Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GenMix: Combining Generative and Mixture Data Augmentation for Medical Image Classification

Hansang Lee, Haeil Lee, Helen Hong

TL;DR

GenMix tackles data scarcity in medical image classification by marrying generative augmentation with MixUp to bolster both sample diversity and decision boundaries. The two-stage approach first generates synthetic images via multiple generative models, then applies MixUp between synthetic and real data to form enhanced training samples, addressing mode collapse and class imbalance. Across focal liver lesion CT data, GenMix consistently improves performance for various generators, with Textual Inversion yielding the strongest gains without task-specific finetuning. The findings highlight GenMix’s potential to generalize to other medical-imaging tasks and datasets, offering a practical path to more robust diagnostic models.

Abstract

In this paper, we propose a novel data augmentation technique called GenMix, which combines generative and mixture approaches to leverage the strengths of both methods. While generative models excel at creating new data patterns, they face challenges such as mode collapse in GANs and difficulties in training diffusion models, especially with limited medical imaging data. On the other hand, mixture models enhance class boundary regions but tend to favor the major class in scenarios with class imbalance. To address these limitations, GenMix integrates both approaches to complement each other. GenMix operates in two stages: (1) training a generative model to produce synthetic images, and (2) performing mixup between synthetic and real data. This process improves the quality and diversity of synthetic data while simultaneously benefiting from the new pattern learning of generative models and the boundary enhancement of mixture models. We validate the effectiveness of our method on the task of classifying focal liver lesions (FLLs) in CT images. Our results demonstrate that GenMix enhances the performance of various generative models, including DCGAN, StyleGAN, Textual Inversion, and Diffusion Models. Notably, the proposed method with Textual Inversion outperforms other methods without fine-tuning diffusion model on the FLL dataset.

GenMix: Combining Generative and Mixture Data Augmentation for Medical Image Classification

TL;DR

GenMix tackles data scarcity in medical image classification by marrying generative augmentation with MixUp to bolster both sample diversity and decision boundaries. The two-stage approach first generates synthetic images via multiple generative models, then applies MixUp between synthetic and real data to form enhanced training samples, addressing mode collapse and class imbalance. Across focal liver lesion CT data, GenMix consistently improves performance for various generators, with Textual Inversion yielding the strongest gains without task-specific finetuning. The findings highlight GenMix’s potential to generalize to other medical-imaging tasks and datasets, offering a practical path to more robust diagnostic models.

Abstract

In this paper, we propose a novel data augmentation technique called GenMix, which combines generative and mixture approaches to leverage the strengths of both methods. While generative models excel at creating new data patterns, they face challenges such as mode collapse in GANs and difficulties in training diffusion models, especially with limited medical imaging data. On the other hand, mixture models enhance class boundary regions but tend to favor the major class in scenarios with class imbalance. To address these limitations, GenMix integrates both approaches to complement each other. GenMix operates in two stages: (1) training a generative model to produce synthetic images, and (2) performing mixup between synthetic and real data. This process improves the quality and diversity of synthetic data while simultaneously benefiting from the new pattern learning of generative models and the boundary enhancement of mixture models. We validate the effectiveness of our method on the task of classifying focal liver lesions (FLLs) in CT images. Our results demonstrate that GenMix enhances the performance of various generative models, including DCGAN, StyleGAN, Textual Inversion, and Diffusion Models. Notably, the proposed method with Textual Inversion outperforms other methods without fine-tuning diffusion model on the FLL dataset.
Paper Structure (8 sections, 1 equation, 4 figures, 1 table)

This paper contains 8 sections, 1 equation, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Synthetic Image Generation with GANs and Diffusion Models
  4. MixUp between Differently Sampled Data
  5. Experiments
  6. Experimental Settings
  7. Results
  8. Conclusion

Figures (4)

  • Figure 1: An overview of the proposed data augmentation method.
  • Figure 2: The evolution of synthetic FLL images during the training process of the diffusion model. Initially, starting from an ImageNet-pretrained model, natural images are generated, which gradually converge to FLL patch images cropped from CT scans.
  • Figure 3: Examples of FLL images for cysts (left), hemangiomas (middle), and metastases (right) in real, synthetic, and GenMix training data.
  • Figure 4: t-SNE visualization of the distribution of training data features extracted from VGG-16 trained on various data augmentation methods. (Blue: Cyst, Green: Hemangioma, Red: Metastasis, Dots: Real data, Triangles: Synthetic data, Circles: GenMix data)