Multiple Instance Learning: State-of-the-Art in Cancer Analysis, Applications & Future Potential

The digitalization of Whole Slide Images (WSIs) in histopathology facilitates the utilization of deep learning (DL) techniques for pathology research and computer-aided diagnosis in medicine. The multiple Instance Learning (MIL) technique is the most trustworthy and reliable method for medical image analysis especially WSI due to its high resolution, it has emerged as a promising approach in cancer analysis, offering unique advantages in handling complex and heterogeneous data. In this paper, we have explained recent techniques and discussed the application of MIL techniques for cancer detection, diagnosis, and prognosis prediction. It explores how MIL leverages the collective information from groups of instances, such as cells or tissue samples, to improve accuracy and robustness in cancer analysis. We also proposed an approach for ovarian cancer image analysis with the MIL technique. Key challenges, MIL in different cancer analyses, are addressed alongside promising future directions for advancing MIL in cancer research and clinical practice.