Bladder tumor segmentation system in cystoscopic images implemented by Tri-scan enhanced images

2023 Fiscal Year Final Research Report

• Project/Area Number: 22K20509
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund
• Review Section: 0403:Biomedical engineering and related fields
• Research Institution: Kyushu University
• Principal Investigator: Mutaguchi Jun 九州大学, 大学病院, 助教 (80961929)
• Project Period (FY): 2022-08-31 – 2024-03-31
• Keywords: 膀胱癌 / 膀胱内視鏡 / 人工知能 / セグメンテーション

Outline of Final Research Achievements

Bladder cancer have high intravesical reccurence rate after transurethral resection of bladder tumors (TURBT), due to intraoperative oversight of the tumor or inadequate resection during TURBT. In this study, we used an artificial intelligence(AI)-based segmentation system and verified its usefulness for tumor detection. We improved the AI algorithm and created a new training image using an image processing technique defined as Tri-scan enhanced image. The improved AI algorithms were verified to be optimal algorithms, and training using these algorithms was performed on a high-performance computer to further improve accuracy. The system is now being implemented on a video to superimpose on the actual endoscope situations.

Free Research Field

泌尿器科学

Academic Significance and Societal Importance of the Research Achievements

膀胱癌は高齢者に多い疾患であり、高齢社会の本邦においては今後も症例数が増えることが予想される。早期の膀胱内再発の一因として、手術中の腫瘍の見落としや、不十分な切除が原因とされる。近年、人工知能によるセグメンテーションシステムを用いることで、腫瘍再発の抑制が可能と考え、これにより再手術を減らすことで、高齢者への不必要な侵襲を減らすことができる。またこれは、医療費の削減や膀胱癌再発率の低下につながり、高齢社会の本邦に於いても意義ある課題と考えられる。