Yasunari Miyagi1,2,3, Katsuhiko Tada4, Ichiro Yasuhi5, Yuka Maekawa6, Naofumi Okura7, Kosuke Kawakami7, Ken Yamaguchi8,9, Masanobu Ogawa10,11, Takashi Kodama12, Makoto Nomiyama13, Tomoya Mizunoe14, Takahito Miyake15. 1. Medical Data Labo, Okayama, Japan. 2. Department of Gynecologic Oncology, Saitama Medical University International Medical Center, Saitama, Japan. 3. Miyake Ofuku Clinic, Okayama, Japan. 4. Department of Obstetrics and Gynecology, National Hospital Organization Okayama Medical Center, Okayama, Japan. 5. Department of Obstetrics and Gynecology, National Hospital Organization Nagasaki Medical Center, Omura, Japan. 6. Department of Obstetrics and Gynecology, National Hospital Organization Mie Chuo Medical Center, Tsu, Japan. 7. Department of Obstetrics and Gynecology, National Hospital Organization Kokura Medical Center, Kitakyushu, Japan. 8. Department of Gynecology and Obstetrics, Kyoto University Graduate School of Medicine, Kyoto, Japan. 9. Department of Obstetrics and Gynecology, National Hospital Organization Kyoto Medical Center, Kyoto, Japan. 10. Research Center for Environment and Developmental Medical Sciences, Kyusyu University, Fukuoka, Japan. 11. Department of Obstetrics and Gynecology/Clinical Research Institute, National Hospital Organization Kyusyu Medical Center, Fukuoka, Japan. 12. Department of Obstetrics and Gynecology, National Hospital Organization Higashihiroshima Medical Center, Higashihiroshima, Japan. 13. Department of Obstetrics and Gynecology, National Hospital Organization Saga National Hospital, Saga, Japan. 14. Department of Obstetrics and Gynecology, National Hospital Organization Kure Medical Center, Kure, Japan. 15. Department of Obstetrics and Gynecology, Miyake Clinic, Okayama, Japan.
Abstract
AIM: To investigate the feasibility of a novel method using artificial intelligence (AI), in which the fibrinogen criterion was determined by the quantitative relation between the distributions of fibrin/fibrinogen degradation products (FDPs) and fibrinogen. METHODS: A dataset of 154 deliveries comprising more than 2000 g of blood lost due to hemorrhage, excluding disseminated intravascular coagulation (DIC), among patients from eight national perinatal centers in Japan from 2011 to 2015 were obtained. The fibrinogen threshold criterion was identified by using the function that best fit the distributions of FDP as determined by AI. FDP production was described by differential equations using a dataset containing fibrinogen levels less than the fibrinogen criterion and solved numerically. RESULTS: A fibrinogen level of 237 mg/dL as the threshold criterion was obtained. The FDP threshold criteria were 2.0 and 8.5 mg/dL for no coagulopathy and a failed coagulation system, respectively. CONCLUSION: The fibrinogen threshold criterion for patients with massive hemorrhage excluding DIC at delivery were obtained by selecting the functions that best fit the distributions of FDP data by using AI.
AIM: To investigate the feasibility of a novel method using artificial intelligence (AI), in which the fibrinogen criterion was determined by the quantitative relation between the distributions of fibrin/fibrinogen degradation products (FDPs) and fibrinogen. METHODS: A dataset of 154 deliveries comprising more than 2000 g of blood lost due to hemorrhage, excluding disseminated intravascular coagulation (DIC), among patients from eight national perinatal centers in Japan from 2011 to 2015 were obtained. The fibrinogen threshold criterion was identified by using the function that best fit the distributions of FDP as determined by AI. FDP production was described by differential equations using a dataset containing fibrinogen levels less than the fibrinogen criterion and solved numerically. RESULTS: A fibrinogen level of 237 mg/dL as the threshold criterion was obtained. The FDP threshold criteria were 2.0 and 8.5 mg/dL for no coagulopathy and a failed coagulation system, respectively. CONCLUSION: The fibrinogen threshold criterion for patients with massive hemorrhage excluding DIC at delivery were obtained by selecting the functions that best fit the distributions of FDP data by using AI.
Authors: Ferdinand Dhombres; Jules Bonnard; Kévin Bailly; Paul Maurice; Aris T Papageorghiou; Jean-Marie Jouannic Journal: J Med Internet Res Date: 2022-04-20 Impact factor: 7.076