CLOG-CD: Curriculum Learning based on Oscillating Granularity of Class Decomposed Medical Image Classification

TL;DR

This work tackles imbalanced and overlapping classes in medical image classification by introducing CLOG-CD, a CNN training framework that fuses class decomposition with oscillating curriculum learning. It uses a convolutional autoencoder to extract latent features and $k$-means clustering with $k=5$ to create granularity levels $G=igl\{g_k,g_{k-1},...,g_1\bigr\}$, then trains in anti-curriculum fashion across these levels to enhance feature transferability. Across four diverse medical-imaging datasets, including CXR, brain tumour, digital knee X-ray, and CRC, CLOG-CD achieves high accuracies, e.g., ResNet-50: 96.08% (CXR), 96.91% (brain), 79.76% (knee), 99.17% (CRC); DenseNet-121: 94.86%, 94.63%, 76.19%, 99.45% respectively, outperforming baseline and other curriculum-based methods. The key contribution is the first integration of class decomposition within curriculum learning for medical image classification, yielding improved robustness and generalization in irregular data distributions and demonstrating practical impact for clinical imaging tasks.

Abstract

Curriculum learning strategies have been proven to be effective in various applications and have gained significant interest in the field of machine learning. It has the ability to improve the final model's performance and accelerate the training process. However, in the medical imaging domain, data irregularities can make the recognition task more challenging and usually result in misclassification between the different classes in the dataset. Class-decomposition approaches have shown promising results in solving such a problem by learning the boundaries within the classes of the data set. In this paper, we present a novel convolutional neural network (CNN) training method based on the curriculum learning strategy and the class decomposition approach, which we call CLOG-CD, to improve the performance of medical image classification. We evaluated our method on four different imbalanced medical image datasets, such as Chest X-ray (CXR), brain tumour, digital knee X-ray, and histopathology colorectal cancer (CRC). CLOG-CD utilises the learnt weights from the decomposition granularity of the classes, and the training is accomplished from descending to ascending order (i.e., anti-curriculum technique). We also investigated the classification performance of our proposed method based on different acceleration factors and pace function curricula. We used two pre-trained networks, ResNet-50 and DenseNet-121, as the backbone for CLOG-CD. The results with ResNet-50 show that CLOG-CD has the ability to improve classification performance with an accuracy of 96.08% for the CXR dataset, 96.91% for the brain tumour dataset, 79.76% for the digital knee X-ray, and 99.17% for the CRC dataset, compared to other training strategies. In addition, with DenseNet-121, CLOG-CD has achieved 94.86%, 94.63%, 76.19%, and 99.45% for CXR, brain tumour, digital knee X-ray, and CRC datasets, respectively

Figures (4)

• Figure 1: An example of the granularity of the class decomposition method: a) the original class, b) the newly generated datasets after applying the granularity of class decomposition (e.g. $k$=4).
• Figure 2: The framework of the CLOG-CD model. The dataset is decomposed into sequential granularity levels with $k$ components based on different speeds. The model fine-tunes initial weights $(w^{0})$, starting with the original classes of the dataset $(g_{1})$ in the (ASG) model. While, in (DEG) and CLOG-CD with different speeds, training starts at the maximum granularity level $(g_{k})$, where the dataset was decomposed into the maximum number of sub-classes.
• Figure 9: The fitted curve at degree 3 of CRC dataset over 20 iterations obtained by: a) ACC, b) PR, c) RE, and d) F1, with ResNet-50.
• Figure 10: The confusion matrix results of CRC dataset obtained by: a) ResNet-50 baseline, b) CLOG-CD (ASG), C) CLOG-CD (DEG), d) CLOG-CD ($\bigtriangleup=1$), e) CLOG-CD ($\bigtriangleup=2$), and f) CLOG-CD ($\bigtriangleup=4$).