DEEP LEARNING TO AUTOMATICALLY CLASSIFY VERY LARGE SETS OF PREOPERATIVE AND POSTOPERATIVE SHOULDER ARTHROPLASTY RADIOGRAPHS.

Researchers

Journal

Modalities

Models

Abstract

Joint arthroplasty registries usually lack information on medical imaging due to the laborious process of observing and recording, and the lack of standard methods to transfer the imaging information to the registries, which can limit the investigation of various research questions. Artificial intelligence (AI) algorithms can automate imaging-feature identification with high accuracy and efficiency. With the purpose of enriching shoulder arthroplasty registries with organized imaging information, it was hypothesized that an automated AI algorithm could be developed to classify and organize pre- and postoperative radiographs from shoulder arthroplasty patients according to laterality, radiographic projection, and implant type.A cohort of 2,303 shoulder radiographs from 1,724 shoulder arthroplasty patients was used. Two observers manually labeled all radiographs according to 1) laterality (left, right), 2) projection (anterior-posterior, axillary, lateral), and 3) whether the radiograph was preoperative or showed an anatomic total shoulder arthroplasty (aTSA), or a reverse shoulder arthroplasty (RSA). All these labeled radiographs were randomly split into developmental and testing sets at the patient level and based on stratification. Using 10-fold cross-validation, a three-task deep learning algorithm was trained on the developmental set to classify the three characteristics mentioned. The trained algorithm was then evaluated on the testing set using quantitative metrics and visual evaluation techniques.The trained algorithm perfectly classified the laterality (F1 scores of 100% on the testing set). When classifying the imaging projection, the algorithm achieved F1 scores of 99.2%, 100%, and 100% on anterior-posterior, axillary, and lateral views, respectively. When classifying the implant type, the model achieved F1 scores of 100%, 95.2%, and 100% on preoperative, aTSA, and RSA radiographs, respectively. Visual evaluation using integrated maps showed that the algorithm focused on the relevant patient body and prosthesis parts for classifications. It took the algorithm 20.3 seconds to analyze 502 images.We developed an efficient, accurate, and reliable AI algorithm to automatically identify key imaging features of laterality, imaging view, and implant types in shoulder radiographs. This algorithm represents the first step to automatically classify and organize shoulder radiographs on a large scale in very little time, which will profoundly enrich shoulder arthroplasty registries.Copyright © 2023. Published by Elsevier Inc.