Medmain Co., Ltd.
Succeeded in Developing Artificial Intelligence that Enables Detection of Prostate Cancer, a Target for Surveillance and Curative Treatment -Paper Published in BMC Cancer-
Medmain Co., Ltd. (Headquarters: Fukuoka City, Fukuoka Prefecture, CEO: Osamu Iizuka, hereinafter referred to as “Medmain”), which provides the digital pathology support solution “PidPort”, uses Deep Learning to improve prostate acupuncture. We have succeeded in developing artificial intelligence for highly accurate screening of prostate cancer, which is a potential target for curative treatment, in pathological tissue digital specimens.
We are pleased to announce that a paper on this study has been submitted and published in BMC Cancer, published by BioMed Central (BMC) (https://www.biomedcentral.com).
(Location: https://bmccancer.biomedcentral.com/articles/10.1186/s12885-022-10488-5)
[Image 1: https://prtimes.jp/i/34505/32/resize/d34505-32-235b0d62542bf325aa33-2.png&s3=34505-32-0d56f997a3a1c1367fd5cd980bca8578-1005×636.png] ■Summary of this research result We have successfully developed artificial intelligence that screens for prostate cancer, which is the target of radical treatment, in prostate needle biopsy
histopathological digital specimens. Background of this study When testing for prostate cancer at a hospital, PSA (prostate-specific antigen) was 4.0 ng/ml for those aged 40 and over, or PSA values by age group exceeded the reference value. At this time, a transrectal ultrasound-guided needle biopsy is performed, and a histopathological examination is performed to check for the presence of prostate cancer (References: Prostate Cancer Clinical Practice Guidelines, Prostate Cancer Screening Guidelines). If prostate cancer is detected, the patient’s treatment policy is determined by assessing prostate cancer, including risk classification. The Gleason classification is an important prognostic factor in prostate cancer. Prostate cancer is histologically characterized by the mixture of cancer components with different degrees of differentiation (high heterogeneity) within the sample. Determine the score and report it at the time of pathological diagnosis. Treatment strategies for prostate cancer can be broadly divided into two. One is called surveillance therapy, which is a method to prevent excessive treatment while observing the progress when it is judged that the cancer detected by prostate biopsy will not affect the life expectancy even if treatment is not started quietly. . The other is called radical treatment, such as surgery or
radiotherapy. Gleason pattern and Gleason score are related to indicators to judge whether it is surveillance therapy or radical treatment. Prostate cancer screening guidelines define pathologic criteria for active surveillance as a Gleason score of 6 or less and no more than 2 positive cores. However, in 2016, the American Society of Clinical Oncology (ASCO) endorsed the criteria for active surveillance advocated by the Cancer Care Ontario (CCO) group, stating, Surveillance therapy is recommended even if there are 4 = 7 cases.” New standards regarding the selection of surveillance therapy and radical treatment have been advocated internationally and have been applied in clinical practice (Reference: Journal of Clinical Oncology , 34: 2182-2190, 2016). Based on the above clinical background, in this research, we decided to use deep learning to develop artificial intelligence that can screen for prostate cancer, which is the target of active surveillance and radical treatment, in prostate needle biopsy histopathological digital specimens. . ■Details of this research Prostate needle biopsy histopathological specimens provided by domestic facilities were digitized, and supervised data including annotation data by multiple pathologists was created. Learning is transfer learning (Document: Proceedings of Machine Learning Research, 143: 338- 353, 2021) and fully supervised learning and weakly supervised learning to visualize prostate cancer, which is a potential target for active surveillance and definitive treatment, using a whole-slide image: We have developed an artificial
intelligence that can screen at the WSI level. We verified the accuracy of the developed artificial intelligence using verification data different from the training data. Results of this research When we compared and examined multiple models developed, we found that a model that performed fully supervised learning using pathologist’s annotation data in addition to weakly supervised learning was verified by the heat map method. Recognized areas of prostate cancer with the highest confidence and potential for active surveillance and definitive therapy. The ROC-AUC at the virtual slide level was 0.846 for prostate cancer targeted for active surveillance and 0.980 for prostate cancer targeted for definitive therapy. In addition, the suspected areas of prostate cancer identified by artificial
intelligence displayed by the heat map and subject to active surveillance and definitive treatment were verified by multiple pathologists and confirmed to be valid. . Based on the above, we have succeeded in developing an artificial intelligence that is capable of highly accurate screening of prostate cancer, which is the target of radical treatment, in prostate needle biopsy pathological tissue digital specimens.
[Image 2: https://prtimes.jp/i/34505/32/resize/d34505-32-6613114b49c66b9ed12f-0.png&s3=34505-32-5304d86231b436fb7037a6530a2a3295-1920×1280.png] In the future, we will proceed with verification tests of the deep learning artificial intelligence model developed this time at multiple facilities and large-scale cases.
■ Original article ▼ Article title: Inference of core needle biopsy whole slide images requiring definitive therapy for prostate cancer ▼DOI: https://doi.org/10.1186/s12885-022-10488-5
Author/Affiliation-Tochigi Prefectural Cancer Center Department of Pathology-Shin Abe-Sapporo Kosei Hospital Department of Pathology Senior Director – Makoto Ichihara – Medmain Co., Ltd. – Masayuki Tsuneki, Fahdi Kanavati ■ Company profile [Company name] Medmain Inc. *Ministry of Economy, Trade and Industry J-START UP selected company https://www.j-startup.go.jp/startups [Establishment date] January 11, 2018 [Business description] Planning, development, operation and sales of medical software and cloud services [Representative Director/CEO] Osamu Iizuka [Location] [Tokyo office] 2-10-11 Minami-Aoyama, Minato-ku, Tokyo Office] 2-4-5 Akasaka, Chuo-ku, Fukuoka City, Fukuoka Prefecture Chatelet Succeeds 104 Various related sites [Corporate site] https://medmain.com [Product site] https://service.medmain.com Contact information Medmain Co., Ltd. Public Relations:
pr-m@medmain.com
Details about this release:
https://prtimes.jp/main/html/rd/p/000000032.000034505.html
MAIL:cr@prtimes.co.jp