Development of an artificial intelligence-based algorithm for predicting the severity of myxomatous mitral valve disease from thoracic radiographs by using two grading systems.

Autor: Valente C; Department of Animal Medicine, Production and Health, University of Padua, Viale dell'Università 16, 35020, Legnaro, Padua, Italy. Electronic address: carlotta.valente@unipd.it., Wodzinski M; Department of Measurement and Electronics, AGH University of Kraków, Al. A. Mickiewicza 30, 30-059 Krakow, Poland; Information Systems Institute, University of Applied Sciences-Western Switzerland (HES-SO Valais), Rue de Technopôle 3, 3960 Sierre, Switzerland., Guglielmini C; Department of Animal Medicine, Production and Health, University of Padua, Viale dell'Università 16, 35020, Legnaro, Padua, Italy., Poser H; Department of Animal Medicine, Production and Health, University of Padua, Viale dell'Università 16, 35020, Legnaro, Padua, Italy., Chiavegato D; AniCura Arcella Veterinary Clinic, Via Cardinale Callegari 48, 35133 Padua, Italy., Zotti A; Department of Animal Medicine, Production and Health, University of Padua, Viale dell'Università 16, 35020, Legnaro, Padua, Italy., Venturini R; AniCura Arcella Veterinary Clinic, Via Cardinale Callegari 48, 35133 Padua, Italy., Banzato T; Department of Animal Medicine, Production and Health, University of Padua, Viale dell'Università 16, 35020, Legnaro, Padua, Italy.
Jazyk: angličtina
Zdroj: Research in veterinary science [Res Vet Sci] 2024 Oct; Vol. 178, pp. 105377. Date of Electronic Publication: 2024 Aug 08.
DOI: 10.1016/j.rvsc.2024.105377
Abstrakt: A heart-convolutional neural network (heart-CNN) was designed and tested for the automatic classification of chest radiographs in dogs affected by myxomatous mitral valve disease (MMVD) at different stages of disease severity. A retrospective and multicenter study was conducted. Lateral radiographs of dogs with concomitant X-ray and echocardiographic examination were selected from the internal databases of two institutions. Dogs were classified as healthy, B1, B2, C and D, based on American College of Veterinary Internal Medicine (ACVIM) guidelines, and as healthy, mild, moderate, severe and late stage, based on Mitral INsufficiency Echocardiographic (MINE) score. Heart-CNN performance was evaluated using confusion matrices, receiver operating characteristic curves, and t-SNE and UMAP analysis. The area under the curve (AUC) was 0.88, 0.88, 0.79, 0.89 and 0.84 for healthy and ACVIM stage B1, B2, C and D, respectively. According to the MINE score, the AUC was 0.90, 0.86, 0.71, 0.82 and 0.82 for healthy, mild, moderate, severe and late stage, respectively. The developed algorithm showed good accuracy in predicting MMVD stages based on both classification systems, proving a potentially useful tool in the early diagnosis of canine MMVD.
Competing Interests: Declaration of competing interest None.
(Copyright © 2024 The Authors. Published by Elsevier Ltd.. All rights reserved.)
Databáze: MEDLINE
Externí odkaz: