Presented at SPIE Medical Imaging 2023 (Image Processing).

Work by Yu Zhou, Jingru Fu, Örjan Smedby, and Rodrigo Moreno.

Abstract

The amount of labelled data in children is much lower than that for adult subjects. This paper proposes a new method to synthesise high-quality pathological paediatric MRI brain images from pathological adult ones.

During the training phase, the whole process is divided into three steps:

• A. Predicting the mass effect induced by a tumour.
• B. Applying the predicted mass effect to the healthy images.
• C. Synthesising plausible tumour MRI scans from the deformed masks.

The method can automatically simulate the mass effect of the tumour given the location and shape of the pathology, producing augmented paediatric-cohort data for downstream segmentation models.

• Paper: SPIE Digital Library — DOI 10.1117/12.2654366