Thyroid Nodule Ultrasound Image Segmentation

Project Code :TCMAPY1979

Objective

The project presents a comprehensive approach to thyroid nodule segmentation, including dataset preparation, training, and evaluation of deep learning models such as U-Net, Residual U-Net, UNet++, and Attention U-Net, and implementation of a Flask-based system. The application enables seamless image upload, automated model prediction, and clear visualization of segmentation results. The work demonstrates effective integration of these deep learning models with a user-friendly interface, efficient frontend–backend communication, and the development of a complete, reliable, and robust medical image segmentation system.

Abstract

This project focuses on segmenting thyroid nodules from ultrasound images using deep learning techniques. Accurate segmentation supports consistent interpretation of thyroid structures and assists in identifying regions of interest with greater clarity. The TN3K Thyroid Nodule Region Segmentation dataset is used to train and evaluate multiple neural network models, specifically U-Net, Residual U-Net, U-Net++, and Attention U-Net. Each model is examined for its ability to capture fine boundaries, preserve structural details, and handle variations in ultrasound image quality.

The system is built using Flask as the back-end framework and HTML, CSS, and JavaScript for the front-end interface. The platform allows users to register, log in, upload an ultrasound image, and obtain a segmented output through the chosen model. The goal is to design a structured, dependable, and user-friendly system that supports segmentation tasks with efficiency.

The project provides a comparative analysis of different segmentation architectures and highlights how architectural improvements influence performance. The study aims to support better understanding of thyroid ultrasound imagery and contribute to improved segmentation-based workflows.

Keywords: Thyroid, Segmentation, Ultrasound, U-Net, Residual U-Net, U-Net++, Attention U-Net, Deep Learning, Flask, TN3K Dataset

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.