Advances in Noninvasive Imaging Techniques to Improve Stone Outcomes

By: Alec Mitchell, MD, MBT; Abdulghafour Halawani, MD, FEBU; Connor Forbes, MD, FRCSC | Posted on: 02 Feb 2023

Figure. Volume of 3 shapes, all of which have a 10 mm diameter. If these were stones, each could be reported as “10 mm” in size, despite having quite different stone volume which could affect operative approach and planning.


Endourologic headlines often feature technical advances in intraoperative care for kidney stones, including miniaturization and improvements in scope and laser technology, which quite rightly have advanced care for our patients. Likewise, preoperative optimization remains an important part of the patient care pathway. In this article, we review emerging techniques in non-invasive preoperative assessment and image processing for kidney stone management.

Advances in CT Acquisition and Analysis

Non-contrast computed tomography (NCCT) provides the most information for preoperative planning for kidney stone surgery, but can be further optimized beyond conventional approaches. Dual energy CT (DECT), for example, has been classically used to distinguish between uric acid and non-uric acid stones. This technology uses x-rays with 2 different energy levels to characterize substances based on their attenuation at each level. Recently, this has been improved upon further to accurately predict the presence or absence of infectious stones via a machine learning model, with an AUC in an external validation set of 0.803.1 Similarly, DECT has been optimized by calibration based on known chemical compounds to develop an algorithm which can be applied to clinical images. Using this technique, researchers were able to differentiate between stone compositions with high accuracy, including 86% accuracy in differentiation between calcium oxalate monohydrate and calcium oxalate dihydrate.2 This type of preoperative differentiation has many potential applications, including in predicting shock wave lithotripsy outcomes. Furthermore, identification of infectious stones may be helpful for counseling between retrograde ureteroscopy (URS) or percutaneous approaches for larger stones.

Ureteral attenuation has also been identified as a technique to improve extracorporeal shock wave lithotripsy (ESWL) outcomes and minimize unnecessary ESWL prior to URS. Measuring the value of CT attenuation of the ureter above and below ureteral calculi can give key insights into ureteral stone impaction.3 In a recent retrospective study, 171 patients with ureteral stones who underwent URS with preoperative NCCT were included. The result showed that thicker ureteral wall, and lower Hounsfield unit ratio of the ureter above compared to below the stone were independent predictors of impacted ureteral calculi (P < .01). Ureteral wall thickness ≥4 mm in combination with Hounsfield unit ratio ≤0.3 corresponded to a 40.5% rate of impaction.3

Volumetric Assessments of Stone Burden

Stone diameter, while convenient to record, provides an incomplete picture of true stone burden: a “10 mm” diameter stone can have volumes anywhere between 10 mm3 and 523 mm3, depending on if it is linear or spherical (see Figure). The intuitive importance of this was underscored in a recent study which confirmed that 3-dimensional (3D) stone volume outperformed linear stone measurement for predicting shock wave lithotripsy success (AUC 0.729 vs 0.672).4 Although not yet widely integrated into clinical workflow, modern 3D volume analyzers can assess stone volume without the need for manual annotation of areas of interest. This automation improves preoperative predictions and is also more reproducible for research purposes, with potential future clinical applications.

Anatomic Reconstruction for Surgical Planning

In one recent study, a 3D-reconstructed virtual model of the location of kidney stones and relevant anatomy was compared to conventional CT pre-percutaneous nephrolithotomy. In the group with 3D imaging, a statistically significantly greater stone-free rate (SFR) (81.9% vs 64.2%), first-time puncture success rate (87.5% vs 47.8%), shorter operative time (62 vs 79 minutes), and fewer complications (8.3% vs 25.4%) were observed.5 The authors propose that these benefits are primarily due to the improvement in planning and improved calyceal access. While this single-surgeon, single-center study is at high risk of bias, the ability to visualize complex anatomy preoperatively may prove useful in larger studies.

Similar to 3D renal anatomy reconstruction, utilization of immersive virtual reality (iVR) renal models to improve preoperative planning and patient education has been explored.6 In a separate study, a virtual surgical model was once again created, but this time displayed in an interactive head-mounted oculus rift virtual reality display. Surgeons and patients interacted with the model preoperatively. Surgeons reported that the model altered their surgical approach in 40% of cases, and patients reported increased understanding and reduced preoperative anxiety. However, iVR is cumbersome and challenging to implement into clinical workflow. As well, the creation of 3D iVR models has a steep learning curve which may limit its immediate applicability to research until there is more automation for their generation.

Artificial Intelligence in Stone Outcomes

Automated intelligence applications are near ubiquitous in medicine and surgery. For example, automated radiomic analysis has been shown to predict likelihood of spontaneous stone passage. Using stone measurements, attenuation, and morphologic features, researchers demonstrated that their automated model predicted spontaneous stone passage at a similar rate to manual measurements (AUC 0.82 vs 0.83).7

More broadly, a recent systematic review demonstrated the widespread nature of machine learning models in ESWL.8 Their analysis included 8 articles and concluded that machine learning models functioned at least as well as human-driven models at predicting SFRs post-ESWL. They do note the significant heterogeneity in their study as a limitation.

Following on this research, an engineered machine learning algorithm was developed to identify kidney stones and stone characteristics without the need for human input.9 In this prospective study, 94 patients with confirmed renal calculi on NCCT were included and a deep machine learning algorithm was compared with human measurements from 3 independent reviewers. This algorithm was able to characterize the stones more accurately, reliably, and faster compared to humans (12 seconds for ALL stones on NCCT vs 10 seconds for each stone stone from human reviewers). These results show the improvements that can be made over the current standard of care by automating a laborious process, allowing radiologists more time to focus on higher level clinical decisions.


Research in noninvasive imaging modalities for nephrolithiasis is undergoing a wave of new innovations. DECT scans, 3D stone analysis with calyceal modeling, ureteral attenuation, and machine learning are all areas with significant promise that are changing how clinicians prepare and operate on renal calculi. While many of these are currently used for research applications primarily, they show potential for integration into clinical workflow to improve prognostication and patient outcomes.

  1. Zheng J, Yu H, Batur J, et al. A multicenter study to develop a non-invasive radiomic model to identify urinary infection stone in vivo using machine-learning. Kidney Int. 2021;100(4):870-880.
  2. Bharati A, Rani Mandal S, Gupta AK, et al. Non-invasive characterisation of renal stones using dual energy CT: a method to differentiate calcium stones. Phys Med. 2022;101:158-164.
  3. Deguchi R, Yamashita S, Iwahashi Y, et al. The ratio of CT attenuation values of the ureter above and below ureteral stones is a useful preoperative factor for predicting impacted ureteral stones. Urolithiasis. 2022;50(5):643-649.
  4. Kobayashi M, Waseda Y, Fuse H, Takazawa R. Variables measured on three-dimensional computed tomography are preferred for predicting the outcomes of shock wave lithotripsy. World J Urol. 2021;40(2):569-575.
  5. Tan H, Xie Y, Zhang X, Wang W, Yuan H, Lin C. Assessment of three-dimensional reconstruction in percutaneous nephrolithotomy for complex renal calculi treatment. Front Surg. 2021;8:701207.
  6. Parkhomenko E, O’Leary M, Safiullah S, et al. Pilot assessment of immersive virtual reality renal models as an educational and preoperative planning tool for percutaneous nephrolithotomy. J Endourol. 2019;33(4):283-288.
  7. Mohammadinejad P, Ferrero A, Bartlett DJ, et al. Automated radiomic analysis of CT images to predict likelihood of spontaneous passage of symptomatic renal stones. Emerg Radiol. 2021;28(4):781-788.
  8. Rice P, Pugh M, Geraghty R, Hameed BMZ, Shah M, Somani BK. Machine learning models for predicting stone-free status after shockwave lithotripsy: a systematic review and meta-analysis. Urology. 2021;156:16-22.
  9. Babajide R, Lembrikova K, Ziemba J, et al. Automated machine learning segmentation and measurement of urinary stones on CT scan. Urology. 2022;169:41-46.