How can data augmentation improve attribution maps for disease subtype explainability? - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2023

How can data augmentation improve attribution maps for disease subtype explainability?

Abstract

As deep learning has been widely used for computer aided-diagnosis, we wished to know whether attribution maps obtained using gradient back-propagation could correctly highlight the patterns of disease subtypes discovered by a deep learning classifier. As the correctness of attribution maps is difficult to evaluate directly on medical images, we used synthetic data mimicking the difference between brain MRI of controls and demented patients to design more reliable evaluation criteria of attribution maps. We demonstrated that attribution maps may mix the regions associated with different subtypes for small data sets while they could accurately characterize both subtypes using a large data set. We then proposed simple data augmentation techniques and showed that they could improve the coherence of the explanations for a small data set.
Fichier principal
Vignette du fichier
Thibeau_Sutre_SPIE_2023.pdf (8.94 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03966737 , version 1 (01-02-2023)

Identifiers

  • HAL Id : hal-03966737 , version 1

Cite

Elina Thibeau-Sutre, Jelmer M Wolterink, Olivier Colliot, Ninon Burgos. How can data augmentation improve attribution maps for disease subtype explainability?. SPIE Medical Imaging, Feb 2023, San Diego, United States. ⟨hal-03966737⟩
54 View
41 Download

Share

Gmail Facebook X LinkedIn More