Speaker
Result
The experimental model achieved a test accuracy of 88.57% and a macro F1-score of 83.60% on the selected proxy dataset. The Low proxy-risk category achieved an F1-score of 0.99, while the High category achieved 0.68, indicating differences in classification performance. These results demonstrate the feasibility of the proposed multimodal framework for experimental analysis. However, they do not establish CTE detection capability or clinical validity. Further research requires genuine CTE-related datasets, reliable patient-level data, and clinical validation.
Aim & Purpose
This study aims to develop a proof-of-concept multimodal AI framework for CTE-related proxy-risk assessment. It explores combining MRI images, radiology reports, clinical, cognitive, and psychological data to identify relevant patterns and evaluate the feasibility of multimodal analysis using available proxy datasets, without claiming to diagnose CTE clinically.
Discussion (methods + interpretation)
The study used publicly available research datasets from Boston University and Kaggle, including MRI images, radiology reports, and other relevant information. Boston University provided access following a formal request and approval process that took approximately one month. Python, PyTorch, Pandas, NumPy, and Scikit-learn were used for data preprocessing and model development. CNN, machine learning, and NLP techniques were explored to analyze and combine available data modalities. The model classified records into project-defined Low, Moderate, and High proxy-risk categories. Since the datasets contain no confirmed CTE cases, the approach represents a proof-of-concept rather than a clinically validated CTE assessment system.
Background
Chronic Traumatic Encephalopathy (CTE) is a neurodegenerative disease associated with repetitive head impacts. Its assessment is challenging because its symptoms overlap with other neurological and psychological conditions, while definitive confirmation requires post-mortem neuropathological examination. This study explores multimodal AI for CTE-related proxy-risk assessment using Alzheimer’s MRI, stroke, dementia, and depression datasets, none of which contains confirmed CTE cases.
| Email address | iamtehreem012@gmail.com |
|---|---|
| Phone number | 03200271791 |