Attention-guided sampling for colorectal cancer analysis with digital pathology.

Autor:	Broad A; School of Computing, University of Leeds, Sir William Henry Bragg Building, Woodhouse Lane, Leeds LS2 9BW, UK.; Leeds Institute for Data Analytics, University of Leeds, Level 11, Worsley Building, Clarendon Way, Leeds LS2 9NL, UK., Wright AI; Leeds Teaching Hospitals NHS Trust, Beckett St, Harehills, Leeds LS9 7TF, UK.; Division of Pathology and Data Analytics, Leeds Institute of Medical Research, University of Leeds, St James's University Hospital, Leeds LS9 7TF, UK., de Kamps M; School of Computing, University of Leeds, Sir William Henry Bragg Building, Woodhouse Lane, Leeds LS2 9BW, UK.; Leeds Institute for Data Analytics, University of Leeds, Level 11, Worsley Building, Clarendon Way, Leeds LS2 9NL, UK.; The Alan Turing Institute, 96 Euston Road, London NW1 2DB, UK., Treanor D; Leeds Teaching Hospitals NHS Trust, Beckett St, Harehills, Leeds LS9 7TF, UK.; University of Leeds, Leeds LS2 9JT, UK.; Department of Clinical Pathology, and Department of Clinical and Experimental Medicine, Linköping University, Linköping, Sweden.; Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, Sweden.
Jazyk:	angličtina
Zdroj:	Journal of pathology informatics [J Pathol Inform] 2022 Jun 24; Vol. 13, pp. 100110. Date of Electronic Publication: 2022 Jun 24 (Print Publication: 2022).
DOI:	10.1016/j.jpi.2022.100110
Abstrakt:	Improvements to patient care through the development of automated image analysis in pathology are restricted by the small image patch size that can be processed by convolutional neural networks (CNNs), when compared to the whole-slide image (WSI). Tile-by-tile processing across the entire WSI is slow and inefficient. While this may improve with future computing power, the technique remains vulnerable to noise from uninformative image areas. We propose a novel attention-inspired algorithm that selects image patches from informative parts of the WSI, first using a sparse randomised grid pattern, then iteratively re-sampling at higher density in regions where a CNN classifies patches as tumour . Subsequent uniform sampling across the enclosing region of interest (ROI) is used to mitigate sampling bias. Benchmarking tests informed the adoption of VGG19 as the main CNN architecture, with 79% classification accuracy. A further CNN was trained to separate false-positive normal epithelium from tumour epithelium, in a novel adaptation of a two-stage model used in brain imaging. These subsystems were combined in a processing pipeline to generate spatial distributions of classified patches from unseen WSIs. The ROI was predicted with a mean F1 (Dice) score of 86.6% over 100 evaluation WSIs. Several algorithms for evaluating tumour-stroma ratio (TSR) within the ROI were compared, giving a lowest root mean square (RMS) error of 11.3% relative to pathologists' annotations, against 13.5% for an equivalent tile-by-tile pipeline. Our pipeline processed WSIs between 3.3x and 6.3x faster than tile-by-tile processing. We propose our attention-based sampling pipeline as a useful tool for pathology researchers, with the further potential for incorporating additional diagnostic calculations. Competing Interests: The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Andrew Broad reports financial support was provided by Roche Tissue Diagnostics. Andrew Broad reports financial support was provided by UK Research and Innovation. (© 2022 The Authors.)
Databáze:	MEDLINE
Externí odkaz: