Can similar OA grades be distinguished more precisely?

Overall KL classification accuracy was 75.7%. This was five-category classification of grades 0–4.

Eigen-CAM visualizations on radiographs spanning KL grades 0–4.
Do Weon Lee, Dae Seok Song, Hyuk-Soo Han, Du Hyun Ro. Knee Surgery & Related Research (2024). DOI: 10.1186/s43019-024-00228-3
Evidence & Original Source Image use terms

The research question

Participants and Methods

A plug-in module (PIM) for subtle image differences was attached to CNN and Transformer models. Four models were ensembled and evaluated on MOST independently of OAI. Models were developed on OAI and tested on 17,040 MOST images. Prospectively collected cohorts were analysed retrospectively.

Main Findings

Overall KL classification accuracy was 75.7%. This was five-category classification of grades 0–4. Borderline KL 1 was difficult to classify. Eigen-CAM examples show image regions attended to by the model, including osteophytes and joint space.

Implications for clinical and surgical practice

This foundational grading study leads to CONNEVO KOA. Alongside technology for subtle differences, clinicians need to assess uncertainty in borderline grades.

Limitations

Some grade-specific values differ between the abstract and results/tables. The overall accuracy consistently reported across them is presented here. Five-category accuracy cannot be compared directly with binary OA (≥2) accuracy in the 2026 multicentre study.

Let's put it briefly.

Which clinical question does this address?

This foundational grading study leads to CONNEVO KOA. Alongside technology for subtle differences, clinicians need to assess uncertainty in borderline grades.

Keywords

Classification
Deep learning
Knee osteoarthritis
AI
KOA
KL
external validation

Abstract · Original text

Do Weon Lee, Dae Seok Song, Hyuk-Soo Han, Du Hyun Ro. Accurate, automated classification of radiographic knee osteoarthritis severity using a novel method of deep learning: Plug-in modules. Knee surgery & related research. 2024;36(1):24. doi:10.1186/s43019-024-00228-3.

BACKGROUND

Fine-grained classification deals with data with a large degree of similarity, such as cat or bird species, and similarly, knee osteoarthritis severity classification [Kellgren-Lawrence (KL) grading] is one such fine-grained classification task. Recently, a plug-in module (PIM) that can be integrated into convolutional neural-network-based or transformer-based networks has been shown to provide strong discriminative regions for fine-grained classification, with results that outperformed the previous deep learning models. PIM utilizes each pixel of an image as an independent feature and can subsequently better classify images with minor differences. It was hypothesized that, as a fine-grained classification task, knee osteoarthritis severity may be classified well using PIMs. The aim of the study was to develop this automated knee osteoarthritis classification model.

METHODS

A deep learning model that classifies knee osteoarthritis severity of a radiograph was developed utilizing PIMs. A retrospective analysis on prospectively collected data was performed. The model was trained and developed using the Osteoarthritis Initiative dataset and was subsequently tested on an independent dataset, the Multicenter Osteoarthritis Study (test set size: 17,040). The final deep learning model was designed through an ensemble of four different PIMs.

RESULTS

The accuracy of the model was 84%, 43%, 70%, 81%, and 96% for KL grade 0, 1, 2, 3, and 4, respectively, with an overall accuracy of 75.7%.

CONCLUSIONS

The ensemble of PIMs could classify knee osteoarthritis severity using simple radiographs with a fine accuracy. Although improvements will be needed in the future, the model has been proven to have the potential to be clinically useful.

Read abstract on PubMed Terms of use

The original paper.

Accurate, automated classification of radiographic knee osteoarthritis severity using a novel method of deep learning: Plug-in modules

Do Weon Lee, Dae Seok Song, Hyuk-Soo Han, Du Hyun Ro
Knee Surgery & Related Research · 2024

DOI · 10.1186/s43019-024-00228-3 PubMed Based data

Academic Discussion

Original Figure 6 · Image regions highlighted by OA grade