Radiomic Synthesis Using Deep Convolutional Neural Networks

TL;DR

RadSynth, a deep CNN model, efficiently generates radiomic images from MRI data, significantly reducing computation time while maintaining high accuracy, thus aiding clinical decision support.

Contribution

The paper introduces RadSynth, a novel deep learning approach that accelerates radiomic image synthesis with high fidelity, outperforming traditional methods in speed and accuracy.

Findings

RadSynth produces synthetic GLCM entropy images with a 0.07% average difference.

The correlation between RadSynth and traditional images is 0.97.

RadSynth significantly reduces computation time for radiomic features.

Abstract

Radiomics is a rapidly growing field that deals with modeling the textural information present in the different tissues of interest for clinical decision support. However, the process of generating radiomic images is computationally very expensive and could take substantial time per radiological image for certain higher order features, such as, gray-level co-occurrence matrix(GLCM), even with high-end GPUs. To that end, we developed RadSynth, a deep convolutional neural network(CNN) model, to efficiently generate radiomic images. RadSynth was tested on a breast cancer patient cohort of twenty-four patients(ten benign, ten malignant and four normal) for computation of GLCM entropy images from post-contrast DCE-MRI. RadSynth produced excellent synthetic entropy images compared to traditional GLCM entropy images. The average percentage difference and correlation between the two techniques were 0.07 $\pm$ 0.06 and 0.97, respectively. In conclusion, RadSynth presents a new powerful tool for fast computation and visualization of the textural information present in the radiological images.