脳卒中試験における計算人工知能を用いた予後予測とリスク評価：FORECAST研究プロトコル

Stroke remains a leading cause of disability and mortality in Bangladesh, with limited data on risk stratification models tailored to the local population. Current prediction models inadequately capture the complex interplay of socioeconomic, environmental, and mental health factors prevalent in Bangladesh. This study protocol was designed to develop an artificial intelligence (AI)-based risk prediction model for stroke occurrence and poor long-term outcomes, specifically for the Bangladeshi population. The Foresighting Outcomes and Risk Evaluation with Computational Artificial Intelligence in Stroke Trials (FORECAST) study is an ambispective case-control study that will enroll 4,000 participants (2,000 stroke cases and 2,000 controls) from the National Institute of Neurosciences and Hospital, Bangladesh, between May 2022 and April 2026. We will develop ensemble machine learning models using comprehensive phenotypic data, including sociodemographic variables, comorbidities, anxiety/depression scores, and clinical parameters. The study will employ explainable AI techniques to enhance clinical interpretability and validation through k-fold cross-validation methods. The FORECAST study will establish the first AI-based stroke prediction model for Bangladesh, accounting for unique local risk factors, including rural-urban disparities and mental health comorbidities, with expected accuracy exceeding 90% based on recent advances in stroke prediction modeling. This study addresses critical gaps in stroke risk assessment for Bangladesh and similar low-middle-income countries by incorporating culturally relevant risk factors into advanced machine learning frameworks.