Knee osteoarthritis network: A hybrid transformer-based approach for enhanced detection and grading of knee osteoarthritis

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27360%2F25%3A10258089" target="_blank" >RIV/61989100:27360/25:10258089 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.sciencedirect.com/science/article/pii/S0952197625017531" target="_blank" >https://www.sciencedirect.com/science/article/pii/S0952197625017531</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.engappai.2025.111751" target="_blank" >10.1016/j.engappai.2025.111751</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Knee osteoarthritis network: A hybrid transformer-based approach for enhanced detection and grading of knee osteoarthritis

Popis výsledku v původním jazyce

Background: Knee osteoarthritis is a prevalent and debilitating condition that significantly affects the elderly population worldwide, leading to pain, joint deformity, and reduced mobility. Early diagnosis and severity grading of knee osteoarthritis are critical for timely intervention but remain challenging due to subjective clinical assessments and the complex progression of the disease. This study aims to develop an automated and reliable artificial intelligence-based deep learning framework for detecting and grading knee osteoarthritis severity using radiographic images.Methods: We propose the Knee Osteoarthritis Network, an innovative hybrid deep learning-based system designed for the automatic detection and grading of knee osteoarthritis. The framework integrates advanced transformer-based models, including Vision Transformer, Swin Transformer, and Faster Vision Transformer, leveraging transfer learning to enhance feature extraction from knee X-ray images. Features extracted from each model are combined using a serial fusion strategy, followed by a hybrid feature selection method employingmutual information and genetic algorithms. The system is evaluated on the publicly available Osteoarthritis Initiative dataset, consisting of 8,260 radiographic images, using stratified five-fold cross-validation.Results: The proposed framework outperforms existing state-of-the-art methods, achieving a classification accuracy of 97.03%, a Cohen’s kappa score of 0.98, a mean absolute error of 0.0312, and a mean squared error of 0.0374. To improve interpretability, we apply explainable artificial intelligence techniques, including Layer Class Activation Mapping and SHapley Additive exPlanations, to visualize class-discriminative regions and identify important features.Conclusions: The experimental results demonstrate that the proposed Knee Osteoarthritis Network offers a reliable and interpretable solution for the automated diagnosis and grading of knee osteoarthritis from radiographic images. By combining transformer-based feature extraction with robust fusion and selection techniques, the framework improves diagnostic accuracy and reduces reliance on manual clinical grading.

Název v anglickém jazyce

Knee osteoarthritis network: A hybrid transformer-based approach for enhanced detection and grading of knee osteoarthritis

Popis výsledku anglicky

Background: Knee osteoarthritis is a prevalent and debilitating condition that significantly affects the elderly population worldwide, leading to pain, joint deformity, and reduced mobility. Early diagnosis and severity grading of knee osteoarthritis are critical for timely intervention but remain challenging due to subjective clinical assessments and the complex progression of the disease. This study aims to develop an automated and reliable artificial intelligence-based deep learning framework for detecting and grading knee osteoarthritis severity using radiographic images.Methods: We propose the Knee Osteoarthritis Network, an innovative hybrid deep learning-based system designed for the automatic detection and grading of knee osteoarthritis. The framework integrates advanced transformer-based models, including Vision Transformer, Swin Transformer, and Faster Vision Transformer, leveraging transfer learning to enhance feature extraction from knee X-ray images. Features extracted from each model are combined using a serial fusion strategy, followed by a hybrid feature selection method employingmutual information and genetic algorithms. The system is evaluated on the publicly available Osteoarthritis Initiative dataset, consisting of 8,260 radiographic images, using stratified five-fold cross-validation.Results: The proposed framework outperforms existing state-of-the-art methods, achieving a classification accuracy of 97.03%, a Cohen’s kappa score of 0.98, a mean absolute error of 0.0312, and a mean squared error of 0.0374. To improve interpretability, we apply explainable artificial intelligence techniques, including Layer Class Activation Mapping and SHapley Additive exPlanations, to visualize class-discriminative regions and identify important features.Conclusions: The experimental results demonstrate that the proposed Knee Osteoarthritis Network offers a reliable and interpretable solution for the automated diagnosis and grading of knee osteoarthritis from radiographic images. By combining transformer-based feature extraction with robust fusion and selection techniques, the framework improves diagnostic accuracy and reduces reliance on manual clinical grading.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10200 - Computer and information sciences

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Engineering Applications of Artificial Intelligence

ISSN

0952-1976

e-ISSN

1873-6769

Svazek periodika

159

Číslo periodika v rámci svazku

111751

Stát vydavatele periodika

GB - Spojené království Velké Británie a Severního Irska

Počet stran výsledku

16

Strana od-do

nestránkováno

Kód UT WoS článku

001540179700005

EID výsledku v databázi Scopus

—