Kenta Kasahara1, Kenji Katsumata2, Akira Saito3, Tetsuo Ishizaki2, Masanobu Enomoto2, Junichi Mazaki2, Tomoya Tago2, Yuichi Nagakawa2, Jun Matsubayashi4, Toshitaka Nagao4, Hiroshi Hirano5, Masahiko Kuroda6, Akihiko Tsuchida2. 1. Department of Digestive and Pediatric Surgery, Tokyo Medical University, 6-7-1 Nishi Shinjuku, Shinjuku-ku, Tokyo, 160-0023, Japan. kasadog@tokyo-med.ac.jp. 2. Department of Digestive and Pediatric Surgery, Tokyo Medical University, 6-7-1 Nishi Shinjuku, Shinjuku-ku, Tokyo, 160-0023, Japan. 3. Department of AI Applied Quantitative Clinical Science, Tokyo Medical University, 6-7-1 Nishi Shinjuku, Shinjuku-ku, Tokyo, 160-0023, Japan. 4. Department of Anatomic Pathology, Tokyo Medical University, 6-7-1 Nishi Shinjuku, Shinjuku-ku, Tokyo, 160-0023, Japan. 5. Diagnostic Pathology Division, Tokyo Medical University Hachioji Medical Center, 1163 Tatemachi, Hachioji-shi, 193-0998, Tokyo, Japan. 6. Department of Molecular Pathology, Tokyo Medical University, 6-7-1 Nishi Shinjuku, Shinjuku-ku, Tokyo, 160-0023, Japan.
Abstract
BACKGROUND: The treatment strategies for colorectal cancer (CRC) must ensure a radical cure of cancer and prevent over/under treatment. Biopsy specimens used for the definitive diagnosis of T1 CRC were analyzed using artificial intelligence (AI) to construct a risk index for lymph node metastasis. METHODS: A total of 146 T1 CRC cases were analyzed. The specimens for analysis were mainly biopsy specimens, and in the absence of biopsy specimens, the mucosal layer of the surgical specimens was analyzed. The pathology slides for each case were digitally imaged, and the morphological features of cancer cell nuclei were extracted from the tissue images. First, statistical methods were used to analyze how well these features could predict lymph node metastasis risk. A lymph node metastasis risk model using AI was created based on these morphological features, and accuracy in test cases was verified. RESULTS: Each developed model could predict lymph node metastasis risk with a > 90% accuracy in each region of interest of the training cases. Lymph node metastasis risk was predicted with 81.8-86.3% accuracy for randomly validated cases, using a learning model with biopsy data. Moreover, no case with lymph node metastasis or lymph node risk was judged to have no risk using the same model. CONCLUSIONS: AI models suggest an association between biopsy specimens and lymph node metastases in T1 CRC and may contribute to increased accuracy of preoperative diagnosis.
BACKGROUND: The treatment strategies for colorectal cancer (CRC) must ensure a radical cure of cancer and prevent over/under treatment. Biopsy specimens used for the definitive diagnosis of T1 CRC were analyzed using artificial intelligence (AI) to construct a risk index for lymph node metastasis. METHODS: A total of 146 T1 CRC cases were analyzed. The specimens for analysis were mainly biopsy specimens, and in the absence of biopsy specimens, the mucosal layer of the surgical specimens was analyzed. The pathology slides for each case were digitally imaged, and the morphological features of cancer cell nuclei were extracted from the tissue images. First, statistical methods were used to analyze how well these features could predict lymph node metastasis risk. A lymph node metastasis risk model using AI was created based on these morphological features, and accuracy in test cases was verified. RESULTS: Each developed model could predict lymph node metastasis risk with a > 90% accuracy in each region of interest of the training cases. Lymph node metastasis risk was predicted with 81.8-86.3% accuracy for randomly validated cases, using a learning model with biopsy data. Moreover, no case with lymph node metastasis or lymph node risk was judged to have no risk using the same model. CONCLUSIONS: AI models suggest an association between biopsy specimens and lymph node metastases in T1 CRC and may contribute to increased accuracy of preoperative diagnosis.
Authors: Al B Benson; Alan P Venook; Lynette Cederquist; Emily Chan; Yi-Jen Chen; Harry S Cooper; Dustin Deming; Paul F Engstrom; Peter C Enzinger; Alessandro Fichera; Jean L Grem; Axel Grothey; Howard S Hochster; Sarah Hoffe; Steven Hunt; Ahmed Kamel; Natalie Kirilcuk; Smitha Krishnamurthi; Wells A Messersmith; Mary F Mulcahy; James D Murphy; Steven Nurkin; Leonard Saltz; Sunil Sharma; David Shibata; John M Skibber; Constantinos T Sofocleous; Elena M Stoffel; Eden Stotsky-Himelfarb; Christopher G Willett; Christina S Wu; Kristina M Gregory; Deborah Freedman-Cass Journal: J Natl Compr Canc Netw Date: 2017-03 Impact factor: 11.908
Authors: Ole-Johan Skrede; Sepp De Raedt; Andreas Kleppe; Tarjei S Hveem; Knut Liestøl; John Maddison; Hanne A Askautrud; Manohar Pradhan; John Arne Nesheim; Fritz Albregtsen; Inger Nina Farstad; Enric Domingo; David N Church; Arild Nesbakken; Neil A Shepherd; Ian Tomlinson; Rachel Kerr; Marco Novelli; David J Kerr; Håvard E Danielsen Journal: Lancet Date: 2020-02-01 Impact factor: 79.321
Authors: Mark C Hornbrook; Ran Goshen; Eran Choman; Maureen O'Keeffe-Rosetti; Yaron Kinar; Elizabeth G Liles; Kristal C Rust Journal: Dig Dis Sci Date: 2017-08-23 Impact factor: 3.199