International Journal of Scientific Research and Engineering Development

Ovarian cancer is one of the deadliest gynecological cancers because it is mostly diagnosed in its late stages and the heterogeneity of the tumor is complicated. Diagnosis and subtype-identification of ovarian cancer is still the gold standard of histopathology or Hematoxylin and Eosin (H&E) stained tissue slide. Nevertheless, the process of manually reviewing Whole Slide Images (WSIs) is time consuming and prone to inter-observer variation among pathologists. Recent developments in computational pathology and deep learning provide prospects of automated histopathological analysis. This paper suggests a Multi-Instance Learning (MIL) model of automated classification of the subtypes of ovarian cancer based on H&E-stained WSIs. In the suggested model, every WSI is represented as a bag of image patches, and the model will learn discriminative morphological patterns based on slide-level labels without the need of pixel-level annotations. The framework combines the use of the GPU to preprocess tissue, extract patches, feature embedding with the help of a pathology foundation model and aggregation with a transformer to obtain slide-level predictions. Attention-based aggregation is used to model spatial relationships between tissue patches to increase the performance of classification. The experimental findings indicate that the proposed framework attains a high classification accuracy and at the same time, it is computationally efficient to be implemented in the clinical environment. The results indicate potential of MIL-based deep learning methods to assist pathologists in automated classification of the subtypes of ovarian cancer and improve the diagnostic processes in the digital pathology.

Sachini Amani Henda Vitharana, Deshan Sachintha Kannangara,"A Multi-Instance Learning Framework for Automated Ovarian Cancer Subtype Classification from Whole Slide Images" International Journal of Scientific Research and Engineering Development, V9(3): Page(715-723) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.