Shedding light on the black box of a neural network used to detect prostate cancer in whole slide images by occlusion-based explainability

AbbreviationsAOPC

Area Over the Perturbed Curve

ASAP

Automated Slide Analysis Platform

DeconvNet

Deconvolution Network

DTD

Deep Taylor Decomposition

FFPE

Formalin-fixed Paraffin-embedded

FROC

Free-response Receiver Operation Characteristics

LIME

Local Interpretable Model-Agnostic Explanations

LRP-z

Layer-Wise Relevance Propagation with Z-rule

LRP-ε

Layer-Wise Relevance Propagation with epsilon rule

MIL

Multiple-instance Learning

OSA

Occlusion Sensitivity Analysis

PDA

Prediction Difference Analysis

RISE

Randomized Input Sampling for Explanations

SHAP

Shapley Additive Explanations

TCAV

Testing with Concept Activation Vectors

xPattern

Explained Pattern

xPOI

Explanation Point of Interest

Keywords

Machine learning

Digital histopathology

Explainable AI

Artificial intelligence

Occlusion sensitivity analysis

Prostate cancer

© 2023 The Authors. Published by Elsevier B.V.

Comments (0)

No login
gif