Automated Gastrointestinal Tract Image Segmentation

The number of people acquiring a gastrointestinal (GI) cancer diagnosis is increasing annually. In the realm of biomedicine, radiation treatment is the accepted gold standard for treating gastrointestinal cancer. An X-ray beam is directed towards the tumour during radiation therapy, sparing the nearby healthy tissue. For an oncologist, manually segmenting healthy organs can be laborious and prone to mistakes. Thus, there is a need for technology that can autonomously segregate the healthy organs of the GI tract. In response to this demand, our work suggests an innovative approach that uses ensemble learning to enhance model performance. Two different deep-learning architectures are used to create the ensemble: the DenseNet and InceptionResNetv2 models. The main goal is to identify healthy organs inside the GI tract and offer a more reliable and efficient segmentation method than manual segmentation, ultimately contributing to more effective radiation therapy treatments for patients diagnosed with GI cancer. Radio oncologists can greatly accelerate the treatment process and provide faster, more precise treatments by utilizing an automated system. By utilizing the GI Tract Image Segmentation dataset from UW-Madison, the suggested approach was implemented, guaranteeing a solid basis for training and testing. By combining the benefits of both InceptionResNetv2 and DenseNet201, ensemble learning seeks to provide a model that is more reliable and accurate for the given job. The key strengths of this approach lie in its robustness, scalability, and computational efficiency. The model attained a commendable Dice coefficient of 0.89, indicating a high level of accuracy in organ segmentation. This indicates the possibility for a smooth integration of the suggested deep learning-based solution into clinical processes in addition to proving its effectiveness.