Empirical Study of Ovarian Tumors Detection and Classification in Ultrasound Images

Ovarian cancer is one of the leading causesof death in women. Ultrasound images often provide initialfindings and diagnosis before further tests are performed.Despite some progress in medical image analysis, the detectionand recognition of ovarian cancers remains still quite limited.Most existing works focused on detecting or classifying twotumor categories, namely benign and malignant lesions. Thispaper presents an initial comprehensive study of ovariantumor detection, and at the same time classifies eight classesof ovarian tumors in ultrasound images using deep-learningmodels. Due to the lack of training data, we implemented adata augmentation process and examined the improvementof performance with and without data augmentation. Experimentscarried out on the OTU-2D set of MMOTU dataset withYOLOv5, YOLOv7, and YOLOv7 variants show promisingdetection and classification results.