Prediction of three-dimensional humanoid avatars for anthropometric modeling

Abstract

A method is described herein comprising generating a predictive shape model using three dimensional representations and feature parameters of a plurality of subjects, wherein the feature parameters include age, weight, height, and percentage body fat, applying the predictive shape model to target feature parameters of a target plurality of subjects to predict three dimensional representations of the target plurality of subjects, wherein the target feature parameters include age, weight, height, and percentage body fat, using the predicted shapes of the target plurality of subjects to measure anthropometric dimensions, and using the anthropometric dimensions to determine at least one clinical state of the target plurality of subject.

Description

[0001] Docket No.: 2932719-000236-WO1 Filed October 15, 2025 UNITED STATES PATENT APPLICATION FOR: PREDICTION OF THREE-DIMENSIONAL HUMANOID AVATARS FOR ANTHROPOMETRIC MODELING RELATED APPLICATIONS This application claims the benefit of US Application No.63 / 707,460, filed October 15, 2025. TECHNICAL FIELD The disclosure herein involves anthropometric dimensions derived from a person’s manifold-regression predicted three-dimensional (3D) humanoid avatar. INCORPORATION BY REFERENCE Each patent, patent application, and / or publication mentioned in this specification is herein incorporated by reference in its entirety to the same extent as if each individual patent, patent application, and / or publication was specifically and individually indicated to be incorporated by reference. SUMMARY OF THE INVENTION In embodiments, a method is described herein comprising generating a predictive shape model using three dimensional representations and feature parameters of a plurality of subjects, wherein the feature parameters include age, weight, height, and percentage body fat, applying the predictive shape model to target feature parameters of a target plurality of subjects to predict three dimensional representations of the target plurality of subjects, wherein the target feature parameters include age, weight, height, and percentage body fat, using the predicted shapes of the target plurality of subjects to measure anthropometric dimensions, and using the anthropometric dimensions to determine at least one clinical state of the target plurality of subject. In embodiments, the generating the predictive model comprises scanning the plurality of subjects to obtain three dimensional mesh representations of each subject. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 In embodiments, the generating the predictive model comprises placing standardized fiducial points on the three dimensional mesh representations. In embodiments, the generating the predictive model comprises conforming a standardized three dimensional template to the fiducial points of each subject of the plurality of subjects. In embodiments, the generating the predictive model comprises identifying a principal components matrix for use in modeling variation amongst the conformed three dimensional templates. In embodiments, the generating the predictive model comprises generating a feature parameter matrix comprising feature parameters of the plurality of subjects. In embodiments, the generating the predictive model comprises generating a pseudoinverse of the feature parameter matrix of the plurality of subjects. In embodiments, the generating the predictive model comprises multiplying the principal components matrix by the pseudoinverse to generate a manifold matrix. In embodiments, applying the predictive shape model comprises generating a target feature matrix using the target feature parameters of the target plurality of subjects. In embodiments, applying the predictive shape model comprises multiplying the manifold matrix by the target feature matrix to produce a target principal component matrix. In embodiments, applying the predictive shape model comprises projecting the target principal component matrix into a cartesian coordinate space to predict the three dimensional representations of the target plurality of subjects. In embodiments, the feature parameters and the target feature parameters include impedance using bioimpedance analysis. In embodiments, the feature parameters and the target feature parameters include resistance using bioimpedance analysis. In embodiments, the feature parameters and the target feature parameters include reactance using bioimpedance analysis. In embodiments, the feature parameters and the target feature parameters include proteins and metabolites. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 illustrates a study plan. The first study phase involved development of manifold regression models and the second phase involved comparisons of predicted and ground-truth avatar body circumferences, volumes, and surface areas. Figure 2 shows an example of avatars generated using two different manifold regression equation inputs and an actual 3D optical scan. The evaluated man was a body builder (left panel) with a waist circumference of 112 cm (designated by arrows) and hip circumference of 119 cm as measured with a 3D optical scan; his body fat was 19% and body mass index 35.5 kg / m2. Manifold predicted waist and hip circumferences (middle panel) were 117 cm and 119 cm, respectively, with age, weight, and height as covariates. Manifold-predicted waist and hip circumferences (right panel) were 111 cm and 117 cm, respectively, with %fat added to the age, weight, and height covariates. Adding %fat to the model including age, weight, and height as covariates brought visual appearance closer to actual appearance and improved waist circumference prediction to within 1 cm of that evaluated with a 3D optical scan. Figures 3A-3F show predicted versus ground-truth body circumferences for the waist (WC) (Figure 3A), hip (HC) (Figure 3C), and waist to hip ratio (left) (Figure 3E) and corresponding Bland-Altman plots (right) (Figure 3B, 3D, 3F). Regression equations and R2 values are shown in each panel of the figures and p-values are provided in Table 1. The regression line in the lefthand panels is in bold and the line of identity is dashed. The mean difference (bold) and 95% confidence intervals are shown in the righthand panels. Figures 4A-4D show predicted versus ground-truth total body volume (Figure 4A) and surface area (SA) (Figure 4C) and corresponding Bland-Altman plots (Figures 4B and 4D). Regression equations and R2 values are shown in each panel of the figures and p-values are provided in Table 2. The regression line in the lefthand panels is in bold and the line of identity is dashed. The mean difference (bold) and 95% confidence intervals are shown in the righthand panels. Figure 5 shows circumference landmarks generated by the Universal Software. Colors represent body regions demarcated by the software that also provides related volumes and surface areas. Figures 6A and 6B. Ground-truth 3D digital avatar (left) and corresponding predicted avatar for a representative man (Figure 6A) and woman (Figure 6B). BMI, body mass index; Docket No.: 2932719-000236-WO1 Filed October 15, 2025 HC, hip circumference; WC, waist circumference. Predicted and ground-truth circumference measurements are in good agreement in the man but deviations are larger in the woman, notably for the waist circumference measurement. DETAILED DESCRIPTION A method is described herein for evaluating anthropometric dimensions derived from a person’s manifold-regression predicted three-dimensional (3D) humanoid avatar. Such dimensions are accurate when compared to their actual circumference, volume, and surface area measurements acquired with a ground-truth 3D optical imaging method. Avatars predicted using this approach, if accurate with respect to anthropometric dimensions, can serve multiple purposes including patient metabolic disease risk stratification in clinical settings. Manifold regression 3D avatar prediction equations were developed on a sample of 570 adults who completed 3D optical scans, dual-energy X-ray absorptiometry (DXA), and bioimpedance analysis (BIA) evaluations. A new prospective sample of 84 adults had ground- truth measurements of 6 body circumferences, 7 volumes, and 7 surface areas with a 20-camera 3D reference scanner.3D humanoid avatars were generated on these participants with manifold regression including age, weight, height, DXA %fat, and BIA impedances as potential predictor variables. Ground-truth and predicted avatar anthropometric dimensions were quantified with the same software. Following exploratory studies, one manifold prediction model was moved forward for presentation that included age, weight, height, and % fat as covariates. Predicted and ground- truth avatars had similar visual appearances; correlations between predicted and ground-truth anthropometric estimates were all high (R2s, 0.75-0.99; all p<0.001) with non-significant mean differences except for arm circumferences (%D ~5%; p<0.05). Concordance correlation coefficients ranged from 0.80-0.99 and small but significant bias (p<0.05-0.01) was present with Bland-Altman plots in 13 of 20 total anthropometric measurements. The mean waist to hip circumference ratio predicted by manifold regression was non-significantly different from ground-truth scanner measurements. 3D avatars predicted from demographic, physical, and other accessible characteristics can produce body representations with accurate anthropometric dimensions without a 3D scanner. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Combining manifold regression algorithms into established body composition methods such as DXA, BIA, and other accessible methods provides new research and clinical opportunities. INTRODUCTION The recent introduction of low-cost three-dimensional (3D) optical imaging systems is revolutionizing anthropometric assessment for children and adults [1-4]. These digital systems, some stationary [5] and others housed in smartphones [6], can capture a person’s whole-body surface data to create a 3D humanoid avatar and estimate anthropometric measures across the whole body using the acquired information [7]. As a result of these technological advances, large 3D avatar databases are accumulating as scanning technology becomes increasingly available in research and clinical settings. One application of these archived humanoid avatars is to serve as a reference sample for developing manifold regression equations that can be used to predict a person’s physical representation in 3D from their demographic (e.g., sex, age, etc.), physical (e.g., weight, height, etc.), and other accessible characteristics (e.g., %fat, segmental impedance, body density, etc.) without requiring 3D scanning equipment [8, 9]. A method is described herein for accurately creating a 3D avatar of an individual based upon either baseline clinical characteristics or for predicting the change in their body shape based on changes in the clinical characteristics. METHODS Study Design The study design is summarized in Figure 1. The first study phase involved development of manifold regression prediction models on a sample of healthy adults. Additional information is provided in Methods on development of the manifold regression models. The second prospective phase then followed with comparison of predicted avatar anthropometric dimensions (6 circumferences, 7 volumes, and 7 surface areas) to corresponding ground-truth estimates in a new sample of healthy adults. Ground-truth anthropometric measurements were acquired with a 20-camera 3D optical scanner (SS20, Size Stream, Cary, NC). Predicted avatars were developed by manifold regression using several different exploratory combinations of demographic, physical, and other accessible characteristics as described in the Methods section. Accessible characteristics in the current study were acquired with dual-energy X-ray absorptiometry (DXA, QDR Discovery, Hologic, Marlborough, Massachusetts) and bioimpedance analysis (BIA, Docket No.: 2932719-000236-WO1 Filed October 15, 2025 InBody S10, Seoul, South Korea). The predicted and actual 3D avatars were analyzed using the same Universal Software [10, 11] developed to identify standard anatomic landmarks. Participants In the first phase of the study, manifold regression model development, participants were evaluated as part of the cross-sectional Shape Up! Adults study (NIH R01 DK109008). The Shape Up! Adults study was designed to investigate associations between body shape and composition with multiple health markers [4, 9]. In the second phase of the study, avatar anthropometric evaluation, participants were a new prospectively evaluated sample of healthy adults at or over the age of 18 years who completed the protocol measurements on the same day. The evaluations were done on the same day. These participants were recruited from the local community through web postings and print media. All participants enrolled in the study self- reported their race / ethnicity. The parent study for this project was approved by the Pennington Biomedical Research Center and University of Hawaii Cancer Center Institutional Review Boards and is posted on ClinicalTrials.gov (ID NCT03637855). The second phase of the current study was approved by the Pennington Biomedical Research Center Institutional Review Board (IRB# PBRC 2022-002). Baseline evaluations included health screening and measurement of body weight and height. Manifold Regression Model Development Statistical Shape Model After 3D optical data acquisition, each raw 3D mesh was registered to a 60,001-vertex template using the methods of Allen et al.

[0012] . Each vertex of the template had an x, y, and z coordinate and was created by a digital artist to resemble a human figure. The raw 3D meshes consisted of thousands of vertices, with the number of vertices varying randomly. Therefore, registration to the template allowed direct anatomical body shape comparisons across the sample. The outcome of this registration provided us with new 3D meshes where all the participants now have a 60,001-vertex makeup. First, fiducial points were manually placed on seventy-five anatomical locations on the raw meshes as defined in the Civilian American and European Surface Anthropometry Resource Project

[0013] by trained and validated personnel using Meshlab 1.3.2 (Consiglio Nazionale delle Ricerche, Rome, Italy). This provided a Cartesian coordinate (x, y, z) for each anatomical location on every raw mesh. Using the software Ganger, developed by Allen et al.

[0012] , the Docket No.: 2932719-000236-WO1 Filed October 15, 2025 template’s fiducial points were transformed to each target mesh’s fiducial points. The vertices of the template then warp to fit the shape of each participant’s mesh using the anatomical fiducial markers as a guide

[0014] . In principle, the algorithm minimizes the errors between the template and target mesh, resulting in the template resembling the target as the outcome. Next, a principal component (PC) transformation of the transformed meshes was performed to create sex-specific statistical shape models. These models described 99% of the body shape variance using fewer than 15 PCs. The space of human body shapes: reconstruction and parameterization from range scans. Acm T Graphic 2003; 22(3): 587-594 and Detailed 3-dimensional body shape features predict body composition, blood metabolites, and functional strength: the Shape Up! studies. Am J Clin Nutr 2019; 110(6): 1316-1326 are incorporated herein by reference in their entireties. Manifold Matrix Manifold regression analysis was performed following the creation of the shape models. The manifold equation is M = P x F+, where M is the manifold, P is the matrix of all PCs for all participants in the shape model, F is the matrix of all feature parameters (e.g., height and weight) for all participants, and+symbolizes the pseudoinverse. Once M was calculated, another matrix was created, W, which contained the target features from a person’s feature parameters (e.g., height = 150 cm and weight = 60 kg). Matrix, M, was then multiplied to matrix W (W x M) creating a new PC matrix where the target features of W have modified M. The new PC matrix was then transformed back into Cartesian space from the PC space to generate the manifold images [16, 17]. Avatar Features The manifold regression models can predict 3D humanoid avatars using demographic covariates such as age and physical characteristics including weight and height. Covariates can include race / ethnicity. Additional characteristics can be included in the equations such as %fat; impedance, resistance, and reactance values obtained from bioimpedance analysis (BIA); and proteins and metabolites obtained from blood samples. Adding more covariates usually refines predictions, especially in samples that have highly varied body shapes. In the current study, we found in exploratory evaluations that the simplest model giving good anthropometric predictions relative to ground-truth included age, weight, height, and %fat (DXA) as covariates. Since the Docket No.: 2932719-000236-WO1 Filed October 15, 2025 shape models were sex-specific, sex was not used as a covariate. This four-variable model was created by modifying F in the manifold equation. An example of the difference in predicted avatars between a model with age, weight, and height and a model that additionally included %fat is shown in Figure 2 for a young muscular adult male. The three-variable model did not distinguish people in the current study who were muscular from their counterparts with greater relative adiposity as was observed in the participant presented in the figure. Manifold regression analysis was performed in R version 4.2.1 (https: / / stat.ethz.ch / pipermail / r- announce / 2020 / 000658.html; R Core Team, 2020). Universal Software Anthropometric body dimensions were evaluated in the predicted and ground-truth avatars with Universal Software. This software operates on Matlab (Mathworks, Natick, MA) [10, 18] and runs four sequences including pre-processing, landmark detection, body partitioning, and surface area calculation. Initial scan processing repairs gaps or imperfections in the 3D mesh. Major anatomic landmarks are next detected

[0010] at the crotch, right / left armpits, shoulders, hips, and toes. The software then partitions body mass into six regions including head-neck, trunk, right / left arm, and right / left leg followed by calculation of body lengths, 6 circumferences (waist, hip, right / left mid- upper arm, right / left thigh) and 7 regional / total volumes (head / neck, torso, right / left arms and legs, whole-body), and the same 7 regional / total surface areas. The circumference sites are shown in Figure 5. Waist Circumference was measured at the lateral border of the right ilium of the pelvis. Hip Circumference was measured at the maximum circumference point within the trochanteric area. Upper Arm Circumference was measured at the midpoint between the acromion process and tip of the elbow. This point was landmarked standing behind the subject as he or she held their arms at a 90° angle with palms facing up and then measured with arms relaxed at his or her sides. Both arms were evaluated. Thigh Circumference was measured at the midpoint between the inguinal crease and the proximal boarder of the patella. This point was landmarked with the subject in a seated position with the legs positioned at a 90° angle and then measured in a standing position with a slight bend at the knee. Both legs were evaluated. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Measurements The SS203D optical reference system includes twenty structured light infrared depth sensors mounted on four vertical columns. Participants stood in the A-pose at the center of the columns and data was acquired during a 4-second scan. The acquired avatars were analyzed using Universal Software. The QDR Discovery DXA was operated with software version V8.26a:3.19 and calibrated at regular intervals according to manufacturer specifications. The National Health and Nutrition Examination Survey scanner option was turned off. Two components were evaluated, total body fat and fat-free mass; percentage (%) fat mass was derived as (fat mass / body mass) x 100. The InBody S10 BIA system used in exploratory studies has touch‐type electrodes that are attached between the heel and ankle bone of the participants’ feet and on the middle finger and thumb of each hand. Impedance of the right arm, left arm, right leg, left leg, and trunk were measured at frequencies of 1, 5, 50, 250, 500, and 1000 kHz. Model exploratory evaluations were completed with data acquired at the commonly used frequency of 50 kHz. Statistical Methods Avatars created using manifold regression (predicted) were compared with the actual (ground truth) participant avatars for selected circumferences, volumes, and surface areas using linear regression analysis (R2) and with means (±SD), root-mean square errors (RMSEs), mean absolute errors (MAEs, X±SE), concordance correlation coefficients (CCCs), and Bland-Altman analyses

[0019] . The predicted and ground-truth avatar comparisons are presented separately for the circumferences and combined for the volumes and surface area evaluations. RESULTS Sample Characteristics The sample used to develop the manifold regression model consisted of 570 adults, including 258 males and 312 females (Table S1). Table S1. Characteristics of the adult sample used in developing the manifold regression equations. Modified from Wong et al1.1Wong MC, McCarthy C, Fearnbach N, Yang S, Shepherd J, Heymsfield SB. Emergence of the obesity epidemic: 6-decade visualization with humanoid avatars. Am J Clin Nutr. 2022;115:1189-1193. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Males Females (n=258) (n=312) Median [Min, Max] 42.5 [18.0, 79.0] 47.0 [18.0, 75.0] Height (cm) Mean (SD) 176 (7.59) 162 (6.95) Median [Min, Max] 176 [151, 202] 162 [144, 181] Weight (kg) Mean (SD) 87.2 (20.4) 71.9 (20.4) Median [Min, Max] 83.8 [40.6, 174] 68.6 [35.4, 153] BMI (kg / m2) Mean (SD) 28.1 (5.86) 27.2 (7.43) Median [Min, Max] 27.4 [17.0, 52.6] 26.4 [14.2, 53.1] The sample in the second study phase included 84 adults, 35 males and 49 females, with a mean age of 45 years (Table S2). Table S2. Prospective sample characteristics. Males Females (n=35) (n=49) Age (y) 43.0 ± 13.9 46.5 ± 16.9 Height (cm) 178 ± 6.2 163.2 ± 6.5 Weight (kg) 95.0 ± 22.0 65.6 ± 14.0 BMI (kg / m2) 30.1 ± 6.6 24.6 ± 5.0 Fat (%) 28.4 ± 8.2 35.2 ± 7.7 (Results are mean±SD. Abbreviation: BMI, body mass index.) Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Males had a larger body mass index than females (~30 vs.25 kg / m2) whereas females had higher percent body fat (~35 vs 28%). There were 70 White, 8 Black, and 6 Asian participants. Circumference Evaluations The results of predicted versus ground-truth avatar circumferences are shown in Table 1 as the mean±SDs, MAEs, RMSEs, CCCs, and Bland-Altman analyses. The correlations and concordance between predicted and ground-truth circumference estimates were all high with R2s ranging from 0.78 to 0.95 (all p<0.001) and CCCs ranging from 0.80 to 0.97. Lower correlations tended to be present in both arms (R2, ~0.78) that also showed small significant (~5%, p<0.05) mean differences between predicted and ground-truth circumferences. There were no other significant predicted-ground-truth mean circumference differences, with small MAEs (2.2-3.3 cm) and RMSEs (2.9-4.2 cm). The correlation between the predicted and measured waist to hip circumference ratio had an R2and CCC of 0.77 and 0.87, respectively; significant bias (p<0.01) was present with a mean bias of 0.001 cm. Figures 3A, 3C, and 3E provides plots of predicted versus ground-truth waist and hip circumferences and the waist to hip circumference ratio. Significant (p<0.05-0.01) bias observed with the Bland-Altman plots was present for the hip, arm, and thigh circumferences with respective mean biases of 0.2-2.0 cm. Examples of generated images with waist and hip circumference measurements and the waist to hip ratio are shown for a representative male and female in Figure 6A and 6B. Predicted and ground-truth circumferences observed in the male are in good agreement while the predicted Docket No.: 2932719-000236-WO1 Filed October 15, 2025 female avatar visually appears leaner than the ground-truth avatar and this leaner appearance is reflected by smaller waist (6-7%) and hip (2-3%) circumferences and a smaller waist to hip circumference ratio (3-4%). Volume and Surface Area Evaluations The results of predicted versus ground-truth avatar volumes and surface areas are shown in Table 2. Ten measurements of right leg volume on the SS20 scanner were technically inadequate and the sample in this cell is reduced accordingly. The correlations and concordance between predicted and ground-truth regional and total volume estimates were all high with R2s ranging from 0.77 to 0.99 (all p<0.001) and CCCs ranging from 0.87 to 0.99. There were no significant predicted- ground-truth mean volume differences with small MAEs (0.01-0.1 l) and RMSEs ranging from 0.01 to 0.06 l. Significant (p<0.05-0.01) bias observed with the Bland-Altman plots was present for the head, arm, trunk, and leg with respective mean differences of -0.2-0.21. The correlation between predicted and ground-truth total volume had an R2(Figure 4A) of 0.99 also a CCC of 0.99; non-significant bias was present with a mean difference of -0.5 l. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 The correlations and concordance between predicted and ground-truth regional and surface area estimates were all high with R2s ranging from 0.74 to 0.97 (all p<0.001) and CCCs ranging from 0.87 to 0.99. There were no significant predicted-ground-truth mean surface area differences with small MAEs (0.01-0.04 m2) and RMSEs ranging from 0.01 to 0.06 m2. Significant (p<0.05) bias observed with the Bland-Altman plots was present for the head, leg, and trunk with respective mean differences of 0.003-0.05 m2. The correlation between predicted and measured total surface area had an R2(Figure 4C) and CCC of 0.97 and 0.99, respectively; non-significant bias was present with a mean difference of 0.02 m2. Composite Summary Overall, differences in the mean predicted and ground-truth circumference, volume, and surface area evaluations were all non-significant, except for the two arm circumferences (∆, ~5%). All of the other measures of agreement were good-to excellent, although small significant bias was present across all three types of anthropometric measurements, primarily those of the arms and legs. DISCUSSION Advances in 3D optical imaging are providing an unprecedented opportunity to amass large databases of humanoid avatars that can be used as reference samples for developing manifold regression prediction models such as those reported in the current study. While the produced images in earlier studies appeared visually accurate [8, 9], a critical question remained: are predicted avatars also accurate with respect to actual physical dimensions? The current study was designed to examine this question by comparing circumferences, volumes, and surface areas on humanoid avatars generated by manifold regression to corresponding measurements made on avatars acquired in healthy adults with a 20-camera 3D optical scanner; identical software was used to process predicted and ground-truth avatars. Our findings answered the question affirmatively: group mean values for 6 circumferences, 7 volumes, and 7 surface areas observed in 84 adults did not differ significantly from those acquired with the optical scanner except for arm circumferences (∆, ~5%). While measures of agreement such as R2s, RMSEs, and CCCs were all strong for predicted versus actual avatars, there was significant bias detected on several of the digitally estimated anthropometric measurements. These small bias effects can potentially be reduced or even eliminated in future studies by expanding the manifold regression sample and / or adding more or different accessible features to the developed prediction equations. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Accessible features that could serve as regression model covariates, other than DXA, include multiple different or combined resistance, reactance, and phase angle whole-body and regional values at a range of frequencies acquired with BIA, %fat as measured with BIA, and body density and %fat as quantified with air-displacement plethysmography (ADP). Potential Applications The current study was prompted by earlier reports employing visual aspects of digital human avatars [8, 9, 20-23]. Our findings show that generated digital humanoid avatars can also have accurate physical dimensions that prove useful in research and clinical settings. One application is with dynamic energy balance models that include predictions of long-term weight and %fat changes with lifestyle and pharmacologic treatments [24, 25]. These dynamic models can be supplemented with visually accurate avatars that additionally provide information on baseline and follow-up body dimensions such as the waist to hip circumference ratio, a marker of metabolic and disease risks

[0026] . Three-dimensional models of human thermoregulation are now used for physiological, medical, and public health applications for which accurate anthropometric features as shown in the current study are important for accurate predictions

[0027] . Developed avatars in exploratory modeling studies can be further processed to show before-after pseudo-DXA

[0028] and whole-body skeletal images

[0029] that can have research and educational value. These pseudo-images mimic their actual counterparts and can be generated from the 3D avatar digital outputs. Another group of applications prevails in the areas of obesity and eating disorders where visualizations of humanoid avatars are now being included in patient evaluation and management studies. Horne et al.

[0022] found that seeing a “future self” in the form of a personalized avatar reinforced motivation to modify behavior and promote engagement in a weight loss program. Three-dimensional avatars are also being used to visually map body image perceptions as an objective means of revealing anorexia nervosa illness severity [20, 21, 30]. Manifold regression models such as those evaluated in the current report improve the visual and anthropometric accuracy (e.g., as for the participant presented in Figure 2) of the avatars generated in these studies. Lastly, outside of research laboratories, our avatar approach when combined with widely available non-X-ray body composition methods such as BIA and ADP can give sufficiently accurate anthropometric estimates (e.g., waist circumference) to improve the clinical diagnosis Docket No.: 2932719-000236-WO1 Filed October 15, 2025 and monitoring of patients with overweight and obesity. Neither of these methods provide patient visualizations, circumferences, or surface areas and manifold regression predictions would thus be complementary to these respective device outputs. Avatars with accurate anthropometric dimensions can also be generated on large population samples such as the U.S. National Health and Nutrition Examination Survey to yield a wealth of information useful in multiple contexts. An essential step in this process, as noted earlier involves use of a manifold regression model with larger and more diverse development samples. Further, combinations of data from methods such as BIA and ADP with those provided by a predicted or actual 3D avatar can improve estimates of body composition and link these evaluations closer to functional and clinical outcomes

[0031] . Using the method described above, one could generate a 3D model of a person’s arm or leg much as we do for the whole body; 3D printing could then follow. Under another embodiment, the method may be used to make 3D armor for warfighters based on their accurately predicted body shapes. Under yet another embodiment, one may input weight, height, age, sex, and an estimate of waist circumference to generate shirt, suit, or pants sizes.

[0002] Docket No.: 2932719-000236-WO1 Filed October 15, 2025 REFERENCES 1. Garber AK, Bennett JP, Wong MC, Tian IY, Maskarinec G, Kennedy SF et al. Cross- sectional assessment of body composition and detection of malnutrition risk in participants with low body mass index and eating disorders using 3D optical surface scans. Am J Clin Nutr 2023; 118(4): 812-821. 2. Minetto MA, Pietrobelli A, Ferraris A, Busso C, Magistrali M, Vignati C et al. Equations for smartphone prediction of adiposity and appendicular lean mass in youth soccer players. Sci Rep 2023; 13(1): 20734. 3. Bennett JP, Liu YE, Quon BK, Kelly NN, Leong LT, Wong MC et al. Three-dimensional optical body shape and features improve prediction of metabolic disease risk in a diverse sample of adults. Obesity (Silver Spring) 2022; 30(8): 1589-1598. 4. Wong MC, Bennett JP, Quon B, Leong LT, Tian IY, Liu YE et al. Accuracy and Precision of 3-dimensional Optical Imaging for Body Composition by Age, BMI, and Ethnicity. Am J Clin Nutr 2023; 118(3): 657-671. 5. Harty PS, Sieglinger B, Heymsfield SB, Shepherd JA, Bruner D, Stratton MT et al. Novel body fat estimation using machine learning and 3-dimensional optical imaging. Eur J Clin Nutr 2020; 74(5): 842-845. 6. Smith B, McCarthy C, Dechenaud ME, Wong MC, Shepherd J, Heymsfield SB. Anthropometric evaluation of a 3D scanning mobile application. Obesity (Silver Spring) 2022; 30(6): 1181-1188. 7. Majmudar MD, Chandra S, Yakkala K, Kennedy S, Agrawal A, Sippel M et al. Smartphone camera based assessment of adiposity: a validation study. NPJ Digit Med 2022; 5(1): 79. 8. Bennett J, Wong MC, McCarthy C, Fearnbach N, Queen K, Shepherd J et al. Emergence of the adolescent obesity epidemic in the United States: five-decade visualization with humanoid avatars. Int J Obes (Lond) 2022; 46(9): 1587-1590. 9. Wong MC, McCarthy C, Fearnbach N, Yang S, Shepherd J, Heymsfield SB. Emergence of the obesity epidemic: 6-decade visualization with humanoid avatars. Am J Clin Nutr 2022; 115(4): 1189-1193. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 Sobhiyeh S, Dunkel A, Dechenaud M, Mehrnezhad A, Kennedy S, Shepherd J et al. Digital anthropometric volumes: Toward the development and validation of a universal software. Med Phys 2021; 48(7): 3654-3664. Sobhiyeh S, Kennedy S, Dunkel A, Dechenaud ME, Weston JA, Shepherd J et al. Digital anthropometry for body circumference measurements: Toward the development of universal three-dimensional optical system analysis software. Obes Sci Pract 2021; 7(1): 35-44. Allen B, Curless B, Popovic Z. The space of human body shapes: reconstruction and parameterization from range scans. Acm T Graphic 2003; 22(3): 587-594. Robinette KM, Blackwell S, Daanen H, Boehmer M, Fleming S. Civilian American and European Surface Anthropometry Resource (CAESAR). Final Report, Volume 1: Summary. In. Dayton, OH: U.S. Air Force Research Laboratory, 2002. Tian IY, Ng BK, Wong MC, Kennedy S, Hwaung P, Kelly N et al. Predicting 3D body shape and body composition from conventional 2D photography. Med Phys 2020; 47(12): 6232-6245. Ng BK, Sommer MJ, Wong MC, Pagano I, Nie Y, Fan B et al. Detailed 3-dimensional body shape features predict body composition, blood metabolites, and functional strength: the Shape Up! studies. Am J Clin Nutr 2019; 110(6): 1316-1326. Bennett J, Wong MC, McCarthy C, Fearnbach N, Queen K, Shepherd J et al. Emergence of the adolescent obesity epidemic in the United States: five-decade visualization with humanoid avatars. International Journal of Obesity 2022: 1-4. Wong MC, McCarthy C, Fearnbach N, Yang S, Shepherd J, Heymsfield SB. Emergence of the obesity epidemic: six decade visualization with humanoid avatars. The American journal of clinical nutrition 2022. Sobhiyeh S, Dechenaud M, Dunkel A, LaBorde M, Kennedy S, Shepherd J et al. Hole Filling in 3D Scans for Digital Anthropometric Applications. Annu Int Conf IEEE Eng Med Biol Soc 2019; 2019: 2752-2757. Lin LIK, Hedayat A, Wu W. Statistical tools for measuring agreement, Springer: New York, 2012. Docket No.: 2932719-000236-WO1 Filed October 15, 2025 De Coster L, Sanchez-Herrero P, Aliaga C, Otaduy MA, Lopez-Moreno J, Tajadura- Jimenez A. Perceived match between own and observed models' bodies: influence of face, viewpoints, and body size. Sci Rep 2020; 10(1): 13991. Ralph-Nearman C, Arevian AC, Moseman S, Sinik M, Chappelle S, Feusner JD et al. Visual mapping of body image disturbance in anorexia nervosa reveals objective markers of illness severity. Sci Rep 2021; 11(1): 12262. Horne M, Hardy M, Murrells T, Ugail H, Hill AJ. Using Personalized Avatars as an Adjunct to an Adult Weight Loss Management Program: Randomized Controlled Feasibility Study. JMIR Form Res 2022; 6(10): e36275. Horne M, Hill A, Murells T, Ugail H, Irving, Chinnadorai R et al. Using avatars in weight management settings: A systematic review. Internet Interv 2020; 19: 100295. Thomas DM, Martin CK, Heymsfield S, Redman LM, Schoeller DA, Levine JA. A Simple Model Predicting Individual Weight Change in Humans. J Biol Dyn 2011; 5(6): 579-599. Thomas DM, Scioletti M, Heymsfield SB. Predictive Mathematical Models of Weight Loss. Curr Diab Rep 2019; 19(10): 93. Criminisi A, Sorek N, Heymsfield SB. Normalized sensitivity of multi-dimensional body composition biomarkers for risk change prediction. Sci Rep 2022; 12(1): 12375. Xu X, Rioux TP, Castellani MP. Three dimensional models of human thermoregulation: A review. J Therm Biol 2023; 112: 103491. Leong LT, Wong MC, Liu YE, Glaser Y, Quon BK, Kelly NN et al. Generative deep learning furthers the understanding of local distributions of fat and muscle on body shape and health using 3D surface scans. Commun Med (Lond) 2024; 4(1): 13. Keller M, Zuffi S, Black MJ, Pujades S. OSSO: Obtaining Skeletal Shape from Outside. Available at: https: / / download.is.tue.mpg.de / osso / OSSO.pdf. Last updated Accessed: June 3 2024. Cornelissen KK, McCarty K, Cornelissen PL, Tovee MJ. Body size estimation in women with anorexia nervosa and healthy controls using 3D avatars. Sci Rep 2017; 7(1): 15773. Heymsfield SB, Prado CM, Gonzalez MC. Skeletal muscle-focused guideline development: hierarchical model incorporating muscle form, function, and clinical outcomes. Appl Physiol Nutr Metab 2023; 48(10): 751-756.

Claims

Docket No.: 2932719-000236-WO1 Filed October 15, 2025 CLAIMS 1. A method comprising, generating a predictive shape model using three dimensional representations and feature parameters of a plurality of subjects, wherein the feature parameters include age, weight, height, and percentage body fat; applying the predictive shape model to target feature parameters of a target plurality of subjects to predict three dimensional representations of the target plurality of subjects, wherein the target feature parameters include age, weight, height, and percentage body fat; using the predicted shapes of the target plurality of subjects to measure anthropometric dimensions; using the anthropometric dimensions to determine at least one clinical state of the target plurality of subject.

2. The method of claim 1, wherein the generating the predictive model comprises scanning the plurality of subjects to obtain three dimensional mesh representations of each subject.

3. The method of claim 2, wherein the generating the predictive model comprises placing standardized fiducial points on the three dimensional mesh representations.

4. The method of claim 3, wherein the generating the predictive model comprises conforming a standardized three dimensional template to the fiducial points of each subject of the plurality of subjects.

5. The method of claim 4, wherein the generating the predictive model comprises identifying a principal components matrix for use in modeling variation amongst the conformed three dimensional templates.

6. The method of claim 5, wherein the generating the predictive model comprises generating a feature parameter matrix comprising feature parameters of the plurality of subjects.Docket No.: 2932719-000236-WO1 Filed October 15, 2025 7. The method of claim 6, wherein the generating the predictive model comprises generating a pseudoinverse of the feature parameter matrix of the plurality of subjects.

8. The method of claim 7, wherein the generating the predictive model comprises multiplying the principal components matrix by the pseudoinverse to generate a manifold matrix.

9. The method of claim 8, wherein applying the predictive shape model comprises generating a target feature matrix using the target feature parameters of the target plurality of subjects.

10. The method of claim 9, wherein applying the predictive shape model comprises multiplying the manifold matrix by the target feature matrix to produce a target principal component matrix.

11. The method of claim 10, wherein applying the predictive shape model comprises projecting the target principal component matrix into a cartesian coordinate space to predict the three dimensional representations of the target plurality of subjects.

12. The method of claim 1, wherein the feature parameters and the target feature parameters include impedance using bioimpedance analysis.

13. The method of claim 1, wherein the feature parameters and the target feature parameters include resistance using bioimpedance analysis.

14. The method of claim 1, wherein the feature parameters and the target feature parameters include reactance using bioimpedance analysis.

15. The method of claim 1, wherein the feature parameters and the target feature parameters include proteins and metabolites.

Images