Thyroid Papillary Carcinoma Segmentation in Ultrasound Image using a MP-Former based Model

Papillary Thyroid Carcinoma (PTC) is a very common type of thyroid cancer. Patients can attain a relatively high survival rate if diagnosed and treated promptly. Hence, the early diagnosis and treatment of PTC are of crucial significance, and PTC segmentation in ultrasound images holds significant research value. In recent years, with the advancement of artificial intelligence (AI), deep learning has demonstrated outstanding performance in computer-aided diagnosis (CAD). Nevertheless, the research efforts on PTC segmentation based on deep learning are extremely limited. To enrich the research domain, we propose a PTC automatic segmentation method based on MP-Former (Mask-Piloted Transformer). In this paper, the PTC segmentation capabilities of MP-Former and other deep learning models have been verified. Experimental results show that the MP-Former model has obtained satisfactory PTC segmentation result with weighted crossover ratio (fwIoU) value of 92.60%, which reveals that MP-Former has the potential to become an effective ultrasound image segmentation for PTC, capable of assisting clinicians in diagnosing and treating PTC more accurately. To the best of our knowledge, this is the first study that utilizes the MP-Former-based model to segment PTC in ultrasound images, and this work can offer a reference for subsequent research in this field.