Skip to main content
Fig. 2 | World Journal of Surgical Oncology

Fig. 2

From: Automatic thyroid nodule recognition and diagnosis in ultrasound imaging with the YOLOv2 neural network

Fig. 2

a, b We analyzed the performance of benign and malignant nodules target detectors using the Precision-Recall curve. In order to get the PR curve, we set the IOU between the bounding box and ground truth greater than 0.3 to be true positive, otherwise false positive. The AP value is used to evaluate the performance of the detector. The AP value can be obtained by calculating the area under the PR curve. As can be seen from the PR curve, the AP values of malignant and benign nodules were 87.94% and 83.90% respectively, and the average (mAP) of the two types of AP was 85.92%

Back to article page