Assistant Professor Yi-Tsen Lin
|
National Taiwan University Hospital, Taiwan
Dr. Yi-Tsen Lin is an assistant professor in the Department of Otolaryngology at the National Taiwan University Hospital and College of Medicine at National Taiwan University. Dr. Lin has over a decade of experience in clinical practice and research. She specializes in rhinology, with a focus on the endotypes and clinical features of chronic rhinosinusitis with nasal polyps (CRSwNP), the impact of biologics, and the application of AI in rhinology.
Dr. Lin earned her medical degree from the National Taiwan University College of Medicine, followed by a master's degree and a Ph.D. from the Institute of Clinical Medicine at National Taiwan University. She has completed observerships in Korea and the United States, and has held a visiting scholarship at Stanford University. Dr. Lin currently serves as Deputy General Secretary of the Taiwan Rhinology Society and is a member of several international organizations, including the American Rhinology Society and the North American Skull Base Society. |
|
AI in Endotype Diagnosis of Chronic Rhinosinusitis: From Eosinophil Quantification to Predicting Disease Phenotypes
The diagnosis and classification of eosinophilic chronic rhinosinusitis (ECRS) is crucial for treatment decision-making and prognosis prediction. We developed two AI-based approaches to enhance the diagnosis and prediction of ECRS. The first, BREATHE (Boosted Rhinosinusitis Evaluation Algorithm Through Haematology and Ethmoid-Maxillary Analysis), analyses computed tomography (CT) imaging and hematologic markers, such as the percentage of eosinophils and the eosinophil-to-neutrophil ratio (ENR), to predict ECRS. Data from 158 patients (102 with ECRS and 56 without) were processed using machine learning models. This approach achieved an area under the receiver operating characteristic curve of 0.901, demonstrating high sensitivity (85.0%) and specificity (90.9%). The second approach focused on developing an automated system that uses deep learning and multi-instance learning to quantify eosinophils in the sinonasal histopathological images. This system achieved a validation performance of 87.63%. By comparing the experimental results and pathology reports, we found that the sensitivity was 86.2% and the specificity was 63.6%. These AI models significantly improve the accuracy of diagnosing ECRS, offering predictive capabilities and automated eosinophil quantification. Combining these technologies could improve the clinical management of CRS by providing timely, non invasive predictions and supporting personalized treatment strategies.
