Lower urinary tract symptoms (LUTS), which include voiding and storage symptoms, have been universally recognized since the standardization of the urinary tract function terminology by the International Continence Society in 2002 [1].   Various assessment methods are utilized, including the assessment of post-void residual volume to determine the type of LUTS. Using ultrasonography (US), bladder volume can be estimated. Some devices allow visualization of the bladder volume, while others provide an estimate without direct visualization of the bladder. The latter method does not provide an accurate assessment as it cannot discriminate between other structures, such as fluid-filled cysts and the bladder. Consequently, an ultrasound that allows for direct observation of the bladder provides the most accurate volume estimate. A common manual method of determining bladder volume from two-dimensional (2D) US images uses the prolate ellipsoid method, which requires three measurements be taken as follows:  in the transverse plane, the height (D1) and width (D2) are acquired, and in the sagittal plane, the diameter (D3) is acquired. This assessment is done manually by the US operator using onscreen calipers. The formula is L x W x H x 0.52 [2]. This process is repetitive and time-consuming due to many factors. It also requires operator training to acquire the requisite skills. Artificial intelligence (AI) can improve the accuracy of bladder volume measurement by identifying the bladder region and eliminating artifacts or noise from the image. Machine learning algorithms can be trained to recognize image patterns and classify them according to the bladder's shape and size. These algorithms or models can automatically recognize the shape and dimensions of the bladder, thus calculating the volume [3]. This can improve image quality, reduce operator variability, and increase accuracy. Since AI can reduce measurement errors and improve efficiency, this can improve diagnostic accuracy and treatment of bladder-related or LUT conditions such as urinary incontinence, urinary retention, and neurogenic bladder dysfunction. More importantly, bladder AI (BLAI) can provide real-time bladder volume measurements and benefit patients requiring frequent bladder volume monitoring. This can positively impact patient care by providing accurate and timely information to healthcare professionals, enabling them to make informed decisions about patient management [3]. The primary objective of this study was to assess the magnitude of the difference between Clarius BLAI and human bladder volume measurement to determine whether the BLAI measurement is non-inferior to human expert measurement. A secondary objective was determining if Clarius BLAI could correctly identify the imaging plane (sagittal or transverse).

This prospective non-inferior study was conducted at a university physical therapy department following institutional review board (IRB) approval. Inclusion criteria included non-pregnant individuals over 18 years who provided informed consent. Fifty-eight subjects were recruited (18 Males: 40 Females), with a mean age of 31 years (range 21 – 61) and a mean BMI of 26.88 kg/m2 (range 18 – 45). Three physical therapists (PTs) with ultrasound imaging experience performed all ultrasound scanning using a Clarius C3HD (curvilinear) and PAHD (phased array) scanner as subjects lay supine on a plinth with a pillow beneath their heads and their lower abdomen exposed to access the bladder. Sagittal and transverse views were used to determine bladder volume. Images were uploaded to the Clarius Cloud and reviewed by the three PTs, who independently and manually measured the bladder volume, indicated the bladder view (sagittal or transverse), and traced the bladder wall for each image (Fig.1 A and B). The three blinded PT reviewers measured each bladder volume. The bladder's length, width, and height were measured, and volume was calculated via the prolate ellipsoid method (L x W x H x 0.52) (Fig 1. C and D).  The absolute percent (%) difference between reviewer pairs was calculated and compared to the absolute percent difference between automatic measurement and mean reviewer measurement using a one-sided t-test and an equivalence margin of 25% (i.e., the mean difference between differences should be no greater than 25% of the measured bladder difference) which has been reported in the literature when comparing different methods of bladder volume measurement [2].  Inter-rater reliability (IRR) was determined via the Intraclass correlation coefficients (ICC) using Bland-Altman plots, while the average Dice similarity coefficient and Jaccard similarity index determined the accuracy of image segmentation measurements.

The BLAI volume measurement was found to be non-inferior (p<.001), with a mean difference between percent differences of the human measurement and BLAI means of  -0.0228 (95% CI -0.074, 0.028). Bland-Altman plots were assessed to determine agreement between the two methods of bladder volume measurement. The mean difference and tight clustering of bladder volume measurements between BLAI and human measurements indicated strong overall agreement with individual human measurements. The ICC showed strong agreement between reviewer 1 vs reviewer 2 of 0.98 ( 95% CI 0.96, 0.99), reviewer 1 vs reviewer 3 of 0.98 (95% CI 0.97,0.99), reviewer 2 vs. reviewer 3 of 0.98 (95% CI 0.97,0.99) (Table 1). Additionally, the ICC comparing BLAI to the reviewer mean was 0.91 (95% CI 0.85,0.95) (Table 1). The average Dice coefficient and Jaccard index scores showed strong agreement between the BLAI model compared to each reviewer and between each reviewer pair. Dice scores for reviewers 1, 2, and 3 compared to BLAI were 0.93 (95% CI 0.92, 0.93), 0.92 (95% CI 0.91, 0.93), and 0.93 (95% CI 0.92, 0.93), respectively. Comparison pair Dice scores for reviewer 1 vs. reviewer 2 was 0.93 (95% CI 0.91, 0.94), reviewer 1 vs reviewer 3 was 0.94 (95% CI 0.93, 0.94), and reviewer 2 vs reviewer 3 was 0.94 (95% CI 0.93, 0.95). Jaccard scores for reviewers 1, 2, and 3 compared to BLAI were 0.87 (95% CI 0.85,  0.88), 0.86 (95% CI 0.84, 0.87), and 0.86 (95% CI 0.85, 0.88) respectively. Comparison pair Jaccard scores for reviewer 1 vs. reviewer 2 was 0.87 (95% CI 0.85, 0.88), reviewer 1 vs reviewer 3 was 0.89 (95% CI 0.87, 0.90), and reviewer 2 vs reviewer 3 was 0.88  (95% CI 0.87,  0.90)(Table 2). A sub-group analysis compared bladder volumes measured with Clarius C3HD with an ICC of 0.90 (95% CI 0.81,  0.94) and Phased array with an ICC of 0.97 (95% CI 0.92,  0.99), thus demonstrating excellent reliability.

Overall, results confirm high levels of agreement and consistency across all bladder volume measurements (BLAI and human). The study's secondary objective to determine whether BLAI can correctly identify transverse and sagittal bladder views was also successfully met.

Accurately assessing bladder volume is crucial in determining a host of LUT conditions. Many health systems and outpatient settings implement point-of-care ultrasound  (POCUS), including bladder scanning, to improve patient care. Including BLAI can reduce measurement errors and improve efficiency, diagnostic accuracy, and the treatment of bladder-related conditions. More importantly, BLAI provides real-time bladder volume measurements in seconds, which can benefit patients who require frequent bladder volume monitoring, allowing for more cost-effective care. This can also positively impact patient care by providing accurate and timely information to healthcare professionals, enabling them to make informed decisions about patient management across multiple settings.

Abrams P, Cardozo L, Fall M, Griffiths D, Rosier P, Ulmsten U., et al. Standardization Subcommittee of the International Continence Society. The standardization of terminology of lower urinary tract function: report from the standardization subcommittee of the International Continence Society. Neurourol Urodyn 2002;21:167—178.
Dicuio M, Pomara G, Menchini Fabris F, Ales V, Dahlstrand C, Morelli G. Measurements of urinary bladder volume: comparison of five ultrasound calculation methods in volunteers. Arch Ital Urol Androl. 2005 Mar;77(1):60-2. PMID: 15906795.
Matsumoto M, Tsutaoka T, Yabunaka K, Handa M, Yoshida M, Nakagami G, Sanada H. Development and evaluation of automated ultrasonographic detection of bladder diameter for estimation of bladder urine volume. PLoS One. 2019 Sep 5;14(9):e0219916. doi: 10.1371/journal.pone.0219916. PMID: 31487299; PMCID: PMC6728037.