Prism Labs:
Validation methods & benchmarks
Intro
1
Proprietary and Confidential
How do we validate our methods at Prism?
Prism continuously improves its estimation methods by running clinical studies.
Prism has partnered with Energy Balance & Body Composition Laboratory at the Texas Tech University.
The most recent study (Fall ‘22/Spring ‘23) had over 60 subjects scanned an iPhone for body composition and body measurements against high-cost high-value commercially available solutions.
2
Proprietary and Confidential
Validation Summary
Prism is the singular mobile-based 3D body reconstruction solution in the market that has undergone clinical validation to affirm its accuracy and precision, employing a true 3D body reconstruction method:
3
Proprietary and Confidential
Body Composition Benchmark for iPhone Body Scans
4
What is our body composition accuracy against GE DXA scans?
Absolute error�(unsigned) | iPhone Body Fat against GE DXA | Ideal Target |
Median | 3.96 % | < 5 % |
Absolute error�(unsigned) | iPhone Body Fat % | Ideal Target |
Median | 0.3 % | < 1 % |
What is the consistency of our body composition estimates when done back-to-back?
Research Publications
Data from 60 subjects across full range of BMI classes.
Proprietary and Confidential
Body Circumference Measurements Benchmark for iPhone Body Scans
5
Relative error | iPhone body circumferences | Ideal Target |
Median | 0.41 % | < 1 % |
Relative error | iPhone circumferences vs Naked Mirror | Ideal Target |
Median | 2.20 % | < 5% |
What is our body measurements accuracy against the Naked Labs Body Scanner?
What is the consistency of our body measurements when done back-to-back?
Research Publications
Data from 60 subjects across full range of BMI classes.
Proprietary and Confidential
iPhone results lead the way in terms of precision
6
Precision across different body composition methods | Fat Mass (lbs) | Lean Mass (lbs) |
Prism iPhone scanner | 0.72 | 0.72 |
*Skinfolds | 1.28 | 1.24 |
*DXA | 1.45 | 1.35 |
*Bod Pod | 3.18 | 3.19 |
*BIA | 5.13 | 5.00 |
Measurements precision for different 3D body scanners | Std. Error (inch) | Std. Error (%) |
Prism iPhone scanner | 0.0834 | 0.41 |
SizeStream | 0.1290 | 0.63 |
Styku | 0.1387 | 0.73 |
*Errors reported from previously published studies, ranging from 1.24-5.13 lbs of fat & lean mass (Ava Farley, 2020) (Adam J Zemski, 2018) (Timothy R. Ackland, 2012).
Proprietary and Confidential
Studies Summary
7
Study | Highlights |
(2024) Accuracy of body composition against DXA |
|
(2024) Precision Validation of Prism Mobile Scanning Tech |
|
(2023) Precision and Accuracy Validation of Prism Mobile Scanning Tech |
|
(2023) Precision Validation of Single Camera Non-rigid Scanner |
|
(2020) Validation of 3 Commercial Body Scanners vs Prism Gen 1 |
|
(2019) Accuracy & Precision of Prism Gen 1 scanner |
|
Back to key highlights
Proprietary and Confidential
Body Composition Methodologies
8
| Body Scanning | Bioimpedance (BIA) | Dexa (DXA) | Volume-based Methods |
Methodology | Depth and/or RGB Camera | Electrical Currents | X-Ray Imaging | Water and Air Displacement |
Safe while pregnant | | | | |
No need for fasting or dietary restrictions | | | | |
Takes less than a minute | | | | |
Consistent Body Fat (BF) Readings | | | | |
3D Visualization with Before/After | | | | |
Body Dimensions & Circumferences | | | | |
Proprietary and Confidential
Low end photo-based mobile body scanning
Mobile-based scanning solutions
Existing solutions in the market (Spren, Bodygram, SizeStream/MeThreeSixty, myBVI, Lean Screen…) only perform photo-based scans and use statistical models to provide body insights. Prism uses a 10 second full video spin of the user per scan and does a true personal 3D body reconstruction.
Typically take 2 to 4 photos of the user in different positions. Prism captures 75x more data per scan.�
Goes beyond BF into circumferential measurements.
These solutions often lacks consistency of results, which means changes can get lost in the noise.
�The 3D scans are not personally accurate since only statistical models are used - the avatar is a generic representation that is fitted to the volume of the photos.
9
Halo
Bodygram
Proprietary and Confidential
Bioelectrical impedance analysis (BIA)
Body fat (BF) accuracy: on average +/- 8 to 12 % BF deviation from DXA equivalents on BIA scales but significantly closer for higher-end InBody systems.
Highest penetration in gyms - Inbody
Relatively compact, quick, no moving parts
High end device lists for $27k, divides body into 5 sections
Withings scale is $60, one score, less accurate and consistent
Not very accurate, and very sensitive to external factors
Less accurate at higher body fat %
Hydration levels significantly impact score
Training intensity and recency impact score
10
Withings Scale
InBody System
Sources:
Proprietary and Confidential
Dexa (Dual Energy X-Ray Absorptiometry)
Gold standard for % BF, typically +/-1.5 to 2 % error.
You need a large X-Ray machine and a technician.
Takes about 20 min.
Average unit price is approximately $40,000.
A single scan can cost $125.
11
Sources:
Proprietary and Confidential
Volume-based methods
Average accuracy at +/- 2 % BF, however more often it has more oscillations than hydrostatic that can go as high as +/- 10%.�
Typically done in a medical facility with need for a technician.�
Equipment costs around $30,000 and each analysis about $50.
Average accuracy between +/- 1.8 to 2.8 % BF.
You need a pool and a technician -> Large investment.
Time-consuming and causes discomfort to user.
Hydrostatic Weighing
Air Displacement Plethysmography (ADP)
12
Proprietary and Confidential