Exploring the Efficacy of Base Data Augmentation Methods in Deep Learning-Based Radiograph Classification of Knee Joint Osteoarthritis

TL;DR

This study explored data augmentation methods, including adversarial techniques, and found potential confounding regions within radiographic images of knee osteoarthritis using adversarial augmentation, shown in models’ accurate classification of extreme KOA grades, even without the knee joint.

Abstract

Diagnosing knee joint osteoarthritis (KOA), a major cause of disability worldwide, is challenging due to subtle radiographic indicators and the varied progression of the disease. Using deep learning for KOA diagnosis requires broad, comprehensive datasets. However, obtaining these datasets poses significant challenges due to patient privacy concerns and data collection restrictions. Additive data augmentation, which enhances data variability, emerges as a promising solution. Yet, it's unclear which augmentation techniques are most effective for KOA. This study explored various data augmentation methods, including adversarial augmentations, and their impact on KOA classification model performance. While some techniques improved performance, others commonly used underperformed. We identified potential confounding regions within the images using adversarial augmentation. This was evidenced by our models' ability to classify KL0 and KL4 grades accurately, with the knee joint omitted. This observation suggested a model bias, which might leverage unrelated features for classification currently present in radiographs. Interestingly, removing the knee joint also led to an unexpected improvement in KL1 classification accuracy. To better visualize these paradoxical effects, we employed Grad-CAM, highlighting the associated regions. Our study underscores the need for careful technique selection for improved model performance and identifying and managing potential confounding regions in radiographic KOA deep learning.

Figures (10)

• Figure 1: The methodological pipeline for the study. Green represents positive/supportive augmentation, red signifies adversarial augmentations. Blue indicates data processing and purple signifies CNN training. Numeric markers indicate the order of operations.
• Figure 2: Sample images showing various KL grades, ranging from 0 (no OA signs) to 4 (severe OA). From left to right, OA severity increases. Joint space narrowing, denoted as JSN.
• Figure 3: Visualization of the study's base data augmentations: red indicates negative/adversarial augmentations and green shows positive/supportive augmentations. Each transformation is demonstrated on provided baseline image.
• Figure 4: EfficientNetV2 base architecture with our post-pooling modifications.
• Figure 5: Confusion matrices for the test set using positive/supportive base data augmentations.