Glottic lesion segmentation of computed tomography images using deep learning

The larynx, a common site for head and neck cancers, is often overlooked in automated contouring due to its small size anatomically complex nature. More than 75% of laryngeal tumors originate the glottis. This paper proposes method automatically delineate glottic present contrast computed tomography (CT) images neck. A novel dataset 340 with was acquired pre-processed, senior radiologist created detailed, manual slice-by-slice tumor annotation. An efficient deep-learning architecture, U-Net, modified trained on our segment automatically. then visualized corresponding ground truth. Using combined metric dice score binary cross-entropy, we obtained an overlap 86.68% train set 82.67% test set. results are comparable limited work done this area. paper’s novelty lies compiled impressive data. Limited research has been detection diagnosis cancers. Automating segmentation process while ensuring malignancies not essential saving clinician’s time.