Applying a portable pocket handheld ultrasound scanner to monitor tumor volume in preclinical studies

Abstract

Background:

Calipers remain the most common method for monitoring tumor volume in preclinical studies, despite known geometric limitations that can lead to substantial measurement error, particularly for irregularly shaped tumors. Portable handheld ultrasound (US) devices have recently become more accessible and may provide a practical alternative; however, their quantitative performance in routine small-animal workflows has not been fully characterized.

Purpose:

To evaluate whether a low-cost handheld US probe is suitable for routine preclinical tumor-volume measurement and to quantify its accuracy relative to calipers using cone-beam computed tomography (CBCT)–derived volumes as the reference standard.

Methods:

Five BALB/c and five C57BL/6J mice (n = 10) bearing subcutaneous CT26 (colorectal) or LLC1 (lung) tumors were measured on day 14 post-inoculation using digital calipers and a handheld US probe. Tumor volumes were estimated using caliper-derived length × width (L×W_Cal), US-derived length × width (L×W_US), US-derived length × width × depth (L×W×D_US), and an US-based region-of-interest method (ROI_US). Excised tumors were imaged with CBCT and segmented to obtain reference volumes. Volumes were normalized to CBCT and analyzed using nonparametric repeated-measures statistics (Friedman test with Holm–Bonferroni–corrected Wilcoxon signed-rank tests; significance defined as adjusted p < 0.05). Effect sizes were calculated using the rank-biserial correlation coefficient.

Results:

CBCT tumor volumes ranged from 30–170 mm³ (82.8 ± 49.6 mm³). Caliper-derived L×W_Cal volumes significantly overestimated CBCT (median 196% of reference; adjusted p = 0.0078; large effect size, r_rb = 1.0). US-based methods demonstrated substantially reduced bias and variability. L×W_US, L×W×D_US, and ROI_US volumes were not significantly different from CBCT after correction, with ROI_US showing the lowest median absolute error (9.6%).

Conclusions:

In this preclinical cohort, handheld US reduced systematic volumetric overestimation relative to calipers and demonstrated improved agreement with CBCT-derived reference volumes. ROI-based US analysis yielded the lowest absolute error. These findings support the use of portable US as a more accurate alternative to calipers for tumor-volume monitoring in small-animal studies.

1 |. INTRODUCTION

Longitudinal monitoring of tumor volume is essential in preclinical biomedical research.^1,2 The gold standard for determining tumor burden is histological analysis, which generally requires euthanasia of animal subjects.^3,4 The best approximation of tumor volume for longitudinal studies is obtained from computed tomography (CT).^3–5 While accurate, CT scans of animals are time-consuming and require a substantial amount of training for a researcher to conduct independently.

Due to these practical limitations, subcutaneous tumor volume in preclinical studies is most commonly estimated using calipers to measure external linear dimensions, which are then converted to volume using ellipsoid-based geometric approximations.^1,2,6–8 While straightforward and widely accessible, caliper-based measurements often produce substantial estimation errors because they rely on simplified geometric assumptions and cannot capture irregular tumor shapes or internal depth.^7–12 These inaccuracies can be particularly problematic in oncological studies, where precise measurements of tumor growth or regression is critical for evaluating therapeutic response^,13,14. Ultrasonography has also been explored as an alternative method for measuring tumor volume in small animals. Compared with calipers, ultrasound (US) can provide cross-sectional imaging of tumors and potentially improve volumetric estimation. Recent advances in US technology have also enabled the development of compact portable imaging devices.¹⁵ However, most US systems used in preclinical research are large console-based machines and typically require anesthesia and specialized training to operate.^16,17 As a result, their routine use for longitudinal tumor monitoring remains limited. Only a small number of studies have evaluated handheld US devices for tumor measurement in mice, and quantitative validation of these approaches against CT-derived volumetric reference measurements remains limited.¹⁸

In this study, we investigated the application of inexpensive, portable US probes as a practical alternative to traditional caliper-based tumor measurements. Using cone-beam CT (CBCT)–derived tumor volumes as a reference standard, we systematically compared multiple US-based volume estimation approaches with conventional caliper measurements. Our goal was to evaluate whether portable US can provide more accurate tumor volume estimation while maintaining the practicality required for routine preclinical use.

2 |. MATERIAL AND METHODS

2.1 |. Animal models

All animal experiments and procedures were approved by the Institutional Animal Care and Use Committee of Stanford University (APLAC 27939) and were performed in accordance with institutional guidelines, the United States (U.S.) National Research Council’s Guide for the Care and Use of Laboratory Animals, and the U.S. Public Health Service’s Policy on Humane Care and Use of Laboratory Animals. Six- to eight-week-old female BALB/c and C57BL/6 (Mus musculus Linnaeus, 1758) mice were obtained from The Jackson Laboratory (Sacramento, CA), and were selected to evaluate the measurement methodology across commonly used syngeneic tumor models with differing growth kinetics and morphological characteristics. The study was designed as a methodological comparison rather than a biological evaluation of strain- or sex-specific differences. Standard animal care and housing was provided by the Stanford University School of Medicine and is under the care and supervision of the Department of Comparative Medicine’s Veterinary Service Center. Mice were maintained on regular chow Envigo Teklad 2018 global diet containing 18% protein and 6% fat. Five BALB/c and five C57Bl6/J mice (total n = 10) were anaesthetized using isoflurane (3% induction; 2 and 1.5% maintenance) in 100% oxygen at a flow rate of 2 L/min, and hair was removed from the right bottom leg and back. Subsequently, 10⁵ colorectal CT26 (ATCC catalog # CRL-2638) or lung carcinoma LLC1 (ATCC catalog # CRL-1642) were injected subcutaneously at the lower back or flank in C57Bl/6J or BALB/c, respectively.

2.2 |. Caliper measurements

On day 14 post–tumor inoculation, tumor length (L) and width (W) were measured sequentially using calipers and US. Mice were anesthetized via inhalation (3% isoflurane in room air for induction and 1.5% isoflurane for maintenance). Tumors were measured first using calipers, followed by US imaging. Digital calipers were used (Fisherbrand^™ Traceable^™ Digital Carbon Fiber Calipers, catalog no. S90187A;Thermo Fisher Scientific, Waltham, MA, USA) to measure tumor L (longest tumor dimension) and W (orthogonal dimension). Each measurement was performed twice, and the mean value was used for analysis and for calculation of caliper-derived ellipsoid volume estimates (L×W_Cal).

2.3 |. US imaging

US imaging was performed using a portable handheld diagnostic US system (Clarius L20 HD3; Clarius Mobile Health, Burnaby, BC, Canada). The L20 HD3 is a high-frequency linear array transducer with a nominal operating frequency range of 8–20 MHz, designed for superficial imaging applications. Images were acquired in B-mode using the manufacturer’s proprietary real-time digital beamforming and image reconstruction software. The imaging depth was set to 1.5 cm for all acquisitions. During the same anesthesia session for each mouse, tumors were measured using a portable handheld US probe following caliper measurements. Minimal transmission gel was applied to ensure optimal acoustic coupling. Two-dimensional US images were acquired in orthogonal cross sections, from which tumor length (L) and width (W), corresponding to caliper-based measurements, were derived. Image acquisitions were saved using a mobile device (iPhone 12 mini, Apple Inc., Cupertino, CA, USA) running the Clarius US App. US-derived L and W measurements were used to calculate ellipsoid volume estimates (L×W_US).

2.4 |. CT imaging

Following completion of in vivo tumor measurements, tumors were excised surgically. Following euthanasia by cervical dislocation under isoflurane anesthesia, tumor-bearing mice were immobilized in the supine position. A circumferential skin incision (~1 cm margin) was made around the tumor, and the skin was retracted and secured to expose the tumor. The tumor mass was carefully dissected from surrounding normal tissue using a scalpel and excised for subsequent imaging.

Within approximately 20 min of collection, excised tumors were aligned together and imaged using cone-beam computed tomography (CBCT; SmART, Precision X-Ray Inc., North Branford, CT, USA). CBCT acquisition was performed using an amorphous silicon flat panel detector, and images were reconstructed using a Feldkamp-Davis-Kress (FDK) reconstruction with an isotropic voxel size of 0.2 mm. The CBCT datasets were analyzed using RT-Image open-source software (version 3.13.1), running on IDL (version 8.5.1).

2.5 |. Tumor volume analysis

Tumors were segmented on CBCT images at regularly spaced planes with an inter-slice distance of 0.2 mm in the z-direction, and volumes were subsequently interpolated to derive tumor volumes. These CBCT-derived tumor volumes were used as the gold-standard volume measurements (Figure 1A,B).

FIGURE 1.

Excised tumors and CBCT-based segmentation. (a) Photograph illustrating the orientation of all excised tumors (n = 10) prior to CBCT imaging. The photograph is presented with cropped tumor volumes using background removal. (b) Representative CBCT cross-section showing corresponding tumor masks, with segmented tumor boundaries outlined in red. These segmentations were used as the reference (“Gold Standard”) for tumor volume normalization and comparison across measurement methods. Tumor labels are defined as follows: C1 denotes the colorectal carcinoma cohort (BALB/c mice), C2 denotes the lung carcinoma cohort (C57BL/6J mice), and M1–M5 indicate individual mice within each cohort.

Tumor volumes estimated from caliper and US measurements using L and W, were calculated using the conventional ellipsoid approximation formula given in Equation (1),

$$V=\frac{1}{2}(L\times W^2)$$ (1)

and are referred to as L×W_Cal and L×W_US volumes, respectively.

For US measurements, where tumor depth (D) was additionally obtained from the width cross-section in the orthogonal view (Figure 2A,B), US-derived tumor volumes were calculated using the traditional ellipsoid formula shown in Equation (2),

$$V=\frac{4}{3}\pi \left(\frac{L}{2}\right)\left(\frac{W}{2}\right)\left(\frac{D}{2}\right)$$ (2)

and are referred to as L×W×D_US volumes. Lastly, US datasets were analyzed using RT-Image open-source software, and regions of interest (ROIs) were selected from two orthogonal cross-sectional images (Figure 2C,D). An algorithm was developed that used these US-derived ROIs to compute tumor volumes, referred to as ROI_US volumes (Figure 3).

FIGURE 2.

Representative US-based tumor measurements and region-of-interest (ROI) delineation (sample C1M3, colorectal carcinoma cohort, mouse 3). (a–b) Two-dimensional B-mode US images of a representative tumor acquired in orthogonal orientations. Panel (a) shows the longitudinal plane used to measure tumor length (L). Panel (b) shows the orthogonal cross-section used to measure tumor width (W) and depth (D). Dashed arrows indicate linear measurements. (c–d) The same US images shown in (a) and (b), with manual delineation of tumor regions of interest (ROIs) in the corresponding orthogonal planes, used for ROI-based volume estimation. The orange scale bar represents 0.5 cm in the lateral (horizontal) direction. Depth markers (white dots along the right margin) indicate 0.2 cm increments. All images were acquired with a 1.5 cm depth setting.

FIGURE 3.

Schematic illustration of the US-based ROI volume estimation algorithm. Tumor volume is estimated from two orthogonal cross-sectional areas (A₁ and A₂) obtained from US images. Each cross-section is discretized into N segments along its principal axis (l₁ or l₂), and the corresponding transverse lengths are used to approximate volume by numerical summation. Two independent volume estimates (V and V′) are obtained by interchanging the principal axes, and the final tumor volume is defined as the average of V and V′ to reduce orientation-dependent bias.

2.6 |. Statistical methods

Tumor volumes derived from caliper and US measurements (L×W_Cal, L×W_US, L×W×D_US, and ROI_US) were normalized to the corresponding CBCT–derived tumor volume for each mouse, which served as the gold-standard reference. Normalized volumes are reported as a percentage of the CBCT volume, and summary statistics are expressed as mean ± standard deviation (SD).

Statistical analyses were performed for colorectal tumors (samples 1–5), lung tumors (samples 6–10), and the combined dataset (n = 10). Because each tumor was evaluated using all volume estimation methods, a repeated-measures design was employed. Overall differences among measurement methods were assessed using the Friedman test. When significant, post hoc pairwise comparisons between each method and CBCT were conducted using two-sided Wilcoxon signed-rank tests. To control the family-wise error rate associated with multiple comparisons, p-values were adjusted using the Holm–Bonferroni method, with statistical significance defined as an adjusted p < 0.05.

To quantify the magnitude and direction of differences relative to CBCT, effect sizes were calculated using the rank-biserial correlation coefficient (r_rb). For paired Wilcoxon signed-rank tests, r_rb was computed according to Equation (3),

$$r_{\mathit{\text{rb}}}=\frac{W^{+}-W^{-}}{W^{+}+W^{-}}$$ (3)

where W⁺ and W⁻ represent the sums of positive and negative signed ranks, respectively. Effect sizes were interpreted using conventional thresholds (|r_rb| ≈ 0.1, small; ≈ 0.3, moderate; ≥ 0.5, large).

Statistical analyses and figure generation were performed in Python (v3.11.2) using NumPy (v1.24.0), SciPy (v1.14.1), pandas (v2.2.3), and Matplotlib (v3.7.5).

2.7 |. AI use

An AI language model (ChatGPT, OpenAI) was used solely for editorial assistance (grammar, wording, and structure). All scientific content, analysis, and conclusions were produced by the authors.

3 |. RESULTS

As shown in Figure 4A, caliper-derived volume estimates based on length and width (L×W_Cal) consistently overestimated tumor volume relative to CBCT. In contrast, US-based methods produced volume estimates that clustered closer to the CBCT reference, with reduced variability. A Friedman test demonstrated a significant overall difference among volume estimation methods (χ² = 21.36, p < 0.001).

FIGURE 4.

Comparison of tumor volume estimation methods relative to CBCT (n = 10). (a) Tumor volumes normalized to the CBCT-derived reference volume (100%) for the pooled combined cohort (n = 10). Violin plots represent the distribution of normalized volumes for each method (L×W_Cal, L×W_US, L×W×D_US, ROI_US), with individual tumor measurements overlaid as points. The dashed horizontal line denotes the CBCT reference (100%). The central line within each violin indicates the median. An asterisk (*) denotes a statistically significant difference relative to CBCT after Holm–Bonferroni correction (paired Wilcoxon signed-rank test, adjusted p < 0.05). (b) Absolute percent error of tumor volume estimates relative to CBCT. Boxplots display the interquartile range (25th–75th percentile) with the median shown as a horizontal line. Whiskers extend to the most extreme values within 1.5 × interquartile range (IQR), and circles represent statistical outliers beyond this range. Panel B is descriptive and intended to illustrate differences in accuracy and variability across methods; no additional statistical testing is indicated.

Post hoc paired comparisons confirmed that L×W_Cal volumes were significantly larger than CBCT-derived volumes after Holm–Bonferroni correction (adjusted p = 0.0078), with a large effect size (rank-biserial correlation r_rb = 1.0), indicating systematic overestimation (Table 1). The median L×W_Cal volume was 196% of the CBCT reference, corresponding to a median absolute error of 96%.

TABLE 1.

Statistical comparison of caliper- and US-based tumor volume estimation methods relative to cone-beam computed tomography (CBCT) for the combined cohort (n = 10). Volumes are expressed as a percentage of the CBCT-derived reference volume. Holm-corrected p-values are from paired Wilcoxon signed-rank tests. Effect sizes are reported as rank-biserial correlation coefficients (r_rb).

Calculation method	Median volume (% of CBCT)	Median absolute error (%)	Holm-corrected p-value vs CBCT	Effect size (rrb)	Interpretation
L×W_Cal	196	96	0.0078*	1	Systematic overestimation
L×W_US	99.9	26.2	1	0.02	No bias versus CBCT
L×W×D_US	83.9	20.2	0.316	−0.60	Moderate underestimation
ROI_US	94.6	9.6	0.316	−0.56	Closest agreement with CBCT

^*Statistically significant after Holm–Bonferroni correction (adjusted p < 0.05), relative to CBCT.

US-based methods showed substantially improved agreement with CBCT. Volumes estimated using US-derived length and width (L×W_US) exhibited minimal median bias (99.9% of CBCT) but greater variability, and did not differ significantly from CBCT after multiple-comparison correction (adjusted p = 1.00; r_rb = 0.02). Incorporation of US-measured depth (L×W×D_US) reduced variability and bias relative to calipers, though a trend toward underestimation was observed (median 83.9% of CBCT), without statistical significance after correction (adjusted p = 0.316; r_rb = −0.60). The ROI-based US approach (ROI_US) demonstrated the closest overall agreement with CBCT, with a median volume of 94.6% of the reference and the lowest median absolute error (9.6%), but likewise did not differ significantly from CBCT after correction (adjusted p = 0.316; r_rb = −0.56).

Analysis of absolute percent error (Figure 4B) further highlighted the reduced error and variability of US-based methods compared with calipers, with ROI_US exhibiting the narrowest error distribution. Subgroup analyses performed separately for colorectal (n = 5) and lung (n = 5) tumors showed similar directional trends across methods, although statistical significance was not reached within individual subgroups due to limited sample size.

4 |. DISCUSSION

The present study quantitatively demonstrates that portable handheld US substantially reduces volumetric error compared with caliper-based measurements when evaluated against CBCT-derived tumor volumes. As shown in Figure 4A and summarized in Table 1, caliper-derived L×W_Cal volumes overestimated CBCT measurements by a median of 96% (196% of reference volume), corresponding to a large effect size (r_rb = 1.0) and a statistically significant difference after Holm–Bonferroni correction (adjusted p = 0.0078). This magnitude of deviation indicates a substantial and practically meaningful bias.

In contrast, US-based methods showed markedly smaller deviations from CBCT. ROI-based US demonstrated the lowest median absolute error (9.6%), a median volume of 94.6% relative to CBCT, and a moderate effect size (r_rb = −0.56) (Table 1; Figure 4B). Although this comparison did not reach statistical significance after correction (adjusted p = 0.316), the moderate effect size indicates a substantially smaller deviation than that observed with calipers, suggesting a practically meaningful reduction in systematic overestimation. Similarly, the L×W_US method demonstrated minimal median bias (99.9% of CBCT) and a negligible effect size (r_rb = 0.02), reflecting very small practical deviation from the reference standard.

The superiority of US in this context is consistent with known limitations of caliper-based tumor measurement. Calipers have repeatedly been reported to overestimate tumor volume in small animal models due to inclusion of skin, fur, and surrounding soft tissue,^1,2,6–11 as well as variability introduced by operator handling.^12–14 Furthermore, the conventional ellipsoid formula assumes symmetry and cannot account for irregular tumor geometry. In the present study, this geometric limitation is reflected quantitatively in the large effect size observed for caliper measurements (Table 1). US overcomes these constraints by directly measuring depth and visualizing cross-sectional morphology, enabling more representative volume estimation.²

Although US acquisition required approximately three times longer than caliper measurements, the observed reduction in volumetric error, from 96% median absolute error with calipers to 9.6% with ROI-based US (Figure 4B), suggests that the additional acquisition time yields substantial gains in measurement accuracy. Most of the additional time was attributable to probe handling and smartphone interface operation. With increased user experience or minor ergonomic refinements (e.g., a 3D-printed probe holder), acquisition time could likely be reduced without compromising accuracy.

Importantly, we do not interpret statistically non-significant differences as proof of equivalence. Rather, we report both statistical significance and effect size magnitude to characterize the direction and practical importance of deviations relative to CBCT. In this small cohort (n = 10), non-significant findings accompanied by small or moderate effect sizes suggest limited detectable deviation rather than confirmed equivalence. Larger sample sizes would permit more precise estimation of deviation magnitude and narrower confidence intervals.

Beyond the direct comparison with calipers, this study highlights the practical implications of having a compact imaging device readily available within preclinical laboratories. While ultrasonography is already established in clinical diagnostic settings, including oral and maxillofacial tumor assessment^19,20 and emergency and cardiovascular applications,^21–24 handheld US probes are increasingly being explored across additional clinical applications.^25,26 However, their integration into routine preclinical tumor monitoring has remained limited due to accessibility and operational complexity. The present findings demonstrate that a handheld device can provide quantitatively improved volumetric accuracy compared with calipers, without requiring access to dedicated imaging cores.

The improved volumetric precision observed here suggests potential utility in several preclinical contexts. For example, more accurate early-stage tumor measurement may improve response monitoring in studies with small sample sizes or tight effect margins. Enhanced depth visualization may also reduce misclassification of irregular tumors or partial regressions. Although these broader applications were not directly evaluated in this study, the substantial reduction in volumetric bias observed relative to calipers (Figure 4, Table 1) supports the feasibility of expanding handheld US use in preclinical tumor research.

Overall, the contribution of this work lies not in claiming equivalence to CBCT, but in quantitatively demonstrating a substantial reduction in systematic volumetric bias relative to caliper-based measurement. The large effect size observed for calipers and the markedly smaller deviations observed for US-based methods provide objective evidence that portable US offers a more accurate and practically meaningful alternative for preclinical tumor volume monitoring.

5 |. CONCLUSIONS

In this study, portable handheld US reduced volumetric error compared with caliper-based tumor measurements when evaluated against CBCT-derived reference volumes. Caliper measurements demonstrated substantial overestimation (large effect size, adjusted p < 0.05), whereas US-based methods showed markedly smaller deviations and lower absolute errors. These findings indicate that handheld US provides more accurate tumor-volume estimation than calipers within the tested preclinical models. While acquisition time was longer than caliper measurement, the observed reduction in systematic bias supports the practical utility of handheld US as an alternative method for preclinical tumor monitoring. Future studies with larger cohorts may further define its performance across additional tumor morphologies and growth stages.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.