# Image Imputation with conditional generative adversarial networks captures clinically relevant imaging features on computed tomography

## TL;DR

AI can generate missing kidney cancer CT scans that closely match real ones, preserving important diagnostic features for better clinical decisions.

## Contribution

Demonstrates that cGANs can impute missing CT images while retaining clinically relevant features for kidney cancer diagnosis.

## Key findings

- Imputed images showed over 85% agreement with true images for 13 categorical clinical features.
- Imputed images preserved key radiomic features like mean intensity and tumor characteristics.
- AI-generated scans maintained diagnostic relevance for benign or malignant kidney tumors.

## Abstract

Kidney cancer is among the top 10 most common malignancies in adults, and is commonly evaluated with four-phase computed tomography (CT) imaging. However, the presence of missing or corrupted images remains a significant problem in medical imaging that impairs the detection, diagnosis, and treatment planning of kidney cancer. Deep learning approaches through conditional generative adversarial networks (cGANs) have recently shown technical promise in the task of imputing missing imaging data from these four-phase studies. In this study, we explored the clinical utility of these imputed images. We utilized a cGAN trained on 333 patients, with the task of the cGAN being to impute the image of any phase given the other three phases. We tested the clinical utility on the imputed images of the 37 patients in the test set by manually extracting 21 clinically relevant imaging features and comparing them to their ground truth counterpart. All 13 categorical clinical features had greater than 85% agreement rate between true images and their imputed counterparts. This high accuracy is maintained when stratifying across imaging phases. Imputed images also show good agreement with true images in select radiomic features including mean intensity and enhancement. Imputed images possess the features characteristic of benign or malignant diagnosis at an equivalent rate to true images. In conclusion, imputed images from cGANs have large potential for clinical use due to their ability to retain clinically relevant qualitative and quantitative features.

Kidney cancer is one of the most common cancers in adults, and doctors often use a special type of CT scan that captures four different phases to detect and understand the disease. However, some of these scans can be missing or unusable due to technical problems, which can affect diagnosis and treatment planning. To address this, we used a type of artificial intelligence called a conditional generative adversarial network (cGAN) to “fill in” missing scans by learning from complete ones. After training our model on data from 333 patients, we tested it on 37 new patients to see how well the AI-generated scans matched the real ones. We compared 21 important imaging features and found that the AI-generated scans closely matched the originals, with over 85% agreement on all 13 key categorical features. These generated images also preserved important tumor characteristics and image details. Our findings suggest that AI can help recover missing medical images in a way that still supports accurate clinical decision-making. This technology could improve care for kidney cancer patients and reduce the impact of incomplete imaging.