20–21 Nov 2026
Online Conference
Asia/Karachi timezone

Multimodel Early Risk Assessment of Chronic Traumatic Encephalopathy (CTE)

20 Nov 2026, 21:15
5m
Online Conference

Online Conference

Online Conference

Speaker

Sadia Tehreem

Discussion (methods + interpretation)

Five publicly available data sources were used: an Alzheimer’s MRI image dataset, 350 DICOM files with derived metadata (brain_mri_reports), the clinical dataset (stroke_prediction.csv, 5,110 records), the cognitive dataset (dementia_patient_health.csv, 1,000 records), and the psychological dataset (depression_anxiety.csv, 783 records). Python-based tools and libraries, including PyTorch, Pandas, NumPy, and Scikit-learn, were used for preprocessing, feature extraction, multimodal integration, and model evaluation. Since confirmed CTE labels and patient-level matching across all datasets were unavailable, project-defined proxy labels were used to classify records into Low, Moderate, and High categories.

Background

Chronic Traumatic Encephalopathy (CTE) is a neurodegenerative disease associated with repetitive head impacts and can currently be confirmed only through post-mortem neuropathological examination. Its overlapping cognitive, behavioural, and psychological symptoms make research assessment challenging. This study explores a multimodal artificial intelligence (AI) approach to investigate patterns across different data sources relevant to CTE-related research.

Result

The proposed model achieved 88.57% test accuracy and a macro F1-score of 83.60%. The Low proxy-risk category achieved an F1-score of 0.99, while the High category achieved 0.68, indicating comparatively weaker classification performance for the High category.

Aim & Purpose

This study aims to develop a proof-of-concept multimodal AI framework that integrates imaging, DICOM-derived metadata, clinical, cognitive, and psychological information to explore CTE-related proxy-risk assessment. It investigates whether combining different data modalities can provide useful experimental insights beyond analysing individual data sources.

Email address iamtehreem012@gmail.com
Phone number 03200271791

Author

Presentation materials