Mesial temporal lobe epilepsy (MTLE) is the most common type of focal refractory epilepsy in adults. Traditional analysis of positron emission tomography (PET) imaging heavily relies on subjective interpretation by physicians, making it challenging to accurately delineate lesion boundaries. This subjectivity often compromises the precision of surgical planning. Furthermore, when the PET signal differences between the two temporal lobes are subtle, accurate diagnosis becomes even more difficult. This forces many patients to undergo invasive intracranial electrode implantation to obtain clearer information about the epileptic focus.

To address these challenges, a research team led by Associate Professor Syu-Jyun Peng at Taipei Medical Universityhas successfully developed an innovative AI-driven PET Quantitative Localization System. By leveraging artificial intelligence (AI), this system offers a groundbreaking solution to the long-standing issue of precise localization of epileptic lesions.

AI-Powered Precision: Advancing Epilepsy Surgery Success

The innovative system, developed by the Taipei Medical University research team, integrates magnetic resonance imaging (MRI) and PET imaging with advanced machine learning algorithms to perform in-depth analysis. The core techniques include:

• Image Preprocessing: Enhancing PET image quality through nonlinear spatial alignment and grayscale intensity normalization.
• Brain Region Segmentation: Using the automated anatomical labeling (AAL) atlas to segment the brain and precisely identify regions of interest.
• Feature Extraction: Calculating standardized uptake values (SUVs) for each brain region and extracting lateralization indices (LIs) and other features.
• Machine Learning Classification: Feeding extracted features into machine learning models to train classifiers that accurately determine the side of the epileptic lesion.

By converting subjective evaluations into objective quantitative analyses, this system significantly improves the accuracy of lesion localization. Its automated analysis also reduces manual operations by physicians, enhancing efficiency. Remarkably, the system achieved a 100% accuracy rate in lesion localization in test datasets, highlighting its immense market potential.

Interview with Associate Professor Peng Syu-Jyun: AI’s Role in Transforming Epilepsy Treatment

In an exclusive interview, Associate Professor Syu-Jyun Peng emphasized that this breakthrough represents a significant milestone in epilepsy treatment. By incorporating machine learning technology, the research team developed a more objective and precise tool for lesion localization, which not only increases the success rate of epilepsy surgeries but also improves the quality of life for patients.

Dr. Peng highlighted the substantial market potential of the PET Quantitative Localization System. The team plans to actively pursue industry-academic collaborations to commercialize the technology into medical devices and introduce it to the international market. This tool can assist physicians in diagnosing and treating epilepsy with greater accuracy and may also have applications in diagnosing other neurological disorders, such as Alzheimer’s disease and Parkinson’s disease, benefiting a broader range of patients.

Resource: 癲癇病灶難判斷 AI加持正子攝影協助量化定位