A Comparative Study of Attention-Based MIL Architectures in Cancer Detection

This article compares two attention-based MIL models—AMIL and AdMIL—used in cancer detection. It examines their architectures, how they compute attention scores, and their implications for model interpretability in medical imaging. AdMIL improves patch-level explainability but narrows focus; a hybrid model is proposed to balance both.

Table of Links

Abstract and I. Introduction

2.2. Model Architectures

Due to the need for interpretability of the results, when choosing and building the models, we focused on MIL models that use an attention pooling mechanism. Since we also prioritized efficiency, we chose models with a relatively simple architecture. Specifically, we focused on the original attention MIL (AMIL), proposed by Ilse et al. [7] and the additive MIL (AdMIL), proposed by Javed et al. [8].

Ilse et al. proposed a weighted average of the instances as a MIL pooling operator, where the weights were trained on a ...