Neurofibromatosis Tumor Segmentation on Whole-body MRI (WBMRI-NF) Challenge

Motivation

Neurofibromatosis (NF), including NF1, NF2, and schwannomatosis, is a group of genetic disorders that cause tumors to grow on nerve tissues. NFs can occur anywhere in the body from a handful of tumors to hundreds of tumors and are broadly classified as discrete (involving a single nerve) or plexiform (involving multiple nerve fascicles).

Whole-Body MRI (WB-MRI) has emerged as the standard non-invasive imaging modality for longitudinal monitoring of NF patients, early detection of NF1-related malignancy, planning of surgical and/or oncologic treatments, and assessment of treatment response in NF-related clinical care and clinical trials.

The technical challenges for automated segmentation of NF tumors on WBMRI are their high variation in tumor location, size, shape, infiltration, and heterogeneity, particularly in plexiform neurofibromas. Our results show that the widely-adopted U-Net model and its variants (such as nnU-Net, UNet++, etc.) exhibit significantly lower accuracy in the segmentation of NFs on WBMRI, when compared to other MRI segmentation tasks.

To address these technical challenges for automated volumetric quantification of NFs on WBMRI, we are developing this benchmarking Challenge to identify novel solutions of deep-learning based segmentation algorithms, with the long-term goal of fostering a research community focused on WBMRI image analysis techniques.

This Challenge will offer the first large-scale WBMRI dataset for NF tumors, and task participants to develop models for automated segmentation and detection of NF tumors on WBMRI. This Challenge will promote image analysis techniques on WBMRI, improve the clinical care of NF patients and the clinical trial evaluation of NF drugs, bring more attention and innovative treatments for rare disease like NF, and bridge interdisciplinary communication between researchers and practitioners of medical image analysis, computer vision, and machine learning, to investigate novel solutions to overcome the technical barriers.