Cross-scale multi-instance learning for pathological image diagnosis.

Autor: Deng R; Vanderbilt University, Nashville, TN 37215, USA., Cui C; Vanderbilt University, Nashville, TN 37215, USA., Remedios LW; Vanderbilt University, Nashville, TN 37215, USA., Bao S; Vanderbilt University, Nashville, TN 37215, USA., Womick RM; The University of North Carolina at Chapel Hill, Chapel Hill, NC 27514, USA., Chiron S; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Li J; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Roland JT; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Lau KS; Vanderbilt University, Nashville, TN 37215, USA., Liu Q; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Wilson KT; Vanderbilt University Medical Center, Nashville, TN 37232, USA; Veterans Affairs Tennessee Valley Healthcare System, Nashville, TN 37212, USA., Wang Y; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Coburn LA; Vanderbilt University Medical Center, Nashville, TN 37232, USA; Veterans Affairs Tennessee Valley Healthcare System, Nashville, TN 37212, USA., Landman BA; Vanderbilt University, Nashville, TN 37215, USA; Vanderbilt University Medical Center, Nashville, TN 37232, USA., Huo Y; Vanderbilt University, Nashville, TN 37215, USA. Electronic address: yuankai.huo@vanderbilt.edu.
Jazyk: angličtina
Zdroj: Medical image analysis [Med Image Anal] 2024 May; Vol. 94, pp. 103124. Date of Electronic Publication: 2024 Feb 27.
DOI: 10.1016/j.media.2024.103124
Abstrakt: Analyzing high resolution whole slide images (WSIs) with regard to information across multiple scales poses a significant challenge in digital pathology. Multi-instance learning (MIL) is a common solution for working with high resolution images by classifying bags of objects (i.e. sets of smaller image patches). However, such processing is typically performed at a single scale (e.g., 20× magnification) of WSIs, disregarding the vital inter-scale information that is key to diagnoses by human pathologists. In this study, we propose a novel cross-scale MIL algorithm to explicitly aggregate inter-scale relationships into a single MIL network for pathological image diagnosis. The contribution of this paper is three-fold: (1) A novel cross-scale MIL (CS-MIL) algorithm that integrates the multi-scale information and the inter-scale relationships is proposed; (2) A toy dataset with scale-specific morphological features is created and released to examine and visualize differential cross-scale attention; (3) Superior performance on both in-house and public datasets is demonstrated by our simple cross-scale MIL strategy. The official implementation is publicly available at https://github.com/hrlblab/CS-MIL.
Competing Interests: Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Keith T. Wilson reports financial support was provided by Leona M and Harry B Helmsley Charitable Trust. Bennett A. Landman reports financial support was provided by National Science Foundation. Keith T. Wilson reports financial support was provided by US Department of Veterans Affairs. Yuankai Huo reports financial support was provided by National Institute of Health. Ken S. Lau reports financial support was provided by National Institutes of Health.
(Copyright © 2024 The Author(s). Published by Elsevier B.V. All rights reserved.)
Databáze: MEDLINE
Externí odkaz: