Multi-task multi-scale learning for outcome prediction in 3D PET images

Amine Amyar (1, 2) , Romain Modzelewski (1, 3) , Pierre Vera (1, 3) , Vincent Morard (2) , Su Ruan (1)

1 QuantIF-LITIS - Equipe Quantification en Imagerie Fonctionnelle
2 GE Healthcare - General Electric Medical Systems [Buc]
3 CLCC Henri Becquerel - Centre de Lutte Contre le Cancer Henri Becquerel Normandie Rouen

Predicting patient response to treatment and survival in oncology is a prominent way towards precision medicine. To this end, radiomics has been proposed as a field of study where images are used instead of invasive methods. The first step in radiomic analysis in oncology is lesion segmentation. However, this task is time consuming and can be physician subjective. Automated tools based on supervised deep learning have made great progress in helping physicians. However, they are data hungry, and annotated data remains a major issue in the medical field where only a small subset of annotated images are available.

Mots clés: Radiomics; Deep learning; Image classification; Image segmentation; Multi-task learning; Positron Emission Tomography

Domaines: Sciences du Vivant [q-bio]

Amine Amyar, Romain Modzelewski, Pierre Vera, Vincent Morard, Su Ruan. Multi-task multi-scale learning for outcome prediction in 3D PET images. Computers in Biology and Medicine, 2022, 151 (Part A), pp.106208. ⟨10.1016/j.compbiomed.2022.106208⟩. ⟨hal-03842217⟩