SHAP-driven explicability of CNN-based computer-aided diagnosis for malaria

Loading...
Thumbnail Image

Date

Co-advisor(s)

Committee members

Graduate program

Journal Title

Journal ISSN

Volume Title

Publisher

Access rights

DOI

Resumo (Portuguese)

Redes Neurais Convolucionais (CNNs) têm grande potencial para classificação de imagens médicas, mas sua adoção clínica depende da confiabilidade dos modelos. Neste estudo, utilizamos Shapley Additive Explanations (SHAP) para explicar um sistema de CAD (Computer-Aided Diagnosis) baseado em CNN para diagnóstico de malária em imagens de microscopia. Aplicamos o Gradient Explainer para destacar as regiões de pixel mais relevantes para as previsões do CAD. Ao oferecer explicações transparentes em nível de pixel, esta abordagem capacita profissionais de saúde a compreenderem as decisões do sistema e fortalece a confiança no diagnóstico automatizado de malária.

Abstract

Convolutional Neural Networks (CNN) hold great promise for medical image classification, but their clinical adoption depends on model reliability. In this study, we leverage Shapley Additive Explanations (SHAP) to explain a CNN-based Computer-Aided Diagnosis (CAD) system for malaria detection from microscopy images. We employ SHAP’s Gradient Explainer to highlight key pixel regions driving the CAD model predictions. By providing transparent, pixel-level insights, this approach empowers healthcare professionals to understand system decisions and enhances trust in automated malaria diagnosis.

Description

Controlled term

Sustainable Development Goals (SDG)

Citation

Endorsement

Review

Supplemented By

Referenced By

Creative Commons license

Except where otherwise noted, this item's license is described as Acesso Aberto