سامانه همایش‌ها TSTA | ثبت‌نام و ارسال مقاله کنفرانس‌ها و کنگره‌های ملی

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

The Adaptive Approach of Ensemble Deep Learning Model in OCT Image Classification

نویسندگان :

Hamed Aghapanah Roudsari¹ Ali Ghaderian² Mrteza Choubin³

1- دانشگاه علوم پزشکی اصفهان، دانشکده فناوری‌های نوین پزشکی 2- دانشگاه ملایر 3- دانشگاه ملایر

کلمات کلیدی :

Adaptive Ensemble Models،Deep Learning،OCT Images

چکیده :

Currently, there exists a global population of over 2.2 billion individuals with visual impairments, among which at least 1 billion could potentially avoid or treat their vision-related issues. The domain of eye care encounters significant challenges on a worldwide scale. Disparities persist in the accessibility, quality, and reach of treatment and rehabilitation services. Integrating eye care within healthcare systems and a scarcity of skilled eye care professionals compound these challenges. The early screening of fundus images offers an economical and accessible means of preventing blindness stemming from ocular disorders. Manual diagnostic approaches are time-consuming and can lead to delayed treatment due to limited medical resources. The advent of deep learning has yielded promising outcomes in the study of eye diseases, although most of these endeavors have focused on specific conditions. Pioneering research demonstrates the cost-effectiveness and efficacy of early eye diagnosis for mitigating blindness due to conditions such as diabetes, glaucoma, and cataracts. This study employs various deep learning architectures, including VGG16, ResNet50, and MobileNetV3Small, showcasing their synergistic integration to outperform individual models. The fundus image recognition task entails categorizing images into four classes, achieving an average recognition accuracy of 96.3%, and precision of 95.63%.