Radiomics-based nomogram using digital radiography for early diagnosis of knee osteoarthritis

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Early diagnosis of knee osteoarthritis (KOA) is crucial but challenging, especially differentiating between Kellgren-Lawrence (KL) grades 1 and 2 on standard radiographs.
• Radiomics offers a potential solution for extracting quantitative imaging features to aid in early radiographic KOA (RKOA) detection.

Purpose of the Study:

• To develop and validate a radiomics-based model using digital radiography (DR) for the early identification of RKOA.
• To assess the performance of machine learning models and a combined nomogram incorporating radiomics and age for KOA diagnosis.

Main Methods:

• Retrospective analysis of 859 patients with KL grade 1 or 2 KOA, divided into training (n=601) and validation (n=258) sets.
• Extraction of 2,632 radiomics features from DR images, followed by feature selection using ICC, correlation analysis, and LASSO regression, resulting in 38 features.
• Development and comparison of five machine learning models, with logistic regression (LR) selected for its generalizability to compute a radiomics score (Radscore) and a combined nomogram.

Main Results:

• The LR model achieved an AUC of 0.821 in the validation set.
• The combined nomogram model significantly outperformed the Radscore model alone, showing AUCs of 0.914 (training) and 0.833 (validation) compared to 0.908 and 0.823, respectively.
• Calibration curves and decision curve analysis confirmed the nomogram's good fit and clinical utility.

Conclusions:

• A DR-based radiomics model, when combined with patient age, provides accurate early diagnosis of KOA.
• This approach demonstrates significant potential for clinical application in the early detection of RKOA.