MSTT-199: MRI Dataset for Musculoskeletal Soft Tissue Tumor Segmentation

TL;DR

This paper introduces a new MRI dataset of 199 musculoskeletal soft tissue tumors, trains segmentation models on it, and benchmarks their performance, achieving state-of-the-art results and highlighting challenges with certain tumor types.

Contribution

The authors collected and publicly shared a large, annotated MRI dataset for soft tissue tumor segmentation and demonstrated its utility by training and benchmarking models with state-of-the-art performance.

Findings

Model achieved a dice score of 0.79 without fine tuning.

Performance decreased on fibrous and vascular tumors.

Dataset diversity contributed to robust model training.

Abstract

Accurate musculoskeletal soft tissue tumor segmentation is vital for assessing tumor size, location, diagnosis, and response to treatment, thereby influencing patient outcomes. However, segmentation of these tumors requires clinical expertise, and an automated segmentation model would save valuable time for both clinician and patient. Training an automatic model requires a large dataset of annotated images. In this work, we describe the collection of an MR imaging dataset of 199 musculoskeletal soft tissue tumors from 199 patients. We trained segmentation models on this dataset and then benchmarked them on a publicly available dataset. Our model achieved the state-of-the-art dice score of 0.79 out of the box without any fine tuning, which shows the diversity and utility of our curated dataset. We analyzed the model predictions and found that its performance suffered on fibrous and vascular tumors due to their diverse anatomical location, size, and intensity heterogeneity. The code and models are available in the following github repository, https://github.com/Reasat/mstt