Whole-Kidney Radiomics Performs Robustly in Assessing Kidney Stones
Key findings
- Radiomics, a new approach to medical imaging, uses artificial intelligence to extract a large number of quantitative features from medical images with the aim of improving clinical decision-making
- In this study, abdomen–pelvis computed tomography (CT) images of 202 patients were processed using a radiomics prototype that automatically identifies and segments the entire kidney volume
- Radiomics was able to detect and quantify kidney stone burden, assess the presence of hydronephrosis and identify patients who received invasive treatment
- In the future, radiomics may allow assessment of stone burden in a more quantitative and reproducible manner, allowing clinicians to better determine the suitable treatment pathway for patients and monitor for changes over time
Subscribe to the latest updates from Radiology Advances in Motion
Recent studies have reported on the role of stone-specific radiomics to detect and evaluate kidney stones. Radiomics, a new approach to medical imaging, uses artificial intelligence to extract a large number of quantitative features from medical images with the aim of improving clinical decision-making.
Extending that research, Fatemeh Homayounieh, MD, research fellow in the Webster Center for Quality and Safety in the Department of Radiology at Massachusetts General Hospital, Mannudeep K. Kalra, MD, staff radiologist in the Division of Thoracic Imaging and director of the Webster Center for Quality and Safety, and colleagues have used radiomics of the whole kidney to quantify kidney stone burden, assess the presence of hydronephrosis and predict the need for invasive treatment. Their report appears in Abdominal Radiology.
Study Methods
This retrospective study was conducted at three hospitals that used multidetector-row CT scanners from three different vendors. The researchers identified 202 adults who underwent unenhanced abdomen–pelvis CT examinations for renal colic or kidney stones between May and June 2019.
A senior radiologist used a picture archiving and communication system workstation to visually detect kidney stones, review each segmented contour, edit the segmentation margins and visually detect hydronephrosis. The workstation calculated the overall stone burden.
Another physician processed the CT images with a standalone radiomics prototype that automatically identifies and segments the entire kidney volume with a single click of an icon.
Results
Radiomics was able to:
- Differentiate patients with and without kidney stones (receiver operating characteristics area under the curve [AUC], 0.84; P<0.003)
- Differentiate between adjacent quartiles of kidney stone volume (first vs. second quartiles: AUC, 0.84; second vs. third: AUC, 0.88; third vs. fourth: AUC, 0.89)
- Differentiate kidneys with and without hydronephrosis among all 404 kidneys (AUC, 0.85; P<0.006), although not among the subset of 184 kidneys with stones
- Identify which patients received lithotripsy or flexible ureteroscopy rather than conservative treatment (AUC, 0.91; P<0.02)
The performance of radiomics did not vary based on the CT scanner.
A New Direction for Kidney Stone Assessment
Manual assessment of the size or dimension of kidney stones can be time-consuming and prone to intra- and inter-observer inconsistencies. The radiomics prototype used in this study addresses this inefficiency. For example, there was no need to correct any segmented renal contour, although this capability exists.
The high AUC (>0.8) in patients from three sites with different scanner types supports the generalizability of radiomics in assessing kidney stones. In the future, radiomics may allow assessment of stone burden in a more quantitative and reproducible manner, allowing clinicians to better determine the suitable treatment pathway for patients and monitor for changes over time.
view original journal article Subscription may be required
Visit the Department of Urology
Refer a patient to the Department of Urology