| Summary: | For many years, convolutional neural networks dominated the field of computer vision, not least in the medical field, where problems such as image segmentation were addressed by such networks as the U-Net. The arrival of self-attention-based networks to the field of computer vision through ViTs seems to have changed the trend of using standard convolutions. Throughout this work, we apply different architectures such as U-Net, ViTs and ConvMixer, to compare their performance on a medical semantic segmentation problem. All the models have been trained from scratch on the DRIVE dataset and evaluated on their private counterparts to assess which of the models performed better in the segmentation problem. Our major contribution is showing that the best-performing model (ConvMixer) is the one that shares the approach from the ViT (processing images as patches) while maintaining the foundational blocks (convolutions) from the U-Net. This mixture does not only produce better results (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mi>D</mi><mi>I</mi><mi>C</mi><mi>E</mi><mo>=</mo><mn>0.83</mn></mrow></semantics></math></inline-formula>) than both ViTs (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.80</mn></mrow></semantics></math></inline-formula>/0.077 for UNETR/SWIN-Unet) and the U-Net (<inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>0.82</mn></mrow></semantics></math></inline-formula>) on their own but reduces considerably the number of parameters (2.97M against 104M/27M and 31M, respectively), showing that there is no need to systematically use large models for solving image problems where smaller architectures with the optimal pieces can get better results.
|
|---|