WPDM: enhancing subarachnoid hemorrhage CT detection with Worley-Perlin noise diffusion on imbalanced data

We propose a diffusion model based on Worley-Perlin noise, which provides higher robustness for image synthesis problems with striated textures, e.g. subarachnoid hemorrhage (SAH) CT slices. Unlike Gaussian noise, which uniformly perturbs images and causes the generated samples to be overly homogeneous, or Simplex noise, which excessively disrupts images and leads to distortion, our method accounts for the morphological structure of textured images. It ensures effective perturbation while avoiding image distortion and generates high-quality samples for downstream tasks. To validate the effectiveness of our proposed generative model in improving SAH CT detection performance on imbalanced datasets, we incorporated the generated SAH samples into the training process and compared the results with those of models trained on the original dataset. The experimental results demonstrated that our synthesized data can help improve the classification accuracy in an imbalanced data situation, with the F1-score improving by nearly 6.4%.