Almășan Oana, Leucuța Daniel-Corneliu, Hedeșiu Mihaela, Mureșanu Sorana, Popa Ștefan Lucian
Department of Prosthetic Dentistry and Dental Materials, Iuliu Hațieganu University of Medicine and Pharmacy, 400006 Cluj-Napoca, Romania.
Department of Medical Informatics and Biostatistics, Iuliu Hațieganu University of Medicine and Pharmacy, 400349 Cluj-Napoca, Romania.
J Clin Med. 2023 Jan 25;12(3):942. doi: 10.3390/jcm12030942.
The aim was to systematically synthesize the current research and influence of artificial intelligence (AI) models on temporomandibular joint (TMJ) osteoarthritis (OA) diagnosis using cone-beam computed tomography (CBCT) or panoramic radiography. Seven databases (PubMed, Embase, Scopus, Web of Science, LILACS, ProQuest, and SpringerLink) were searched for TMJ OA and AI articles. We used QUADAS-2 to assess the risk of bias, while with MI-CLAIM we checked the minimum information about clinical artificial intelligence modeling. Two hundred and three records were identified, out of which seven were included, amounting to 10,077 TMJ images. Three studies focused on the diagnosis of TMJ OA using panoramic radiography with various transfer learning models (ResNet model) on which the meta-analysis was performed. The pooled sensitivity was 0.76 (95% CI 0.35-0.95) and the specificity was 0.79 (95% CI 0.75-0.83). The other studies investigated the 3D shape of the condyle and disease classification observed on CBCT images, as well as the numerous radiomics features that can be combined with clinical and proteomic data to investigate the most effective models and promising features for the diagnosis of TMJ OA. The accuracy of the methods was nearly equivalent; it was higher when the indeterminate diagnosis was excluded or when fine-tuning was used.
目的是系统地综合人工智能（AI）模型对使用锥形束计算机断层扫描（CBCT）或全景X线摄影的颞下颌关节（TMJ）骨关节炎（OA）诊断的当前研究及影响。我们在七个数据库（PubMed、Embase、Scopus、Web of Science、LILACS、ProQuest和SpringerLink）中检索了有关TMJ OA和AI的文章。我们使用QUADAS-2评估偏倚风险，同时使用MI-CLAIM检查有关临床人工智能建模的最低信息。共识别出203条记录，其中7条被纳入，共计10,077张TMJ图像。三项研究聚焦于使用全景X线摄影和各种迁移学习模型（ResNet模型）诊断TMJ OA，并在此基础上进行了荟萃分析。合并敏感度为0.76（95%CI 0.35 - 0.95），特异度为0.79（95%CI 0.75 - 0.83）。其他研究调查了髁突的三维形状以及在CBCT图像上观察到的疾病分类，以及众多可与临床和蛋白质组数据相结合的放射组学特征，以研究诊断TMJ OA的最有效模型和有前景的特征。这些方法的准确性几乎相当；排除不确定诊断或使用微调时准确性更高。
