of this thesis is to leverage MRI-based texture analysis, an image processing technique that quantifies intuitive features of an image, to delineate the progression of cerebral degeneration in ALS. Through a series of aims, the thesis seeks to utilize texture analysis on structural MR images to assess changes in gray and white matter structures based on a model of disease progression, detect longitudinal cerebral changes and evaluate the clinical classification of patients, analyze degeneration within specific neural tracts, and classify ALS patients based on cerebral degeneration patterns. This multifaceted investigation endeavors to provide a deeper understanding of ALS complex pathology and to pave the pathway toward developing personalized medicine strategies. In the first experiment, texture analysis of T1-weighted images of ALS patients revealed an association between textural feature autocorrelation in both gray and white matter and disease accumulation, independent of disease aggressiveness. This study demonstrated the method’s capability as a unimodal tool for assessing disease accumulation in ALS. The second experiment indicated that clinical trial criteria for patient selection do not accurately reflect the degree of cerebral pathology, in vivo. Disease progression rate and the King’s staging system showed a better correlation with cerebral pathology, both at baseline and longitudinally. Analysis supporting the early involvement of the corticospinal tract following by changes in the frontotemporal regions as the disease progresses. The third experiment conducted a detailed analysis of white matter tracts in ALS, including the corticospinal tract and corpus callosum, and detected early and progressive changes in the bilateral corticospinal tract followed by later changes in the middle and anterior parts of the corpus callosum. This study supported the potential of texture analysis for detailed disease monitoring. Finally, the last experiment identified two subgroups of patients based on the disease propagation pattern in gray matter regions of the brain. This multimodal study utilized texture analysis of T1-weighted and T2-weighted MRI scans, along with cortical thickness measurements. Comparisons across modalities highlighted the importance of T1-weighted texture in diagnosis and T2-weighted texture in the subtyping of patients with ALS. In summary, this thesis has revealed important insights into the progression of gray and white matter deterioration in ALS brains and their association with clinical measures. It highlights the early involvement of the corticospinal tract in the disease, followed by a pattern of frontotemporal spread, rather than a deterioration within motor regions over the same period. The texture analysis of T1-weighted images demonstrated the limitations of current clinical classifications of ALS patients in reflecting the extent of cerebral pathology. Texture abnormalities in the corticospinal tract and corpus callosum were associated with clinical indicators of upper motor neuron (UMN) dysfunction, highlighting their significance as dependable markers for UMN pathology. This thesis also reveals the capability of simple structural MRI scans through texture analysis to unravel the disease’s heterogeneity.

Journal Articles

1. T1-weighted MRI texture analysis in amyotrophic lateral sclerosis patients stratified by the D50 progression model

Brain Communications, 2024

Summary:
This study explores the potential of texture analysis on T1-weighted MRI images to reveal unique insights into the diverse progression patterns of ALS. We pushed the limits of texture analysis on 1.5 Tesla T1-weighted scans, stratifying patients by disease severity and stage using the D50 disease progression model. This model, which captures the sigmoidal trajectory of functional decline, allowed us to differentiate ALS progression with greater precision. Our findings indicate that texture analysis could be a promising biomarker for staging ALS.

2. Mismatch between clinically defined classification of ALS stage and the burden of cerebral pathology

Journal of Neurology, 2024

Summary:
In our recent article, my colleagues and I embarked on a detailed investigation of the clinical stages of amyotrophic lateral sclerosis (ALS) and the extent of cerebral pathology detected by texture analysis of T1-weighted brain images.

3. Accurate personalized survival prediction for amyotrophic lateral sclerosis patients

Scientific Reports, 2023

Summary:
Our new article delves into an innovative approach aimed at enhancing the management of ALS. It introduces a machine-learned tool designed to accurately predict the survival time of ALS patients, a critical factor in strategizing future treatments and care.

This tool uniquely integrates clinical features with cortical thickness data derived from brain magnetic resonance (MR) images, offering a novel method to estimate the time until a composite respiratory failure event in ALS patients. The article presents the prediction as individual survival distributions (ISDs), which provide a probIn the realm of neurodegenerative diseases, Amyotrophic Lateral Sclerosis (ALS) stands out for its rapid progression and the significant challenges it poses in patient care and treatment planning. Our new article delves into an innovative approach aimed at enhancing the management of ALS. It introduces a machine-learned tool designed to accurately predict the survival time of ALS patients, a critical factor in strategizing future treatments and care.

This tool uniquely integrates clinical features with cortical thickness data derived from brain magnetic resonance (MR) images, offering a novel method to estimate the time until a composite respiratory failure event in ALS patients. The article presents the prediction as individual survival distributions (ISDs), which provide a probabilistic view of survival at various future time points for each patient. This approach represents a significant advancement in the predictive modeling for ALS and stands as a beacon of hope for patients and clinicians in tailoring more effective treatment plans.abilistic view of survival at various future time points for each patient. This approach represents a significant advancement in the predictive modeling for ALS and stands as a beacon of hope for patients and clinicians in tailoring more effective treatment plans.

4. Effects of MRI scanner manufacturers in classification tasks with deep learning models

Scientific Reports, 2023

Summary:
This study explores the challenges of domain shift in deep learning applications for medical imaging, specifically focusing on multi-center MRI data from different scanner manufacturers (GE, Philips, and Siemens). Please check it out for further details.

Conference Abstracts

Podium Presentations

Poster Presentations