This website requires JavaScript.

Medical Diffusion -- Denoising Diffusion Probabilistic Models for 3D Medical Image Generation

Firas KhaderGustav Mueller-FranzesSoroosh Tayebi Arasteh ...+11 Daniel Truhn
Nov 2022
摘要
Recent advances in computer vision have shown promising results in imagegeneration. Diffusion probabilistic models in particular have generatedrealistic images from textual input, as demonstrated by DALL-E 2, Imagen andStable Diffusion. However, their use in medicine, where image data typicallycomprises three-dimensional volumes, has not been systematically evaluated.Synthetic images may play a crucial role in privacy preserving artificialintelligence and can also be used to augment small datasets. Here we show thatdiffusion probabilistic models can synthesize high quality medical imagingdata, which we show for Magnetic Resonance Images (MRI) and Computed Tomography(CT) images. We provide quantitative measurements of their performance througha reader study with two medical experts who rated the quality of thesynthesized images in three categories: Realistic image appearance, anatomicalcorrectness and consistency between slices. Furthermore, we demonstrate thatsynthetic images can be used in a self-supervised pre-training and improve theperformance of breast segmentation models when data is scarce (dice score 0.91vs. 0.95 without vs. with synthetic data).
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答