MedSegNet: Self-Supervised Medical Image Segmentation

Medical image segmentation plays a vital role in the diagnosis, treatment planning, and
monitoring of various diseases. However, most existing deep learning methods rely heavily on
large annotated datasets, which are often difficult and costly to obtain in the medical domain.
To overcome this limitation, MedSegNet employs a Self-Supervised Learning (SSL) approach
to achieve efficient medical image segmentation using limited labeled data. The model
leverages an autoencoder-based SSL framework for pretraining on unlabeled medical images,
learning strong visual representations. These pretrained features are then fine-tuned using a
small number of labeled scans for precise segmentation.
To ensure accessibility and real-world usability, the trained model is integrated into a Java Full

Stack Web Application. The backend, built with Spring Boot, communicates with a Python-
based AI service for inference, while the frontend (developed using React.js) enables users to

upload scans, visualize segmented results, and download reports. A database layer ensures
secure user management and storage of processed data.
Experimental evaluation using open-source medical datasets demonstrates improved accuracy
and segmentation quality with minimal labeled samples. The proposed system provides a
scalable and cost-effective solution for hospitals and researchers, reducing dependency on
large-scale annotations. Future enhancements include extending support for multi-modal
imaging (CT, MRI, X-ray) and developing a mobile version for clinical deployment.