Abstract—This paper proposes a diagnostic support method that simultaneously detects and tracks abdominal tumors from moving camera images taken by ultrasonography. Ultrasound diagnosis requires a doctor’s experience because it is necessary to operate the equipment and perform the diagnosis simultaneously. Also, since the purpose of ultrasonic diagnosis is to find the presence of tumors, it is necessary to avoid overlooking them. We aim to develop a diagnostic support system that detects tumor cross-sections in each input frame using YOLOv3 and then tracks them using DeepSORT considering the instances of detected tumors. Here, even if a tumor disappears in an input frame due to doctor’s probe manipulation, it can be identified as the same instance between the input frames and finally be grouped into the same cluster. If diagnostic support information can be ensembled for each cluster, it is expected that the accuracy of tumor detection should be improved. Experimental results showed that our system could perform the clustering of tumor cross-sections in real-time.

Index Terms—liver tumor detection, tracking, ultrasound image, convolutional neural network

Cite: Shoya Yamagishi, Keisuke Doman, Yoshito Mekada, Naoshi Nishida, and Masatoshi Kudo, "Detection and Tracking of Liver Tumors for Ultrasound Diagnostic Support Using Deep Learning," Journal of Image and Graphics, Vol. 10, No. 1, pp. 50-55, March 2022.

Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.