Related Experiment Video
Updated: Apr 15, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.6K
Automated Analysis of Split Kidney Function from CT Scans Using Deep Learning and Delta Radiomics
Ramon Luis Correa-Medero1, Jiwoong Jeong1, Bhavik Patel1,2
1School of Computing and Augmented Intelligence, Arizona State University, Arizona, USA.
Journal of Endourology
|May 2, 2024
Summary
This study shows that deep learning and radiomic features from computed tomography (CT) scans can accurately assess differential kidney function. This method may reduce the need for nuclear medicine scans in preoperative evaluations.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Differential kidney function (DKF) assessment is crucial for preoperative planning of urological interventions.
- Current DKF assessment relies on nuclear medicine imaging, which involves radiation and is not integrated with conventional imaging.
- There is a need for non-invasive, readily available methods for DKF assessment.
Purpose of the Study:
- To evaluate the feasibility of assessing DKF using contrast-enhanced computed tomography (CT) scans.
- To develop and validate a deep learning and radiomic feature-based pipeline for DKF estimation.
- To determine if this novel approach can reduce reliance on nuclear medicine scans.
Main Methods:
- A retrospective analysis of patients who underwent kidney nuclear scanning and contrast-enhanced CT scans was performed.
- A segmentation model was used to isolate kidneys, followed by extraction of 2D and 3D radiomic features.
- Delta radiomics features were computed and used to train a random forest model to predict DKF.
- Model performance was validated on internal and external datasets using receiver operating characteristic (ROC) curves, sensitivity, and specificity.
Main Results:
- The random forest model utilizing 3D delta radiomic features achieved an area under the curve (AUC) of 0.85 (internal) and 0.81 (external).
- The model demonstrated good predictive performance with internal set sensitivity/specificity of 0.84/0.68 and external set sensitivity/specificity of 0.70/0.65.
- The developed automated pipeline successfully derived differential kidney function information from routine CT scans.
Conclusions:
- An automated pipeline using deep learning and radiomics from contrast-enhanced CT can effectively assess differential kidney function.
- This CT-based method offers a promising alternative to nuclear medicine scans for early-stage DKF assessment.
- The findings establish a machine learning methodology for DKF evaluation from routine CT, avoiding radioactive tracers and associated costs.
