Browse Teach Curate

Generative Multi-Attribute Explanations for Prostate MRI Classifiers with Feature-Pyramid Encoders

How a generative model can explain a medical-image classifier by projecting a prostate-MRI scan into a generator's latent space and perturbing disentangled attributes to expose which image features drive the prediction. How adding feature pyramids to the encoder supplies multiscale feedback that refines the latent representation and sharpens the resulting visual explanations.

KITT Text

ProjectedEx: Enhancing Generation in Explainable AI for Prostate Cancer Xuyin Qi1,2,∗,†, Zeyu

Prostate-cancer diagnosis increasingly relies on multiparametric MRI together with deep-learning classifiers, but clinicians need to know why a model flags a lesion. Explainable-AI methods that use a…