Abstract

Lowery urinary tract symptoms (LUTS) affect a large majority of the aging population. 3D Dynamic MRI shows promise as a noninvasive diagnostic tool that can assess bladder anatomy and function (urodynamics) while overcoming challenges associated with current urodynamic assessment methods. However, validation of this technique remains an unmet need. In this study, an anatomically realistic, bladder-mimicking in vitro flow model was created and used to systematically benchmark 3D dynamic MRI performance using a highly controllable syringe pump. Time-resolved volumes of the synthetic bladder model were obtained during simulated filling and voiding events and used to calculate volumetric flowrate. During MRI acquisitions, pressure during each event was recorded and used to create PV loops for work assessment. Error between control and MRI-derived volume for voiding and filling events exhibited 3.36% and 4.66% differences, respectively. A slight increase in average error was observed for MRI-derived flowrate when compared to the control flowrate (4.90% and 7.67% for voiding and filling, respectively). Overall, average error in segmented volumes increased with decreasing volume flowrate. Pressure drops were observed during voiding. Pressure increased during filling. Enhanced validation of novel 3D MRI urodynamics is achieved by using high-resolution PIV for visualizing and quantifying velocity inside the bladder model, which is not currently possible with 3D Dynamic MRI.

Introduction

Lower urinary tract symptoms (LUTS) and changes in bladder function often occur as individuals age [14]. LUTS affect the majority of the aging population and individuals affected by LUTS may experience symptoms of bladder and urethral dysfunction pertaining to incontinence/storage or voiding dysfunction (diminished stream or emptying) [14]. Specifically in men, these issues are also attributed to obstruction from an enlarged prostate also known as benign prostatic hyperplasia (BPH) [3,4].

Individuals with LUTS are commonly evaluated through multichannel urodynamic studies (UDS) to determine bladder flow and pressure during voiding and filling to gain insight into the progression and severity of symptoms [5,6]. However, these tests are invasive and do not provide comprehensive (3D) flow visualization or quantification. Additionally, insight into anatomical changes of the bladder as it contracts or expands is often inferred [7]. As a result, there is an imperative need to develop a comprehensive, noninvasive diagnostic method to assess bladder voiding and filling performance in real time to improve LUTS assessment and treatment planning.

To overcome these challenges recent efforts have focused on utilizing noninvasive imaging of the bladder based on magnetic resonance imaging (MRI) or ultrasound to more comprehensively and quantitatively assess bladder anatomy and function [811]. Three-dimensional, time-resolved representations of the bladder can be acquired as the bladder fills or voids. These data can be used to quantify changes in bladder morphology and volume in real-time, while also assessing anatomical changes of the entire lower urinary tract (bladder, prostate, and urethra). Additionally, information obtained from MRI can be used to drive computational fluid dynamics (CFD) simulations to analyze flow, pressure, and wall motion with high spatial and temporal resolution [12,13].

Although 3D Dynamic MRI shows promise as a tool to comprehensively assess bladder and lower urinary tract biomechanics, systematic validation of this technique remains an unmet need before expanded clinical use. To accomplish this, anatomically realistic, in vitro models can be used to validate experimental MRI measurements and calculated flow dynamics. in vitro flow studies are recognized as the gold standard for flow-based MRI [14]. Using an in vitro modeling approach, systematic control of anatomy and flow conditions can be achieved and used to directly compare MRI quantification to reference standard quantitative methods. Additionally, in vitro flow models can be evaluated with optical imaging techniques, such as particle image velocimetry (PIV). PIV is a velocity-sensitive flow acquisition technique that has been employed for patient-specific flow phantoms as a reference standard for flow quantification benchmarking [1518]. As it applies to this study, PIV can be used to enhance bladder biomechanics assessment through quantification of fluid behavior and velocity inside a model with high spatio-temporal resolution, the latter of which is currently impossible with 3D Dynamic MRI [15,16].

Therefore, the aim of this study was to develop an MRI- and PIV-compatible anatomically realistic, in vitro model of the bladder to validate and expand upon a recently developed real-time volumetric (3D) MRI protocol [19]. The modeling pipeline is as follows:

  1. Develop an anthropomorphic, in vitro, flow model of the bladder using a hybrid additive-molding manufacturing process.

  2. Acquire 3D volumes of the bladder during simulated voiding and filling events that can be used to benchmark volumetric MRI performance (total volume change and volumetric flow rate).

  3. Using invasive, but MRI-compatible pressure sensors, acquire voiding and filling pressure data at the bladder model outlet allowing for the determination of filling/voiding energetics, such as work, during 3D MRI acquisitions.

  4. Enhance validation through the creation of a PIV-compatible experimental pipeline which allows for the capture of time-resolved velocity distributions inside the bladder model.

Materials and Methods

In Vitro Model Fabrication.

In this IRB-approved, HIPPA-compliant study (University of Wisconsin-Madison, Ethical Approval Number 2021-1247), a healthy, 37-year-old male subject with no history of LUTS was recruited to void in the scanner in the supine position during a 3D dynamic MRI acquisition based on a 3D Differential Subsampling with Cartesian Ordering (DISCO) Flex acquisition sequence [13,19]. A flexible, transparent in vitro patient-specific model of the bladder was constructed using a hybrid additive-casting manufacturing process as follows:

  1. Full bladder segmentation from 3D Dynamic MRI dataset was obtained at the initiation of voiding to generate computerized model (Fig. 1(a)).

  2. Manual manipulation of bladder model to add a synthetic, idealized urethra. This model serves as the “core” of what will become the final bladder flow model. (Fig. 1(b)).

  3. Creation of a hollow mold, which is generated via subtraction of the core from a rectangular block (Fig. 1(c)).

  4. Careful placement and sealing of the mold around the core so that the bladder model sits at the center of the mold.

  5. Casting of silicone into the gap between the mold and core.

  6. Curing of silicone at 100 °C for 35 min (Fig. 1(d)). Removal of core mold cast in silicone and dissolving process of core using water (Fig. 1(e)).

Fig. 1
Flexible, transparent, in vitro bladder model constructions process. Volumetric MRI images were acquired for a healthy, male volunteer (a). The MRI data was used to generate a computerized model of the bladder, where a synthetic urethra of 6.3 mm and fixturing components were added (b, c). This component serves as the molding core and was printed fabricated using water-soluble PVA material. The bladder anatomy was enlarged and subtracted from a block to create a mold to facilitate a casting process. After assembly, silicone was poured into the mold cavity and cured (d). The final step of the process involved dissolving the core with water and removing it from the mold, producing a hollow bladder flow model that can be used in MRI and PIV experimentation (e).
Fig. 1
Flexible, transparent, in vitro bladder model constructions process. Volumetric MRI images were acquired for a healthy, male volunteer (a). The MRI data was used to generate a computerized model of the bladder, where a synthetic urethra of 6.3 mm and fixturing components were added (b, c). This component serves as the molding core and was printed fabricated using water-soluble PVA material. The bladder anatomy was enlarged and subtracted from a block to create a mold to facilitate a casting process. After assembly, silicone was poured into the mold cavity and cured (d). The final step of the process involved dissolving the core with water and removing it from the mold, producing a hollow bladder flow model that can be used in MRI and PIV experimentation (e).
Close modal

Segmentation was performed using the software package MIMICS/3-Matic (Materialise, Leuven, Belgium) which allowed for 3D bladder segmentation and manipulation. The computerized model was smoothed and exported into stereolithography (STL) format. A 20 mm long, 6.3 mm diameter cylindrical rod was added at the base of the bladder STL to create an outlet representing a urethra. The urethra geometry was selected to represent the mean diameter of the membranous urethra, typically the narrowest portion of the urethra [20]. This constituted the final bladder geometry that served as the flow model and core for casting. The mold was generated via subtraction of the original bladder model from a block and designed to allow a mean gap spacing of 4 mm between the core and mold, representing the anatomical thickness of the bladder wall [21]. The bladder core was fabricated using desktop material-extrusion-based 3D printing and constructed using water-soluble polyvinyl alcohol (PVA) (Ultimaker, Utrecht, The Netherlands) (Fig. 1(b)). The mold was printed using a desktop stereolithography (SLA) printer (Formlabs, Somerville, MA) (Fig. 1(c)).

To physically manufacture the hollow flow model, the mold was placed around the core and sealed using watertight sealant. Once sealed, optically clear silicone (Sylgard-184, Dow Chemical, Midland, MI) was poured inside the mold, placed in an oven at 100 °C, and cured over a timespan of 35 min The choice of silicone is important, as it provides both flexibility and a refractive index (R.I.) that can be matched with the fluid required for PIV experimentation (R.I. = 1.41). Once the silicone was cured, the bladder core was dissolved with water, leaving an anatomically realistic, hollow, in vitro bladder flow model.

Flow Experimental Setup.

To simulate the voiding or filling cycle, the transparent silicone bladder model was connected to a syringe pump (Chemyx, Stafford, TX). 6.35 mm diameter × 1.59 mm thick (1/4″ × 1/16″) vascular tubing was connected between the outlet of the bladder model (base) and the syringe pump. Because ferrous material cannot be brought inside the MRI scan room, 8 m of tubing was required to link the syringe pump housed in the control room (Fig. 2(a)) and the bladder phantom placed in the MRI scanner (Fig. 2(b)). The working fluid for MRI experimentation was water. Prior to MRI acquisition, the synthetic bladder and flow tubing were manually filled until full (approximately 270 cc total), taking special care to minimize and remove any air in the system. Before data collection, a 300 cc syringe (diameter = 48mm) was filled with 100 cc of water and connected to the urethra tubing. After placement, water could be withdrawn or injected into the bladder model to simulate voiding or filling events, respectively. Using the syringe pump, the total voided, or filled, volume and volume flowrate could be prescribed and controlled during MRI acquisitions.

Fig. 2
Schematic of 3D Dynamic MRI experimental setup consisting of flow loop driven by programable syringe pump to direct flow. The syringe pump sits inside the control room (a). 6.35 mm (1/4″) tubing connects the syringe pump to the bladder model sitting in the MRI scanner (b). A slip-fit pressure port is placed just upstream of the connection between the bladder model and tubing, allowing for an MRI-compatible pressure transducer to be fed into the bladder model. Example image acquired using bSSFP (c) MRI sequence is shown.
Fig. 2
Schematic of 3D Dynamic MRI experimental setup consisting of flow loop driven by programable syringe pump to direct flow. The syringe pump sits inside the control room (a). 6.35 mm (1/4″) tubing connects the syringe pump to the bladder model sitting in the MRI scanner (b). A slip-fit pressure port is placed just upstream of the connection between the bladder model and tubing, allowing for an MRI-compatible pressure transducer to be fed into the bladder model. Example image acquired using bSSFP (c) MRI sequence is shown.
Close modal

In Vitro MRI Acquisitions.

For all tests, the bladder model was placed inside a clinical 3.0T MRI scanner (GE Healthcare, Waukesha, WI) using a high-density flexible surface coil array (AIR Coil, GE Healthcare, Waukesha, WI), where real-time, volumetric MRI data could be acquired as the model was filled or emptied. For every MRI acquisition, 150 cc was either withdrawn or injected. The dynamic MRI acquisition was performed using a 3D Balanced Steady-State Free Procession (bSSFP) acquisition sequence optimized for noncontrast urine imaging. The temporal resolution for the bSSFP sequence was 1.65 s, which is defined as the time to acquire one whole bladder volume. The total acquisition time was tailored to the time required to completely inject or withdraw the total volume of fluid from the model as prescribed by the volumetric flowrate of the syringe pump. This determines the number of bladder volumes acquired during each MRI acquisition. A summary of acquisition parameters for the bSSFP MRI sequence is shown in Table 1.

Table 1

Summary of volumetric MRI acquisition parameters

MRI acquisition sequence
 Dynamic MRI Sequence3D Balanced Steady-State Free Procession (bSSFP)
Acquisition parameters
 TR, repetition time (ms)2.5
 TE, echo time (ms)0.848
 Flip angle (°)20
 Scan planeSagittal
 Spatial resolution (x,y) (mm)1
 Slice thickness (mm)2
 Voxel volume (mm3)1 × 1 × 2
MRI acquisition sequence
 Dynamic MRI Sequence3D Balanced Steady-State Free Procession (bSSFP)
Acquisition parameters
 TR, repetition time (ms)2.5
 TE, echo time (ms)0.848
 Flip angle (°)20
 Scan planeSagittal
 Spatial resolution (x,y) (mm)1
 Slice thickness (mm)2
 Voxel volume (mm3)1 × 1 × 2

Pressure Measurements.

To facilitate direct pressure measurements at a location inside the bladder, an MRI-compatible, fiber-optic pressure transducer (OPP-M200, Opsens Inc., Québec, CA) was inserted into the bladder through a slip-fit connector downstream of the outlet of the model. The location of the transducer tip was placed approximately 1 cm inside the model and fixed in location which allowed for pressure measurements at the bladder model outlet. Because the pressure transducer is placed inside the bladder model, the recorded pressure can be likened to intravesical pressure that can be determined by invasive catheterization during standard urodynamics studies [22]. Pressure measurements were acquired for the entirety of each flow test at a frequency of 100 Hz. Before each flow test the pressure system was zeroed, reflecting that the pressure reading measures the relative change in pressure due to the fluid and bladder system acting to withdraw or inject fluid during each event.

Optical Imaging Experiments—Particle Image Velocimetry.

Particle image velocimetry-velocities were acquired for one voiding and one filling event using a flowrate of 5.37 cc/s controlled using the syringe pump. The PIV flow setup consisted of the bladder flow model surrounded by an acrylic container that was open to atmospheric pressure. PIV data were acquired using a Flowmaster system (LaVision, Göttingen, Germany) and consisted of a dual-pulse 527 nm Nd:YLF laser opposite one high-speed camera (Phantom VEO E-340 L, Vision Research, NJ), oriented perpendicular to the imaging plane (Fig. 3(a)). The camera was equipped with a 60 mm f/2.8D lens (Nikon Inc., Melville, NY). The acrylic box was filled with fluid matching the refractive index of the silicone model (R.I. = 1.41). This fluid consisted of a mixture of glycerol and water (58:42% by weight) that our lab typically employs for index-matching based PIV experimentation [16]. This fluid was also utilized to fill the bladder model, however, it was seeded with polymethyl methacrylate (PMMA) particles (diameter = 10μm) coated with Rhodamine B dye. A laser sheet of 2 mm thickness was projected and aligned near the center of the bladder phantom which illuminated particles inside the bladder model (Fig. 3(b)). To further limit the effect of refraction, the camera was equipped with a 570 nm cutoff long-pass filter, which is tailored to the wavelength of light that can be captured from the fluorescent particles. 360 double-frame image pairs were obtained using an imaging rate of 12 Hz. The time between the two frames (dt) was tailored to achieve an average particle displacement of 7-10 pixels between image pairs [23]. Camera calibration was carried out using a scaling operation where an image of a ruler was placed inside a custom-made silicone housing (Sylgard-184) and surrounded by the index-matching fluid utilized in the flow experiments, which reproduced the optical conditions of the PIV flow experiments and maintained consistency between the PIV calibration and acquisition procedures.

Data Analysis

All pressure and volume data were exported to a custom MATLAB script to calculate total volume change, flowrate, root-mean-squared error (RMSE) and construct PV-loops.

3D Dynamic MRI.

Segmentation was performed for all MRI volumes acquired during each voiding or filling test using MIMICS, resulting in one computed volume for each timepoint of the MRI acquisition (one bladder volume rendering every 1.65 s). During segmentation, only the volume of fluid inside of the bladder model was considered and any resulting fluid inside the urethra was manually removed from each volume. The results of these segmentations were used to calculate time-resolved volumes for each test. The time-resolved volumetric flowrate in the bladder model was determined by calculating the difference between two successive volumes and the temporal resolution of the bSSFP MRI sequence. The error between pump and MRI flowrate at each time-step was determined and averaged to yield the average flowrate error. The first and last segmented volumes for each test were used to determine total voided, or filled, volume. The time-resolved volume change and flowrate obtained from MRI were then directly compared to the control settings of the syringe pump using the RMSE.

Pressure.

To effectively analyze pressure as a function of the volume change inside the bladder model, pressure data were smoothed using a low-pass filter and resampled to match the temporal resolution of the bSSFP MRI acquisitions. Acquiring pressure during each MRI acquisition also allowed for the generation of pressure–volume (PV) loops. From the area under the PV curve, a calculation of work can be carried out according to Eq. (1)
W=PdV
(1)

Particle Image Velocimetry.

Particle image velocimetry images were analyzed and reconstructed using Davis 10 postprocessing software (LaVision, Göttingen, Germany) (Fig. 3(c)). Acquired particle images were optimized using a combination of local minimum background subtraction, Gaussian smoothing (3 × 3), and local intensity normalization. These reduced unwanted effects resulting from light scattering at the fluid-model interface. To calculate velocity vectors from particle volumes, two-dimensional (2D) cross-correlation was performed using multipass iteration, with a starting window size of 128 × 128 and a final window size of 8 × 8 in five total iterations. This resulted in a final spatial resolution of 0.41 mm. Cut planes at the bladder model base could be used to calculate an average fluid velocity that could be used to determine volumetric flowrate from the cross-sectional area of the model at the location of the velocity plane. The large-scale deformation of the bladder wall required the use of a temporal masking procedure to restrict the calculation region to the region of interest that only contained fluid remaining inside of the bladder at each time point. This filtered the model boundary from interfering with the velocity calculations.

Results

3D MRI Volume Analysis.

3D renderings of the in vitro bladder model were successfully acquired using the bSSFP (Fig. 2(c)) MRI sequences for three voiding and two filling events. Figure 4 depicts volume renderings at the start, 25%, 50% and 75% of the total volume change, for a voiding (Fig. 4(a)) and filling (Fig. 4(b)) cycle with prescribed volumetric flow rates of 5.37 and 1.67 cc/s, respectively. Results comparing the pump control and MRI volume analysis for all flow tests are presented in Table 2.

Fig. 3
Setup for particle image velocimetry (PIV) experiments. An open-top box was designed to house the bladder model. The same fluid utilized inside the model is poured inside the box, surrounding the bladder with index-matched fluid to limit optical distortions for the laser light path (a). The laser is placed adjacent to the phantom so that a rectangular laser sheet transects the model at its center (cutting a plane through and just above the outlet (urethra). A high-speed camera acquires images of illuminated particles inside the model (b) that can be used derive PIV-velocity maps (c).
Fig. 3
Setup for particle image velocimetry (PIV) experiments. An open-top box was designed to house the bladder model. The same fluid utilized inside the model is poured inside the box, surrounding the bladder with index-matched fluid to limit optical distortions for the laser light path (a). The laser is placed adjacent to the phantom so that a rectangular laser sheet transects the model at its center (cutting a plane through and just above the outlet (urethra). A high-speed camera acquires images of illuminated particles inside the model (b) that can be used derive PIV-velocity maps (c).
Close modal
Fig. 4
3D bSSFP MRI-derived volumes during (a) voiding and (b) filling events. One volume is acquired every 1.65 s and the total acquisition time was tailored to the speed of the bladder flow state being simulated. Volumes were calculated from the anatomical segmentation. Volumetric flowrate was determined from the difference between two successive volumes (ΔVstep) and the time between each volume acquisition (Δt). Total voided, or filled, volume was calculated via the difference between the initial and final volumes of each acquisition.
Fig. 4
3D bSSFP MRI-derived volumes during (a) voiding and (b) filling events. One volume is acquired every 1.65 s and the total acquisition time was tailored to the speed of the bladder flow state being simulated. Volumes were calculated from the anatomical segmentation. Volumetric flowrate was determined from the difference between two successive volumes (ΔVstep) and the time between each volume acquisition (Δt). Total voided, or filled, volume was calculated via the difference between the initial and final volumes of each acquisition.
Close modal
Table 2

Summary of bSSFP 3D MRI-derived total voided, or filled, volume and volumetric flow rates compared to those imposed by the pump during MRI experiments

Bladder eventImposed volume change (cc)Imposed flowrate (cc/s)3D MRI flowrate (cc/s)3D MRI volume change (cc)Average flowrate error (%)Average volume error (%)RMSE flowrate (cc/s)RMSE volume (cc)
Voiding1505.375.24 ± 1.72147.72.581.570.862.1
2.502.63 ± 0.64148.65.390.940.641.0
0.830.89 ± 0.18138.76.727.560.200.32
Filling1501.671.78 ± 0.25154.37.342.880.280.47
0.830.90 ± 0.25140.38.006.440.260.42
Bladder eventImposed volume change (cc)Imposed flowrate (cc/s)3D MRI flowrate (cc/s)3D MRI volume change (cc)Average flowrate error (%)Average volume error (%)RMSE flowrate (cc/s)RMSE volume (cc)
Voiding1505.375.24 ± 1.72147.72.581.570.862.1
2.502.63 ± 0.64148.65.390.940.641.0
0.830.89 ± 0.18138.76.727.560.200.32
Filling1501.671.78 ± 0.25154.37.342.880.280.47
0.830.90 ± 0.25140.38.006.440.260.42

Pressure Analysis.

The fiber optic pressure transducers utilized in this experiment allowed for the capture of high-temporal resolution pressure traces during each MRI experiment. Figure 5 depicts the pressure traces for the two filling, and three voiding, events in this study. Pressure drops are seen during voiding while pressure increases are observed during filling. Consistency between the pressure curves for each voiding effort is observed (Fig. 5(a)). Voiding pressure drops varied between 8.96 and 10.2 cmH2O. Filling pressure curves exhibited more discrepancy when comparing repeated tests at different flow rates (Fig. 5(b)). Filling pressure increases of 10.4 and 5.54 cmH2O for the 1.67 and 0.83 cc/s filling rates were observed, respectively. For both voiding and filling events, an increase in work done is associated with higher flowrate events. Work done in voiding and filling was 76.1 and 99.7 millijoules (mJ), respectively. Figure 6 outlines the difference between voiding and filling PV-loops for the same volume flowrate of 0.83 cc/s. The difference in work done between these tests was 27.9 mJ, highlighting a slight difference between voiding and filling states for the same volume flowrate. A summary of all pressure results is summarized in Table 3.

Fig. 5
PV-loops (pressure as a function of volume change) acquired during MRI acquisitions for each voiding (a) and filling (b) event. Pressure is acquired using fiber optic pressure sensors and volume from dynamic MRI.
Fig. 5
PV-loops (pressure as a function of volume change) acquired during MRI acquisitions for each voiding (a) and filling (b) event. Pressure is acquired using fiber optic pressure sensors and volume from dynamic MRI.
Close modal
Fig. 6
Comparison of voiding and filling PV-loops for sequential tests conducted at the same volumetric flowrate (0.83 cc/s). Clear differences are observed between voiding and filling, where the pressure drop induced in voiding is not fully recovered during filling of the bladder model. Pressure is acquired using fiber optic pressure sensors and volume change from dynamic MRI.
Fig. 6
Comparison of voiding and filling PV-loops for sequential tests conducted at the same volumetric flowrate (0.83 cc/s). Clear differences are observed between voiding and filling, where the pressure drop induced in voiding is not fully recovered during filling of the bladder model. Pressure is acquired using fiber optic pressure sensors and volume change from dynamic MRI.
Close modal
Table 3

Summary of pressure-derived metrics calculated from fiber optic pressure measurements during each filling or voiding event

Bladder eventFlowrate (cc/s)ΔP (cmH2O)Work (mJ)
Voiding5.379.5871.6
2.5010.264.8
0.838.9649.7
Filling1.6710.499.7
0.835.5454.2
Bladder eventFlowrate (cc/s)ΔP (cmH2O)Work (mJ)
Voiding5.379.5871.6
2.5010.264.8
0.838.9649.7
Filling1.6710.499.7
0.835.5454.2

Particle Image Velocimetry Velocity Quantification.

Particle image velocimetry-derived velocity maps obtained for a 5.37 cc/s voiding and filling test are illustrated in Fig. 7. During voiding, PIV velocity distributions show a region of slow flow near the top of the model (anatomical dome) that accelerates toward the model outlet (anatomical base) as the dome deforms downward. This contrasts with the PIV velocity maps for the filling test, which shows a region of jetting flow originating from the base of the model that travels to the top of the bladder and impacts the dome. This results in the formation of flow vortices in the lateral sections of the bladder model. Peak velocities above 20 mm/s are observed for both voiding and filling. In voiding, average velocity at the base (exit) of the model was 1.2 cm/s while average flowrate was 4.87 cc/s, which differed from prescribed flowrate by 9.31%.

Fig. 7
PIV-derived velocity maps inside the bladder model during the PIV-specific voiding and filling events. Velocities were quantified at four select timepoints corresponding to the start, 25%, 50%, and 75% of the total voided volume (150 cc). Pump control flowrate was 5.37 cc/s.
Fig. 7
PIV-derived velocity maps inside the bladder model during the PIV-specific voiding and filling events. Velocities were quantified at four select timepoints corresponding to the start, 25%, 50%, and 75% of the total voided volume (150 cc). Pump control flowrate was 5.37 cc/s.
Close modal

Discussion

To our knowledge, this is the first study to establish an experimental pipeline to investigate the validity of 3D MRI for urodynamic assessment of bladder voiding and filling using an anatomically realistic (thin-walled), in vitro model of the bladder. Very few studies to date have utilized dynamic 3D MRI to investigate bladder urodynamics, highlighting the need for thorough and systematic evaluation of MRI performance applied to this territory of the body. We successfully fabricated a distensible and compliant silicone-based in vitro flow model based on an anatomically realistic bladder. To maintain anatomical consistency, care was taken to ensure bladder wall thickness, urethral diameter, and total bladder volume were maintained in the fabrication process.

To systematically assess how well dynamic MRI could resolve bladder volumes and derive flowrate, multiple voiding, and filling events were studied and compared to control parameters from a highly controllable syringe pump. While validation of MRI is challenging to achieve, this experimental pipeline overcomes this difficulty as we know exactly what is injected into, or taken out of, the bladder flow model at any given time point. The results show that, on average, errors for total volume change and volumetric flowrate calculated from 3D MRI for all tests were within 10% when compared to prescribed values. While no previous study assesses the validity of 3D MRI in the bladder, the error trends match previously published studies establishing the validity of flow-based MRI in other regions of the body, such as those examining flow-based MRI to assess hemodynamics in the cardiovascular system [24]. Trends in RMSE for both voiding and filling display an increasing RMSE value as flowrate increases. We postulate that the reason for this is that at lower flow rates MRI can better capture the time-resolved volume change. Additionally, as flowrate increases fewer volumes are used to determine the time-resolved volume change and flow rates per test and higher variability in these volume measurements may lead to higher measurement error. Examining average errors in total volume change and computed volumetric flowrate we tend to see the opposite effect we see with RMSE and higher average errors are observed for lower flow rates. This may be due to human error effects which are hard to quantify, such as errors in volume segmentation of the MRI datasets or volume averaging errors that could influence this measurement. Further analysis of additional experimental conditions (such as including a wider range of flow rates) may be needed to more rigorously investigate the resolution limits of 3D MRI applied to dynamic quantification of the bladder.

To enhance 3D dynamic MRI for urodynamic assessment, we employ a pressure acquisition system to acquire time-resolved pressure measurements during MRI acquisitions in the bladder model with the use of fiber optic pressure transducers. Although this technique is invasive, this strategy allows us to directly and simultaneously examine the pressure and volume relationship in our experimental setup during MRI, which is typically challenging. PV loops are not new to the bladder, but to our knowledge, this is the first time MRI-based volumes have been utilized to investigate this relationship. Previous studies have examined PV-loops in the bladder to investigate work done by the bladder during micturition and shed light on energy storage by the bladder wall during filling [25]. In cardiovascular applications, the PV relationship is often examined as a tool to assess ventricular function of the heart [26]. In examining the pressure and work done in our system across all events, higher pressures were associated with higher flow rates, suggesting it takes more effort to deform the model as flowrate increases. It is interesting that while the voiding PV-loops are relatively consistent, filling PV-loops show inconsistencies. For the two filling cases studied, over a twofold increase in pressure drop between the 1.67 and 0.83 cc/s volume flow cases is observed. This explains the stark difference between the work done during filling in the system for these two cases. Overall, voiding pressure was relatively insensitive to volume flowrate, suggesting that the in vitro bladder may deform similarly regardless of flowrate during emptying. Looking further, we also investigate the difference in voiding and filling PV-loops for repeated tests at the exact same flowrate (0.83 cc/s). It should be noted that this plot (Fig. 6) can be viewed as analogous to a hysteresis curve. There is a clear differentiation between the pressure increases that accompany the filling of the bladder model with the pressure drop observed during voiding. The difference in work done between these tests was 27.9 mJ, further suggesting that the behavior of the bladder model may be different in filling and voiding. One explanation may be that the silicone exhibits viscoelastic behavior in response to increased strain rate (i.e., higher flowrate) [27]. We recognize that further experimentation is required to better understand this behavior.

Particle image velocimetry is an optical technique providing high spatio-temporal resolution that is often considered a reference standard validation method for fluid flow and velocity benchmarking. Because it is a benchtop experimental technique, it is coupled with in vitro models that aim to yield detailed control of flow conditions which is difficult to achieve in the body. PIV has been extensively used to validate MRI-velocimetry employed in many territories of the body to study hemodynamics in the cardiovascular system and airway tracts [14,15,28]. To our knowledge, this is the first study to of PIV applied to a system aimed at reproducing bladder flow characteristics. PIV-derived velocity maps successfully capture velocity on a 2D plane spanning the entire bladder model cross section during both voiding and filling. These velocities can be tracked even as the bladder deforms using adaptive masking which helps to maintain the fidelity and continuity of the velocity estimations even though the bladder model undergoes large deformations and begins to fold in on itself. In voiding, smooth velocity streamlines indicate limited to no swirling flow regions (vorticity, helicity) suggesting there is limited disturbance in fluid flow in the bladder model. This is in contrast to the velocity maps obtained for the filling event, where chaotic and vortex flow is observed. This is an interesting finding as we expect the behavior observed in filling here would be similar to the behavior one might see during a typical UDS study where the bladder is filled via urethral catheter instead of the ureters. As a result, we hypothesize that filling in this way may dictate bladder response during a UDS study, and a future aim of this work is to explore replicating these experiments to assess the difference between filling from the urethra compared to the ureters.

We understand that limitations to the current work presented exist. First and foremost, the in vitro modeling experimental pipeline does not represent the physiology associated with true bladder urodynamics. Currently, the in vitro bladder model deformation is largely governed by its shape. It tends to fold in on itself as the silicone, while distensible and compliant, cannot be expanded or filled much past its anatomical shape and volume. The human bladder behaves in a much more complex way, where the detrusor (bladder wall) muscle contracts during voiding and relaxes during filling [29,30]. Additionally, the urethra plays an inhibiting role during the voiding process, where the bladder must overcome urethral resistance to properly urinate, likely rendering both flow and pressure curves different from what is presented in this study. Further limitations related to the requirement of manual layup of the core and mold led to difficulties in maintaining exactly 4 mm wall thickness everywhere in the final model. Differences in wall thickness potentially led to changes in how the model wall deforms during filling and voiding because the deformation is largely determined by the design shape. However, because one of the main aims of this study is to determine the overall volume change from MRI and not to assess how the bladder changes, we believe the influence of the thickness on the results of this study is relatively minor.

During both filling and voiding, the human bladder has been shown to expand and contract with the aim to maximize efficiency and maintaining shape when receiving or delivering urine [19]. These conditions are difficult to reproduce in an in vitro setup and future works aim to employ a pressurized system where the conditions surrounding the in vitro bladder model can be better matched to the environment of the human bladder, such as mimicking abdominal effects. Presently, all pressure measurements reflect a relative pressure difference between the pressure at the initiation of a bladder event and subsequent pressure measurements where each event begins with zero (relative) pressure. As a result, we do not measure a simulated detrusor or abdominal pressure, only the pressure inside the bladder model. This means we cannot make comparisons to physiological bladder pressures. The experimental decision to only acquire and assess the relative pressure changes due to flow during each bladder event was chosen because the absolute pressure in this study will be variable depending on experimental setup and acquisition conditions, such as the model starting volume, slight changes in pressure transducer position, or sensor drift. Additionally, while the pressure acquisition system employed in this study is also invasive, which is a limitation of current urodynamic assessment, we hope our experimental pressure setup may set the basis for future work that may validate pressure that can be obtained noninvasively. A few different approaches could be carried out to accomplish this including MR-based velocimetry (2D or 3D-Phase Contrast MRI). From the MRI-derived velocity field, we can estimate a relative pressure difference map by iteratively solving modified Navier–Stokes equations [31]. The volumes we obtain using our in vitro model can also be used to drive computational fluid dynamics to determine pressures [13]. Both the aforementioned approaches are aims of future work of this study.

Future work also aims to expand upon the velocity quantification derived from the PIV experiments presented in this study by performing additional experimentation at different flow rates and extending the 2D quantification to 3D. We recognize that 2D PIV is suboptimal to characterize bladder flow, especially during filling, and will be susceptible to particles leaving the laser illumination plane during the acquisition. The specific effect of this on the results presented here can be addressed by employing stereoscopic or tomographic (both 3D) PIV. Resolving the out-of-plane components with 3D-PIV could have important implications for study the effect of different diseases of the lower urinary tract affecting bladder flow, such as BPH, prostatic protrusion, and involuntary bladder micromotions [32,33]. Specifically for the case of prostatic protrusion, we hypothesize that the concave regions developed near the base of the bladder may create regions with a high propensity of swirling or static flow, altering the flow characteristic observed in the bladder and exacerbating symptoms of bladder dysfunction. Developing a method to quantify and benchmark flow for these anatomies may highlight how the energetics of urine flow in the bladder affect function.

Similar in vitro based experimental pipelines have been used to benchmark MRI performance applied to the study of various diseases of the body. Our lab has employed similar silicone-based phantoms to validate hemodynamics derived from 4D flow MRI in the case of cerebral aneurysms, aortic coarctation, and portal venous flow distributions, among others [16,34,35]. Gonzalez-Pereria et al. first explored the use of the DISCO Flex to generate 3D bladder anatomies from MRI data, analyzing how bladder volume, shape, and orientation may change during voiding in vivo [19]. Using quantitative 3D dynamic MRI, anatomical changes in the bladder in relation to the entire lower urinary tract can be achieved. 3D MRI employed to image the bladder is a new and exciting application, however, the need for robust and systematic validation is necessary to determine its accuracy and reliability. This modeling pipeline represents a novel technique to systematically assess the validity of experimental MRI in the bladder.

Conclusion

A distensible and anatomically realistic in vitro model of the bladder was used to develop an experimental pipeline that can benchmark 3D dynamic MRI applied to the bladder. This method validates 3D Dynamic MRI by assessing potential errors in MRI-derived volume and flowrate when compared to a known reference. Coupling MRI-derived volumes with simultaneous pressure measurements can enhance MRI quantification by constructing PV-loops to analyze work. High-resolution particle image velocimetry for the analysis of flow inside the bladder model was demonstrated. Future efforts will focus on benchmarking multiple different 3D dynamic MRI sequences and quantifying velocities 3D.

Acknowledgment

We gratefully acknowledge funding by the National Institute of Diabetes and Digestive and Kidney Diseases: R01 DK126850-01 and the National Institute of Diabetes and Digestive and Kidney Diseases Complications Consortium (Diacomp): DK 076169, and GE Healthcare for their continued assistance and support.

Funding Data

  • National Institute of Diabetes and Digestive and Kidney Diseases (Grant No. R01 DK126850-01; Funder ID: 10.13039/100000062).

Data Availability Statement

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.

Nomenclature

Abbreviations
bSSFP =

balanced steady-state free procession

BPH =

benign prostatic hyperplasia

CC =

cubic centimeter

CFD =

computational fluid dyanmics

DIC =

digital image correlation

DISCO =

differential subsampling with Cartesian ordering

LUT =

lower urinary tract

mJ =

milijoule

PIV =

particle image velocimetry

PMMA =

polymethyl methacrylate

PVA =

polyvinyl alcohol

RMSE =

root-mean-squared error

R.I. =

refractive index

STL =

stereolithography

UDS =

urodynamic studies

3D =

three-dimensional

References

1.
Sarma
,
A. V.
, and
Wei
,
J. T.
,
2012
, “
Benign Prostatic Hyperplasia and Lower Urinary Tract Symptoms
,”
N. Engl. J. Med.
,
367
(
3
), pp.
248
257
.10.1056/NEJMcp1106637
2.
Wei
,
J. T.
,
Calhoun
,
E.
, and
Jacobsen
,
S. J.
,
2005
, “
Urologic Diseases in America Project: Benign Prostatic Hyperplasia
,”
J. Urol.
,
173
(
4
), pp.
1256
1261
.10.1097/01.ju.0000155709.37840.fe
3.
Güzel
,
Ö.
,
Aslan
,
Y.
,
Balcı
,
M.
,
Tuncel
,
A.
,
Keten
,
T.
,
Erkan
,
A.
, and
Atan
,
A.
,
2015
, “
Can Bladder Wall Thickness Measurement Be Used for Detecting Bladder Outlet Obstruction?
,”
Urology
,
86
(
3
), pp.
439
444
.10.1016/j.urology.2015.06.023
4.
Maserejian
,
N. N.
,
Chen
,
S.
,
Chiu
,
G. R.
,
Wager
,
C. G.
,
Kupelian
,
V.
,
Araujo
,
A. B.
, and
McKinlay
,
J. B.
,
2013
, “
Incidence of Lower Urinary Tract Symptoms in a Population-Based Study of Men and Women
,”
Urology
,
82
(
3
), pp.
560
564
.10.1016/j.urology.2013.05.009
5.
Blaivas
,
J.
,
1984
, “
Multichannel Urodynamic Studies
,”
Urology
,
23
(
5
), pp.
421
438
.10.1016/S0090-4295(84)80005-9
6.
Yao
,
M.
, and
Simoes
,
A.
,
2023
, “
Urodynamic Testing and Interpretation
,”
StatPearls
,
StatPearls Publishing
,
Treasure Island, FL
.
7.
Bushman
,
W.
,
2009
, “
Etiology, Epidemiology, and Natural History of Benign Prostatic Hyperplasia
,”
Urol. Clin. North. Am.
,
36
(
4
), pp.
403
415
.10.1016/j.ucl.2009.07.003
8.
Venkatesan
,
M.
,
Nasto
,
L.
,
Tsegaye
,
M.
, and
Grevitt
,
M.
,
2019
, “
Bladder Scans and Postvoid Residual Volume Measurement Improve Diagnostic Accuracy of Cauda Equina Syndrome
,”
Spine
,
44
(
18
), pp.
1303
1308
.10.1097/BRS.0000000000003152
9.
Dicuio
,
M.
,
Pomara
,
G.
,
Menchini Fabris
,
F.
,
Ales
,
V.
,
Dahlstrand
,
C.
, and
Morelli
,
G.
,
2005
, “
Measurements of Urinary Bladder Volume: Comparison of Five Ultrasound Calculation Methods in Volunteers
,”
Arch. Ital. Urol. Androl.
,
77
(
1
), pp.
60
62
.https://pubmed.ncbi.nlm.nih.gov/15906795/
10.
Diaz
,
T. A.
,
Benson
,
B.
,
Clinkenbeard
,
A.
,
Long
,
J. R.
,
Kawashima
,
A.
, and
Yano
,
M.
,
2022
, “
MRI Evaluation of Patients Before and After Interventions for Benign Prostatic Hyperplasia: An Update
,”
Am. J. Roentgenol.
,
218
(
1
), pp.
88
99
.10.2214/AJR.21.26278
11.
Hocaoglu
,
Y.
,
Roosen
,
A.
,
Herrmann
,
K.
,
Tritschler
,
S.
,
Stief
,
C.
, and
Bauer
,
R. M.
,
2012
, “
Real-Time Magnetic Resonance Imaging (MRI): Anatomical Changes During Physiological Voiding in Men
,”
BJU Int.
,
109
(
2
), pp.
234
239
.10.1111/j.1464-410X.2011.10255.x
12.
Zhang
,
B.
,
Liu
,
S.
,
Liu
,
Y.
,
Wu
,
B.
,
Zhang
,
X.
,
Wang
,
X.
,
Liang
,
X.
,
Cao
,
X.
,
Wang
,
D.
, and
Wu
,
C.-L.
,
2021
, “
Novel CFD Modeling Approaches to Assessing Urine Flow in Prostatic Urethra After Transurethral Surgery
,”
Sci. Rep.
,
11
(
1
), p.
663
.10.1038/s41598-020-79505-6
13.
Pewowaruk
,
R.
,
Rutkowski
,
D.
,
Hernando
,
D.
,
Kumapayi
,
B. B.
,
Bushman
,
W.
, and
Roldán-Alzate
,
A.
,
2020
, “
A Pilot Study of Bladder Voiding With Real-Time MRI and Computational Fluid Dynamics
,”
PloS One
,
15
(
11
), p.
e0238404
.10.1371/journal.pone.0238404
14.
Rutkowski
,
D. R.
,
Medero
,
R.
,
Ruesink
,
T. A.
, and
Roldán-Alzate
,
A.
,
2019
, “
Modeling Physiological Flow in Fontan Models With Four-Dimensional Flow Magnetic Resonance Imaging, Particle Image Velocimetry, and Arterial Spin Labeling
,”
ASME J. Biomech. Eng.
,
141
(
12
), pp.
121004
.10.1115/1.4045110
15.
Roloff
,
C.
,
Stucht
,
D.
,
Beuing
,
O.
, and
Berg
,
P.
,
2019
, “
Comparison of Intracranial Aneurysm Flow Quantification Techniques: Standard PIV versus Stereoscopic PIV versus Tomographic PIV versus Phase-Contrast MRI versus CFD
,”
J. NeuroInterv. Surg.
,
11
(
3
), pp.
275
282
.10.1136/neurintsurg-2018-013921
16.
Medero
,
R.
,
Falk
,
K.
,
Rutkowski
,
D.
,
Johnson
,
K.
, and
Roldán-Alzate
,
A.
,
2020
, “
In Vitro Assessment of Flow Variability in an Intracranial Aneurysm Model Using 4D Flow MRI and Tomographic PIV
,”
Ann. Biomed. Eng.
,
48
(
10
), pp.
2484
2493
.10.1007/s10439-020-02543-8
17.
McCormick
,
N.
, and
Lord
,
J.
,
2010
, “
Digital Image Correlation
,”
Mater. Today
,
13
(
12
), pp.
52
54
.10.1016/S1369-7021(10)70235-2
18.
Corrado
,
P. A.
,
Medero
,
R.
,
Johnson
,
K. M.
,
François
,
C. J.
,
Roldán-Alzate
,
A.
, and
Wieben
,
O.
,
2021
, “
A Phantom Study Comparing Radial Trajectories for Accelerated Cardiac 4D Flow MRI Against a Particle Imaging Velocimetry Reference
,”
Magn. Reson. Med.
,
86
(
1
), pp.
363
371
.10.1002/mrm.28698
19.
Gonzalez-Pereira
,
J. P.
,
Johnson
,
C. J.
,
Wells
,
S.
,
Bushman
,
W.
, and
Roldan-Alzate
,
A.
,
2023
, “
Technical Feasibility of Uro-Dynamic MRI Study of Voiding Biomechanics: A Pilot Study
,”
Int. Urol. Nephrol.
,
56
(
3
), pp.
893
899
.10.1007/s11255-023-03823-7
20.
Johnson
,
C. J.
,
Kounga
,
M. C.
,
Gonzalez-Pereira
,
J. P.
,
Wells
,
S.
,
Bushman
,
W.
, and
Roldán-Alzate
,
A.
,
2022
, “
MRI-Based Assessment of Urethral Biomechanics During Voiding
,”
Proceedings of the 2022 ISMRM & SMRT Annual Meeting & Exhibition. International Society for Magnetic Resonance in Medicine (ISMRM)
,
London, UK
, May 7–12, Paper No. 2578.10.58530/2022/2578
21.
Blatt
,
A. H.
, and
Chan
,
L.
,
2009
, “
The Importance of Bladder Wall Thickness in the Assessment of Overactive Bladder
,”
Curr. Bladder Dysfunct. Rep.
,
4
(
4
), pp.
220
224
.10.1007/s11884-009-0031-1
22.
Griffiths
,
D.
,
1995
, “
Basics of Pressure-Flow Studies
,”
World J. Urol.
,
13
(
1
), pp.
30
33
.10.1007/BF00182663
23.
Raffel
,
M.
,
Willert
,
C. E.
,
Scarano
,
F.
,
Kähler
,
C. J.
,
Wereley
,
S. T.
, and
Kompenhans
,
J.
,
2018
, “
PIV Uncertainty and Measurement Accuracy
,”
Particle Image Velocimetry: A Practical Guide
,
M.
Raffel
,
C.E.
Willert
,
F.
Scarano
,
C.J.
Kähler
,
S.T.
Wereley
, and
J.
Kompenhans
, eds.,
Springer International Publishing
,
Cham, Switzerland
, pp.
203
241
.
24.
Roldán-Alzate
,
A.
,
Frydrychowicz
,
A.
,
Niespodzany
,
E.
,
Landgraf
,
B. R.
,
Johnson
,
K. M.
,
Wieben
,
O.
, and
Reeder
,
S. B.
,
2013
, “
In Vivo Validation of 4D Flow MRI for Assessing the Hemodynamics of Portal Hypertension
,”
J. Magn. Reson. Imag.
,
37
(
5
), pp.
1100
1108
.10.1002/jmri.23906
25.
Abrams
,
P. H.
,
Skidmore
,
R.
,
Poole
,
A. C.
, and
Follett
,
D.
,
1977
, “
The Concept and Measurement of Bladder Work
,”
Br. J. Urol.
,
49
(
2
), pp.
133
138
.10.1111/j.1464-410X.1977.tb04085.x
26.
Burkhoff
,
D.
,
2013
, “
Pressure-Volume Loops in Clinical Research: A Contemporary View
,”
J. Am. Coll. Cardiol.
,
62
(
13
), pp.
1173
1176
.10.1016/j.jacc.2013.05.049
27.
Glover
,
J. D.
,
McLaughlin
,
C. E.
,
McFarland
,
M. K.
, and
Pham
,
J. T.
,
2020
, “
Extracting Uncrosslinked Material From Low Modulus Sylgard 184 and the Effect on Mechanical Properties
,”
J. Polym. Sci.
,
58
(
2
), pp.
343
351
.10.1002/pol.20190032
28.
Kim
,
S. K.
, and
Chung
,
S. K.
,
2009
, “
Investigation on the Respiratory Airflow in Human Airway by PIV
,”
J. Vis.
,
12
(
3
), pp.
259
266
.10.1007/BF03181864
29.
Coolsaet
,
B.
, and
Blok
,
C.
,
1986
, “
Detrusor Properties Related to Prostatism
,”
Neurourol. Urodyn.
,
5
(
5
), pp.
435
447
.10.1002/nau.1930050502
30.
Andersson
,
K.-E.
,
1999
, “
Pathways for Relaxation of Detrusor Smooth Muscle
,”
Advances in Bladder Research
,
L. S.
Baskin
, and
S. W.
Hayward
eds.,
Springer US
,
Boston, MA
, pp.
241
252
.
31.
Bley
,
T. A.
,
Johnson
,
K. M.
,
François
,
C. J.
,
Reeder
,
S. B.
,
Schiebler
,
M. L.
,
R. Landgraf
,
B.
,
Consigny
,
D.
,
Grist
,
T. M.
, and
Wieben
,
O.
,
2011
, “
Noninvasive Assessment of Transstenotic Pressure Gradients in Porcine Renal Artery Stenoses by Using Vastly Undersampled Phase-Contrast MR Angiography
,”
Radiology
,
261
(
1
), pp.
266
273
.10.1148/radiol.11101175
32.
Nagle
,
A. S.
,
Cullingsworth
,
Z. E.
,
Li
,
R.
,
Carucci
,
L. R.
,
Klausner
,
A. P.
, and
Speich
,
J. E.
,
2021
, “
Bladder Wall Micromotion Measured by Non-Invasive Ultrasound: Initial Results in Women With and Without Overactive Bladder
,”
Am. J. Clin. Exp. Urol.
,
9
(
1
), pp.
44
52
.http://www.ncbi.nlm.nih.gov/pmc/articles/pmc8012835/
33.
Chia
,
S. J.
,
Heng
,
C. T.
,
Chan
,
S. P.
, and
Foo
,
K. T.
,
2003
, “
Correlation of Intravesical Prostatic Protrusion With Bladder Outlet Obstruction
,”
BJU Int.
,
91
(
4
), pp.
371
374
.10.1046/j.1464-410X.2003.04088.x
34.
Shahid
,
L.
,
Rice
,
J.
,
Berhane
,
H.
,
Rigsby
,
C.
,
Robinson
,
J.
,
Griffin
,
L.
,
Markl
,
M.
, and
Roldán-Alzate
,
A.
,
2022
, “
Enhanced 4D Flow MRI-Based CFD With Adaptive Mesh Refinement for Flow Dynamics Assessment in Coarctation of the Aorta
,”
Ann. Biomed. Eng.
,
50
(
8
), pp.
1001
1016
.10.1007/s10439-022-02980-7
35.
Rutkowski
,
D. R.
,
Medero
,
R.
,
Garcia
,
F. J.
, and
Roldán-Alzate
,
A.
,
2019
, “
MRI-Based Modeling of Spleno-Mesenteric Confluence Flow
,”
J. Biomech.
,
88
, pp.
95
103
.10.1016/j.jbiomech.2019.03.025