{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,31]],"date-time":"2026-10-31T22:13:53Z","timestamp":1793484833073,"version":"4.5.0"},"reference-count":28,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,2,5]],"date-time":"2025-02-05T00:00:00Z","timestamp":1738713600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec><jats:title>Purpose<\/jats:title><jats:p>This study aims to classify Kellgren\u2013Lawrence (KL) osteoarthritis stages using knee anteroposterior X-ray images by comparing two deep learning (DL) methodologies: a traditional single-model approach and a proposed multi-model approach. We addressed three core research questions in this study: (1) How effective are single-model and multi-model deep learning approaches in classifying KL stages? (2) How do seven convolutional neural network (CNN) architectures perform across four distinct deep learning tasks? (3) What is the impact of CLAHE (Contrast Limited Adaptive Histogram Equalization) on classification performance?<\/jats:p><\/jats:sec><jats:sec><jats:title>Approach<\/jats:title><jats:p>We created a dataset of 14,607 annotated knee AP X-rays from three hospitals. The knee joint region was isolated using a YOLOv5 object detection model. The multi-model approach utilized three DL models: one for osteophyte detection, another for joint space narrowing analysis, and a third to combine these outputs with demographic and image data for KL classification. The single-model approach directly classified KL stages as a benchmark. Seven CNN architectures (NfNet-F0\/F1, EfficientNet-B0\/B3, Inception-ResNet-v2, VGG16) were trained with and without CLAHE augmentation.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>The single-model approach achieved an F1-score of 0.763 and accuracy of 0.767, outperforming the multi-model strategy, which scored 0.736 and 0.740. Different models performed best across tasks, underscoring the need for task-specific architecture selection. CLAHE negatively impacted most models, with only one showing a marginal improvement of 0.3%.<\/jats:p><\/jats:sec><jats:sec><jats:title>Conclusion<\/jats:title><jats:p>The single-model approach was more effective for KL grading, surpassing metrics in existing literature. These findings emphasize the importance of task-specific architectures and preprocessing. Future studies should explore ensemble modeling, advanced augmentations, and clinical validation to enhance applicability.<\/jats:p><\/jats:sec>","DOI":"10.3389\/frai.2025.1413820","type":"journal-article","created":{"date-parts":[[2025,2,5]],"date-time":"2025-02-05T07:31:55Z","timestamp":1738740715000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":8,"title":["Deep learning in gonarthrosis classification: a comparative study of model architectures and single vs. multi-model methods"],"prefix":"10.3389","volume":"8","author":[{"given":"Sahika Betul","family":"Yayli","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kutay","family":"K\u0131l\u0131\u00e7","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Salih","family":"Beyaz","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2025,2,5]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","first-page":"166","DOI":"10.3109\/14397595.2013.854046","article-title":"Quality of life and self-reported disability in patients with knee osteoarthritis","volume":"24","author":"Alkan","year":"2014","journal-title":"Mod. Rheumatol."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"689","DOI":"10.1186\/s12909-023-04698-z","article-title":"Revolutionizing healthcare: the role of artificial intelligence in clinical practice","volume":"23","author":"Alowais","year":"2023","journal-title":"BMC Med. Educ."},{"key":"ref3","doi-asserted-by":"crossref","first-page":"376","DOI":"10.1007\/978-3-319-62416-7_27","article-title":"Automatic detection of knee joints and quantification of knee osteoarthritis severity using convolutional neural networks","volume-title":"Machine learning and data mining in pattern recognition","author":"Antony","year":"2017"},{"key":"ref4","author":"Brock","year":"2021"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"84","DOI":"10.1016\/j.compmedimag.2019.06.002","article-title":"Fully automatic knee osteoarthritis severity grading using deep neural networks with a novel ordinal loss","volume":"75","author":"Chen","year":"2019","journal-title":"Comput. Med. Imaging Graph."},{"key":"ref6","doi-asserted-by":"publisher","first-page":"299","DOI":"10.52312\/jdrs.2021.64","article-title":"Lack of association between MMP13 (rs3819089), ADAM12 (rs3740199-rs1871054) and ADAMTS14 (rs4747096) genotypes and advanced-stage knee osteoarthritis","volume":"32","author":"Haberal","year":"2021","journal-title":"Jt. Dis. Relat. Surg."},{"key":"ref7","doi-asserted-by":"publisher","first-page":"57","DOI":"10.1016\/j.procs.2022.12.111","article-title":"Impact of CLAHE-based image enhancement for diabetic retinopathy classification through deep learning","volume":"216","author":"Hayati","year":"2023","journal-title":"Procedia Comput. Sci."},{"key":"ref8","volume-title":"ultralytics\/yolov5: v7.0 - YOLOv5 SOTA realtime instance segmentation","author":"Jocher","year":"2022"},{"key":"ref9","doi-asserted-by":"publisher","first-page":"1609","DOI":"10.3390\/diagnostics13091609","article-title":"Hybrid techniques of X-ray analysis to predict knee osteoarthritis grades based on fusion features of CNN and handcrafted","volume":"13","author":"Khalid","year":"2023","journal-title":"Diagnostics (Basel)"},{"key":"ref10","doi-asserted-by":"publisher","first-page":"1886","DOI":"10.1007\/s11999-016-4732-4","article-title":"Classifications in brief: Kellgren-Lawrence classification of osteoarthritis","volume":"474","author":"Kohn","year":"2016","journal-title":"Clin. Orthop. Relat. Res."},{"key":"ref11","doi-asserted-by":"publisher","first-page":"1981","DOI":"10.32604\/cmc.2023.035245","article-title":"Deep learning model ensemble for the accuracy of classification degenerative arthritis","volume":"75","author":"Lee","year":"2023","journal-title":"Comput. Mater. Contin."},{"key":"ref12","doi-asserted-by":"publisher","first-page":"3587","DOI":"10.21037\/qims-22-1250","article-title":"Deep learning-assisted knee osteoarthritis automatic grading on plain radiographs: the value of multiview X-ray images and prior knowledge","volume":"13","author":"Li","year":"2023","journal-title":"Quant. Imaging Med. Surg."},{"key":"ref13","doi-asserted-by":"publisher","first-page":"418","DOI":"10.1007\/s11282-022-00654-7","article-title":"Evaluation of histogram equalization and contrast limited adaptive histogram equalization effect on image quality and fractal dimensions of digital periapical radiographs","volume":"39","author":"Mehdizadeh","year":"2023","journal-title":"Oral Radiol."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"2","DOI":"10.1186\/s40634-021-00442-y","article-title":"Radiographic vs. MRI vs. arthroscopic assessment and grading of knee osteoarthritis - are we using appropriate imaging?","volume":"9","author":"Newman","year":"2022","journal-title":"J. Exp. Orthop."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"844","DOI":"10.1186\/s12891-021-04722-7","article-title":"Automating classification of osteoarthritis according to Kellgren-Lawrence in the knee using deep learning in an unfiltered adult population","volume":"22","author":"Olsson","year":"2021","journal-title":"BMC Musculoskelet. Disord."},{"key":"ref16","year":"2023"},{"key":"ref17","author":"Perez","year":"2017"},{"key":"ref18","doi-asserted-by":"crossref","DOI":"10.1109\/VBC.1990.109340","article-title":"Contrast-limited adaptive histogram equalization: speed and effectiveness","author":"Pizer","year":"1990"},{"key":"ref19","volume-title":"opencv\/cvat: v1.1.0","author":"Sekachev","year":"2020"},{"key":"ref20","doi-asserted-by":"publisher","first-page":"881391","DOI":"10.3389\/fmed.2022.881391","article-title":"The burden of osteoarthritis in the Middle East and North Africa region from 1990 to 2019","volume":"9","author":"Shamekh","year":"2022","journal-title":"Front. Med."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1186\/s40537-019-0197-0","article-title":"A survey on image data augmentation for deep learning","volume":"6","author":"Shorten","year":"2019","journal-title":"J. Big Data"},{"key":"ref22","author":"Simonyan","year":"2014"},{"key":"ref23","article-title":"Inception-v4, inception-ResNet and the impact of residual connections on learning","author":"Szegedy","year":"2016"},{"key":"ref24","author":"Tan","year":"2019"},{"key":"ref25","doi-asserted-by":"publisher","first-page":"1727","DOI":"10.1038\/s41598-018-20132-7","article-title":"Automatic knee osteoarthritis diagnosis from plain radiographs: a deep learning-based approach","volume":"8","author":"Tiulpin","year":"2018","journal-title":"Sci. Rep."},{"key":"ref26","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1155\/2021\/5586529","article-title":"An automatic knee osteoarthritis diagnosis method based on deep learning: data from the osteoarthritis initiative","volume":"2021","author":"Wang","year":"2021","journal-title":"J. Healthc. Eng."},{"key":"ref27","volume-title":"rwightman\/pytorch-image-models: v0.8.10dev0 Release","author":"Wightman","year":"2023"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"869","DOI":"10.1186\/s12891-023-06951-4","article-title":"Assessment of a novel deep learning-based software developed for automatic feature extraction and grading of radiographic knee osteoarthritis","volume":"24","author":"Yoon","year":"2023","journal-title":"BMC Musculoskelet. Disord."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1413820\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,2,5]],"date-time":"2025-02-05T07:32:00Z","timestamp":1738740720000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1413820\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,2,5]]},"references-count":28,"alternative-id":["10.3389\/frai.2025.1413820"],"URL":"https:\/\/doi.org\/10.3389\/frai.2025.1413820","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,2,5]]},"article-number":"1413820"}}