Deep Learning With Conformal Prediction for Hierarchical Analysis of Large-Scale Whole-Slide Tissue Images

Abstract

With the increasing amount of image data collected from biomedical experiments there is an urgent need for smarter and more effective analysis methods. Many scientific questions require analysis of image subregions related to some specific biology. Finding such regions of interest (ROIs) at low resolution and limiting the data subjected to final quantification at full resolution can reduce computational requirements and save time. In this paper we propose a three-step pipeline: First, bounding boxes for ROIs are located at low resolution. Next, ROIs are subjected to semantic segmentation into sub-regions at mid-resolution. We also estimate the confidence of the segmented sub-regions. Finally, quantitative measurements are extracted at full resolution. We use deep learning for the first two steps in the pipeline and conformal prediction for confidence assessment. We show that limiting final quantitative analysis to sub-regions with full confidence reduces noise and increases separability of observed biological effects.

Translating polygenic risk scores for clinical use by estimating the confidence bounds of risk prediction

“…The predicted errors for the two models were compared in the hold-off testing set (Methods). In line with previous studies on CP 27 , 29 , the prediction errors from MCCP were perfectly aligned with those expected across all simulated scenarios (Fig. 2 and Supplementary Figures 1 – 4 ).…”

Rapid development of cloud-native intelligent data pipelines for scientific data streams using the HASTE Toolkit

“…For example in a classi cation setting the class labels predicted can be highly uncertain. If in the top tier of the hierarchy we would place only those data points for which we are con dent in the predicted label, downstream analysis would see a reduction in noise and an increased separability of the (biological) e ects under study, as discussed in [27].…”

Editorial Computational Pathology

“…The last set of papers by Mormont et al [20] Stenman et al [21] and Wieslander et al [22] present novel AI-based technology to support development of future computational pathology systems for decision support. [20], the authors present a valid alternative to commonly used DL models pre-trained on ImageNet.…”

“…Finally, [22] the authors proposed a hierarchical approach to the analysis of WSI, to tackle the problem of memory limitation due to the size of digital pathology images. Focusing on COPD and idiopathic pulmonary fibrosis (IPF) and on quantifying region-specific drug response in lung tissue, they first decompose the slide into multiple regions of interest, detected automatically using CNNs at low-resolution, followed by segmentation of multiple morphological components at a medium level of resolution.…”