EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Early diagnosis and grading of knee osteoarthritis from radiographs using KOA-DenseNet++

Michael Joseph Shundi, Chiranjeevi Karri, Rajesh Kumar P.

Imaging · 17.04.2026

DOI: 10.1556/1647.2026.00372 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Imaging, erschienen am 17.04.2026. Verfasst von Michael Joseph Shundi und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 56 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background and Aim Early diagnosis of knee osteoarthritis (KOA) is crucial for slowing disease progression and improving patient outcomes. However, radiographic assessment using the Kellgren–Lawrence (KL) grading system can be subjective and inconsistent. Although recent deep learning techniques have improved diagnostic accuracy, many models still face challenges in identifying subtle features in the early and intermediate stages of KOA. This study aims to develop and evaluate KOA-DenseNet++ , an enhanced DenseNet201-based framework, for accurate and reliable KOA diagnosis and severity grading from radiographic images. Patients and Methods A total of 9,786 knee X-rays from the Osteoarthritis Initiative (OAI) dataset were used to train and evaluate the proposed model. KOA-DenseNet++ was assessed on two tasks: (i) binary classification of healthy (KL0–1) versus osteoarthritic (KL2–4) cases, and (ii) multi-class severity grading for KL2, KL3, and KL4. The model integrates dense connectivity with refined multi-scale feature extraction and advanced data preprocessing to improve feature sensitivity and robustness. Comparative analyses were performed against baseline architectures, including VGG16, ResNet50, EfficientNetB0, and MobileNetV2.

Results

The proposed KOA-DenseNet++ achieved a classification accuracy of 96.1% for binary diagnosis, demonstrating strong performance in early-stage detection, and 90.2% accuracy for severity grading. Both results exceeded the performance of the benchmark deep learning models.

Conclusions

These results demonstrate the potential of KOA-DenseNet++ as a clinically viable tool for early diagnosis and consistent staging of KOA.

Zitieren

APA

Shundi, M. J., Karri, C., & P., R. K. (2026). Early diagnosis and grading of knee osteoarthritis from radiographs using KOA-DenseNet++. Imaging. https://doi.org/10.1556/1647.2026.00372

Vancouver

Shundi MJ, Karri C, P. RK. Early diagnosis and grading of knee osteoarthritis from radiographs using KOA-DenseNet++. Imaging. 2026. doi:10.1556/1647.2026.00372

BibTeX
@article{Shundi2026,
  author = {Shundi, Michael Joseph and Karri, Chiranjeevi and P., Rajesh Kumar},
  title = {Early diagnosis and grading of knee osteoarthritis from radiographs using KOA-DenseNet++},
  journal = {Imaging},
  year = {2026},
  doi = {10.1556/1647.2026.00372}
}
Kostenlos starten →