前立腺癌の局在診断・治療支援のための臓器抽出手法に関する研究

抄録

In this study, to develop an AI-assisted navigation system for prostate cancer localization and tracking technology, we investigated a prostate and tumor segmentation method on TRUS and MRI images using YOLACT++. The results showed that YOLACT++ achieved very high accuracy in prostate segmentation on TRUS and MRI images, and its effectiveness was confirmed. On the other hand, tumor segmentation performed only on MRI images was often difficult to detect, and even when contour extraction could be performed, it could not be done with high accuracy. In the future, we will construct a system to identify the location and spread of tumors by comprehensively examining multiple MRI imaging modalities. In addition, we aim to integrate and implement a registration system between MRI and TRUS images into the automatic segmentation system we have considered this time.