Segment STN

PythonIAHealthImage Processing3D images

Conv-Setr model for 3D segmentation in MRIs, published research on Springer Nature journal


The substantia nigra and subthalamic nucleus are vital components in the treatment of Parkinson’s disease (PD); however, they are difficult to identify on T2-weighted magnetic resonance imaging (MRI).

Segmenting the subthalamic nucleus is particularly important because treatments such as ablations, DBS and HIFU require precise localization of this element.

Automatic segmentation of the subthalamic nucleus subthalamic nucleus and substantia nigra using Deep Learning techniques in MRI for the monitoring of the progression and treatment of Parkinson’s disease.

So, my final degree project was develop a Deep Learning modal to achieve that objective. In result, I create a transformer-based approach (Conv-Setr) achieved the segmentation of volumes of interest with a segmentation of the volumes of interest with a DICE coefficient of 0.81 and AVD of 0.06. These results were contrasted with other architectures,such as CLCI and U-Net. However, these architectures did not perform at the level of the SETR-based architecture.

As result, my research was published on Springer Nature journal at Information Systems and Technologies conference. You can see my research and get the chapter here.

Tech used