Statistical Testing for Multiple Instance Learning via Selective Inference with Applications to Computational Pathology
Abstract
Multiple instance learning (MIL) is widely used in computational pathology because it enables weakly supervised analysis of whole-slide images (WSIs) without requiring patch-level annotations. In attention-based MIL, instances with high attention scores are often interpreted as diagnostically important regions and used as visual explanations. However, attention scores alone cannot determine whether selected high-attention instances are significantly different from normal instances, limiting the reliability of attention-based explanations. In this paper, we formulate the evaluation of high-attention instances as a statistical hypothesis testing problem. Specifically, we assess whether a selected high-attention instance significantly deviates from a representative normal reference instance selected based on feature similarity. A major challenge is that both the target instance and the reference instance are selected through data-dependent procedures, rendering standard hypothesis testing invalid. To address this issue, we introduce a selective inference (SI) framework that explicitly accounts for the selection events induced by attention-based instance selection and adaptive reference selection, thereby enabling the computation of valid selective p-values conditional on these events. Experiments demonstrate Type-I error control on synthetic and MNIST-based data and practical applicability to pathological WSIs, with higher statistical power than the conventional over-conditioning approach.
Related
- Statistical inference and applications of Kumaraswamy Odd Perks power Lomax distribution
- Spectral Analysis of Gram Matrices with Missing at Random Observations: Convergence, Central Limit Theorems, and Applications in Statistical Inference
- The Wrapped Q-Lindley Distribution: A New Circular Model with Statistical Properties, Classical and Bayesian Inference, and Applications to Directional Data
- Second-Order Least Squares Method for Statistical Inference in Linear Models with Asymmetric Errors
- Riemannian Newton Methods with Statistical Inference
BibTeX
@misc{hashimoto2026statistical,
title = {{Statistical Testing for Multiple Instance Learning via Selective Inference with Applications to Computational Pathology}},
author = {Noriaki Hashimoto and Shuichi Nishino and Teruyuki Katsuoka and Tomohiro Shiraishi and Daiki Miwa and Hiroyuki Hanada and Jun Sakuma and Hidekata Hontani and Hiroaki Miyoshi and Ichiro Takeuchi},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.31712},
eprint = {2609.31712},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.31712},
}
RIS
TY - UNPB
TI - Statistical Testing for Multiple Instance Learning via Selective Inference with Applications to Computational Pathology
AU - Noriaki Hashimoto
AU - Shuichi Nishino
AU - Teruyuki Katsuoka
AU - Tomohiro Shiraishi
AU - Daiki Miwa
AU - Hiroyuki Hanada
AU - Jun Sakuma
AU - Hidekata Hontani
AU - Hiroaki Miyoshi
AU - Ichiro Takeuchi
PY - 2026
DA - 2026/09/21
PB - arXiv
DO - 10.48550/arxiv.2609.31712
UR - https://arxiv.org/abs/2609.31712
AB - Multiple instance learning (MIL) is widely used in computational pathology because it enables weakly supervised analysis of whole-slide images (WSIs) without requiring patch-level annotations. In attention-based MIL, instances with high attention scores are often interpreted as diagnostically important regions and used as visual explanations. However, attention scores alone cannot determine whether selected high-attention instances are significantly different from normal instances, limiting the reliability of attention-based explanations. In this paper, we formulate the evaluation of high-attention instances as a statistical hypothesis testing problem. Specifically, we assess whether a selected high-attention instance significantly deviates from a representative normal reference instance selected based on feature similarity. A major challenge is that both the target instance and the reference instance are selected through data-dependent procedures, rendering standard hypothesis testing invalid. To address this issue, we introduce a selective inference (SI) framework that explicitly accounts for the selection events induced by attention-based instance selection and adaptive reference selection, thereby enabling the computation of valid selective p-values conditional on these events. Experiments demonstrate Type-I error control on synthetic and MNIST-based data and practical applicability to pathological WSIs, with higher statistical power than the conventional over-conditioning approach.
ER -