Digital simulation of kidney collecting system and stone: geometry and texture mapping

Abstract

Background

Digital simulation is the first step in many, virtual reality, computer vision, and automation pipelines. However, to our knowledge, no open source highly realistic three-dimensional (3D) models of a kidney with a textured collecting system exists. Herein we describe our technique for rendering a high-fidelity digital shadow of the human kidney collecting system with kidney stone.

Methods

Using medical imaging data, iterative 3D modeling, and advanced texture mapping, an anatomically accurate and visually realistic kidney collecting system was produced. Similarly, we generated calcium oxalate monohydrate (COM) and uric acid (UA) stones. The workflow included computed tomography (CT)-derived geometry correction, custom texture design and shader-based rendering. The structures were sent to twenty-three urologists who were given a Likert-rating scale to rate the fidelity of the renderings in terms of the kidney and stone, geometry and texture. Average, median and intraclass correlation coefficients (ICCs) for the responses were calculated to characterize the responses and reliability.

Results

Fifteen urologists responded to each question. The simulated collecting system was generally evaluated as realistic {median 4 [interquartile range (IQR): 3, 5]}. UA stone texture was significantly more realistic than COM texture [median: 4 (IQR: 4, 5) vs. 3 (IQR: 3, 4); U=47.50, P=0.005]. Similarly, there was a trend for more realistic UA vs. COM stone geometry [Median:4 (IQR: 4, 5) vs. Median: 4 (IQR: 3, 4); U=69.5, P=0.06]. Although individual ratings varied widely [ICC (2, 1) =0.10], the aggregated expert consensus was moderate [ICC (2, k) =0.63], supporting overall consensus in model realism.

Conclusions

This work addresses a key gap in realistic anatomical modeling for ureteroscopy, which may support improved training and potentially enable simulation of kidney stone removal procedures. While the simulation was evaluated favorably especially for the UA stone, there is room for improvement of the urothelial mucosa, and texture of the renal papillae, and COM stone.

Highlight box.

Key findings

• Creation of a high-fidelity digital twin of the kidney collecting system and kidney stones for simulated ureteroscopy.

• Validation of model by endourologists with expertise in ureteroscopy, nephrolithiasis, and the endoscopic appearance of the collecting system.

What is known and what is new?

• While surgical simulation is a vital tool for medical education, there is currently a lack of open-source three-dimensional models that feature highly realistic textures of the renal collecting system or common calculi.

• This manuscript describes the development of a high-fidelity virtual representation of the renal collecting system, with performance validated across both textural and geometric dimensions.

What is the implication, and what should change now?

• Accurately replicating expected surgical anatomy of endourologic procedures is the first step for creating digital surgical scenes. These scenes are used to train vision models that enable computer-assisted surgery.

• Similarly, digital simulation is the first step in developing realistic digital surgical trainers, which can enable trainees to refine their techniques without the costs and burdens associated with physical simulation.

Introduction

In the USA, national data continues to show that kidney stones are an increasing healthcare burden, with prevalence increasing from 8.8% in 2007–2010 (1) to 11.0% in 2015–2018 and a 12-month incidence of 2% (2). For stones that will not pass on their own, ureteroscopy (URS)with laser lithotripsy is the most commonly performed procedure (3). Achieving competency in performing the operation has been estimated to require approximately 60 cases (4). However, decreasing learning time and developing simulation platforms is important for medical training because it allows trainees to practice and refine clinical skills in a safe, controlled environment, thereby reducing the risks incurred by learning on patients (5).

Similar to open or laparoscopic surgical techniques, urologic endoscopy is difficult to practice outside of the operating room, hence simulated environments have been developed. This has been studied extensively in the context of robotic-assisted laparoscopic surgery, where training on a simulated environment has been shown to improve resident performance in camera and needle control as well as tissue handling (6,7). Similarly in orthopedic surgery, simulation of the knee facilitated the development of a total knee arthroplasty trainer, which enabled trainees to visualize patient-specific anatomy, practice procedural steps, and receive immediate feedback, thereby improving technical skills (8). In endourologic surgery, high-fidelity visual replication of the renal collecting system and urinary calculi, characterized by both geometric accuracy and textural realism, is a fundamental prerequisite for high-quality digital surgical simulation.

Beyond the training of medical students and residents, high fidelity simulation also affords the development of computer vision models and control algorithms to develop surgical assistance (9). Therefore, the present work focuses on creating a visually realistic, medically accurate three-dimensional (3D) digital twin of the renal collecting system and a kidney stone for a diverse array of simulation-based training. We present this article in accordance with the SURGE reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0157/rc).

Methods

Kidney model development

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of University of Miami (IRB #20180511). Informed consent was not required for this retrospective chart review. In this study, we extracted the digital imaging and communications in medicine (DICOM) data from a single patient’s computerized tomography performed at University of Miami Health System to evaluate for possible causes of microscopic hematuria. The steps for the development of kidney stone model are depicted in Figure 1.

Figure 1.

Flow chart for the steps followed for the development of the kidney model. DICOM, digital imaging and communications in medicine.

We ensured there was no genitourinary abnormality and then imported the scan into RadiANT^TM, a DICOM viewer. There, the delayed phase was segmented and exported as an.stl, which was imported into Blender for further editing. Using Adobe Photoshop and hand-drawings, a texture was created to replicate the urothelium of the collecting system in high detail and color. A custom shader node graph was constructed using layered albedo, roughness, and subsurface-scattering channels, with node-based modulation used to emphasize vascular gradients and replicate endoscopic light falloff (Figure 2). The mesh contained artifacts or distortions requiring repair: correction of the vertices, closing of the surface holes, and rescaling to match the expected geometry. Fine-scale topology smoothing and proportional-editing tools were applied iteratively to restore anatomical curvature, eliminate segmentation artifacts, and ensure that the surface was suitable for visually realistic texture adherence (Figure 3). Texture mapping was performed using Blender’s automatic UV-unwrapping tools, which generated optimized UV islands with minimal stretching across the curved surfaces of the collecting system.

Figure 2.

Custom shader node architecture for realistic rendering. (A) Front view of developed model. (B) Perspective view of developed model. (C) Internal view of developed model. (D) Texture PNG with vascular and urothelial details. (E) Arrangement of shader editor nodes in Blender via visual scripting.

Figure 3.

Edited external mesh with applied internal texture of collecting system. (A) Raw model (internal view). (B) Raw model with texture mapping. (C) Corrected model (external view). (D) Corrected model with texture mapping.

Kidney stone model development

Two representative kidney stone types, calcium oxalate monohydrate (COM), and uric acid (UA) stones, were selected because they are among the most common in patients, yet remain visually and structurally distinct from each other. A layered texture mapping approach combined albedo maps for base color in order to capture the subtle tonal variations such as the yellowish-brown surface of COM stones or the darker, orange hues of UA stones. For reflectivity we utilized roughness maps. COM stones are assigned a high roughness variance to simulate their glittery, crystalline quality under lighting. To simulate the geometrical differences between rough COM and smooth UA stones, we used a combination of a Normal Map, which simulates fine surface details to create the illusion of tiny lacunae, and a Displacement Map to add actual depth by modifying the geometry to distinguish the smoother UA stone from the edgier COM stone (Figure 4). Both the kidney and stone structures are available as .usd files at https://github.com/uro-glidar/kidney-com-ua-stone-usds.

Figure 4.

Kidney stone mesh and texture. (A) Calcium oxalate monohydrate stone. (A1) Pre-texturing with vertices shown. (A2,A3) Texturing applied. (B) Uric acid stone. (B1) Pre-texturing with vertices shown. (B2,B3) Texturing applied.

Statistical analysis

The assets were optimized for real-time simulation and then static images were sent out to 23 endourologists. Each endourologist was given a Likert rating scale to rate the fidelity of the renderings in terms of kidney geometry, collecting system texture, and for both the COM and UA stone geometry and texture. The Likert rating was on a scale of 1–5 (1—strongly disagree, 2—disagree, 3—neutral, 4—agree, 5—strongly agree). We then calculated the mean, standard deviation and median for each question. Normality of the data distribution was assessed using the Shapiro-Wilk test and because the data were not normally distributed, we used the Mann-Whitney U test to compare the difference in accuracy of representation of stone geometry and texture between the COM and UA stones. Inter-rater agreement among experts was evaluated using the intraclass correlation coefficient (ICC) (two-way random effects, absolute agreement model). All analyses were conducted at the level of individual items as well as pooled domain scores (texture realism, geometric realism). All statistical analysis performed with Python 3.13.

Results

The single patient’s CT scan was successfully extracted using the aforementioned pipeline and texture was successfully applied as outlined above. Similarly, both the COM and UA stones were able to be rendered.

Fifteen out of 23 urologists responded to all questions. The simulated collecting system geometry {mean: 3.80±1.32, median: 4 [interquartile range (IQR): 3, 5]} and texture [mean: 3.80±0.86, median: 4 (IQR: 3, 4)] were generally evaluated as reflecting reality. Regarding the kidney stones, the UA stone texture was found to be significantly more realistic than the COM stone [mean: 4.27±0.70, median: 4 (IQR: 4, 5) vs. mean: 3.27±0.96, median: 3 (IQR: 3, 4); U=47.50, P=0.005]. Similarly, the geometry of the UA stone was found to be more realistic than the COM stone, though this did not meet statistical significance [mean: 4.13±0.83, median: 4 (IQR: 4, 5) vs. mean: 3.53±0.92, median: 4 (IQR: 3, 4); U=69.50, P=0.06] (Table 1).

Table 1. Survey responses evaluating the realism of the simulated collecting system and stones.

Question	Average response (±SD)	Median (IQR)
The simulated collecting system’s geometry is an accurate representation of kidney anatomy	3.80±1.32	4 (3, 5)
The simulated collecting system’s urothelial lining accurately reflects the endoscopic view of the collecting system	3.80±0.86	4 (3, 4)
The simulated calcium oxalate stone geometry reflects reality	3.53±0.92	4 (3, 4)
The simulated calcium oxalate stone texture reflects reality	3.27±0.96	3 (3, 4)
The simulated uric acid stone geometry reflects reality	4.13±0.83	4 (4, 5)
The simulated uric acid stone texture reflects reality	4.27±0.70	4 (4, 5)

IQR, interquartile range; SD, standard deviation.

Interrater reliability was assessed using a two-way random-effects ICC with absolute agreement. The ICC for single raters [ICC (2, 1)] was 0.10, indicating poor agreement among individual experts. However, the reliability of the average rating across the 15 raters [ICC (2, k)] was 0.63, reflecting moderate consensus and reliability (P=0.002). These results indicate that while individual judgments varied, the collective expert assessment was moderately consistent.

Discussion

This work addressed a critical gap in high-fidelity anatomical modeling with specific attention to geometry and texture for surgical simulation by combining advanced graphic techniques, cross-platform integration, and domain-expert validation. Relevant research in texture mapping and surgical simulation (10,11) informed the choice of modeling strategies, ensuring that the visual and anatomical accuracy aligned with current best practices. Beyond these methods, dynamic volume texture mapping techniques developed for deformable surgical models (12), along with voxel-based texture generation methods used in volumetric rendering (13), provide complementary strategies for medical simulation, but neither captures the fine-grained optical detail required for realistic endoscopic visualization of the renal collecting system.

We found that the UA stone simulation was the most realistic, with its texture being statistically significantly more realistic than the COM stone. Furthermore, the median score by endourologists was agreeing that the texture and geometry of the kidney’s collecting system and both COM and UA stones accurately reflected reality, with the exception of COM texture whose median score was only neutral.

We measured ICC to assess the agreement between endourologists to ensure the reliability of our findings. While the ICC for single raters was poor, when averaging across raters we found a moderate to good agreement in perceived realism, leading us to feel confident that these simulated stones and kidney collecting system are realistic.

Several other strategies have been utilized for texturing organs. Azkue described a surface-guided color sampling method for creating renderings of human anatomy but utilized cadaveric specimens, with obvious limitations regarding vascularity and hence coloring, limiting the ability to generate realistic high-fidelity rendering of in vivo human collecting system (14). In contrast, Selver and Caban et al. generated geometry of abdominal organs by using volumetric medical imaging data (such as CT or MRI) (15,16), similar to the technique we used. Selver textured the organs by applying a technique called brushlet expansion, which analyzes the data to capture both fine and coarse details to assign colors and opacities, while Caban et al. used volumetric differences in neighboring voxels to assign texture (15,16). While these techniques work well for distinguishing organs it does not capture the subtleties that make a 3D rendering as realistic as seen during endoscopic procedures.

Limitations of our work include that we only simulated a single kidney, however this open-source pipeline can be applied to any number of kidneys and can be used to create custom rendering for individual patients based on their CT scans. Additionally, though not shown, no renal papilla was created given difficulty with generating a realistic geometry and texture. Similarly, we only rendered two types of kidney stones COM and UA. While they are the among most common, many more stones exist and often in reality most kidney stones are actually of mixed composition (17). Lastly, our sample of surveyed endourologists was small and only included static images, which limits both the quality of the evaluation of a 3D rendering and the statistical power of our findings. The framework presented here serves as a foundation for generating more complex digital surgical scenes. Its ultimate utility will be measured by its effectiveness in two key domains: the development of computer vision-based surgical assistance and its role as a high-fidelity adjunct in pre-clinical training for medical students and residents. Future work will focus on integrating more complex physical dynamics, including the simulation of variable stone hardness and the anatomical realism of soft tissue deformation during instrument interaction.

Conclusions

This work provides a critical foundation for developing a digital simulation model for training surgical residents and computer vision algorithms. The process leveraged theoretical knowledge from computer graphics and applied it to medical modeling, texture theory, shader programming, and UV mapping, resulting in validated kidney and stone models. The endourologists surveyed for this study found the geometry and texture of the simulated collecting system and kidney stones to be visually realistic compared to in vivo ureteroscopy. We hope that these validated, open-source models will provide the foundational architecture for generating more complex synthetic surgical scenes. Such environments are critical for the development of computer vision-based intraoperative assistants and high-fidelity simulators for pre-clinical training.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of University of Miami (IRB #20180511). Informed consent was not required for this retrospective chart review.

Data Sharing Statement

Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0157/dss

tau-15-06-195-dss.pdf ^{(72.9KB, pdf)}

DOI: 10.21037/tau-2026-1-0157