BitDoctor.ai
  • Abstract
  • Introduction
  • Problem Statement
  • Market Insights
  • Preventive Healthcare
  • Unique Value Proposition
  • BitDoctor Blockchain Technology Stack
  • Our AI Technology
    • Training BitDoctor's AI
    • Hardware And Imaging Requirement
  • Clinical Measurement Reports
    • Heart Rate
    • Breathing Rate
    • Irregular Heartbeat
    • Heart Rate Variability
    • Hypertension Risk
    • Type 2 Diabetes Risk
    • Cardiovascular Diseases Risk (incl. Heart Attack & Stroke Risks)
    • Hypercholesterolemia
    • Hypertriglyceridemia
    • Fatty Liver Disease
    • Morning Fasting Blood Glucose
    • Hemoglobin A1C
    • Image-Based Age
  • DePIN Shared Economy
  • Strategic Opportunities
    • Clinical Trial Agencies
    • Preventive Healthcare Brands
    • Active Wear & Equipment
    • Insurance Company
    • Pharmaceutical Company
    • Supplement Company
    • Crypto Firms
  • Tokenomics
    • Token Utility
  • Community
    • About $LiV Points
  • Roadmap
  • Links
  • Appendix
  • Team Info
    • Advisors
Powered by GitBook
On this page
  • Description
  • Participants
  • Additional Data Collection Procedure
  • Modeling Approach
  • Model Performance
  1. Clinical Measurement Reports

Morning Fasting Blood Glucose

Description

Blood glucose concentration is controlled to approximately 5 mmol/L in the fasting state before breakfast. The importance of maintaining the constant blood glucose concentration is because it is the only nutrient that can be used by the brain and retinal cells in sufficient quantities to supply them with the required energy levels.

Fasting blood glucose risk corresponds to the percentage of people with the user's risk profile who had their blood glucose levels above 5.5 mmol/L when tested after 8-10 hours of fasting, indicating a high risk of prediabetes. Their risk profile is based on facial blood flow and demographics.

Participants

Adults (18+ years of age) recruited from several hospital health clinics.

Additional Data Collection Procedure

Subjects were asked to provide their blood work report to determine their morning fasting blood glucose (mFBG) levels.

Modeling Approach

Blood flow signal was extracted and processed from facial video, and then blood flow features were extracted from blood flow signal.

Feature selection was carried out on blood flow and demographic features to identify features predictive of mFBG levels above 5.5 mmol/L. A machine-learning based classifier was then created to predict an individual’s percentage likelihood of having mFBG above 5.5 mmol/L based on those features.

The model was created/trained using 80% of the subjects. The distribution of sex and mFBG level above 5.5 mmol/L in this group is summarized in Table 23.

Subjects in videos
Composition

Sex distribution

52 % Male; 48% Female

mFBG > 5.5 mmol/L distribution

30%

Table 23: Morning fasting blood glucose > 5.5 mmol/L distribution of training set

Model Performance

The mFBG model was then validated for accuracy on an independent portion of the dataset (n=917) that was not used in training - validation set. Then the final accuracy was obtained on an independent portion of the dataset (n=917) that was not used in training - test set.

The accuracy on the test set calculated as area under the curve was 80.4% as shown in Figure 28.

Figure 28: AUC of prediction for morning fasting blood glucose value > 5.5 mmol/L

Predictions are displayed to the user as a percentage likelihood of having prediabetes. A percentage of 45% or more suggests a risk of prediabetes, while 55% or more suggests high risk of prediabetes.

PreviousFatty Liver DiseaseNextHemoglobin A1C

Last updated 4 months ago

Page cover image