Comparative Analysis of Prostate MRI Image Segmentation: Watershed, U-Net, SegNet, and SegFormer Approaches
In this project, we explore the performance of different approaches to the segmentation of the central gland and peripheral zones of the prostate. We include older approaches such as the Watershed algorithm up to state of the art visual transformers such as fine-tuning a SegFormer model, which achieves close to state-of-the-art performance.
Segmentation example
Segfromer predictions
Results
The Dataset used is from the NCI-ISBI 2013 Challenge: Automated Segmentation of Prostate Structures (ISBI-MR-Prostate-2013)
You can find the dataset at: NCI-ISBI 2013 Challenge dataset