An Automatic ROI Detection in Ultrasound Images for Thyroid Cancer
T. Veda Reddy, Shadab Siddiqui · 2024
Thyroid cancer cases seem to be increasing rapidly across the globe. The best way to get detailed information about thyroid is ultrasound image. By analyzing ultrasound image features an expert can advise whether the nodule may be cancer or benign. Recently, artificial intelligence algorithms employed not only for efficient early diagnosis of thyroid cancer but also for over diagnosis. The paper proposes an automatic region of interest detection in ultra-sonic images using intensity and edge based segmentation. Intensity based segmentation is an image processing approach which divides an image into distinct regions based on intensity values of the pixels. Next edge-based segmentation is employed which detects edges within the image that might define anatomical shapes. For this a canny edge detector is used to identify potential boundaries of the ROI. The extracted ROI is evaluated using support vector machine algorithm to classify whether the extracted nodule features are benign or malignant. The proposed method achieves an accuracy of 98.8% and 97.4% in training and test phases respectively. The experimental results are also compared with existing literature work.