DOCSEARCHbyAlgolia
Submit Docs

AI-READI Dataset Documentation

website·Indexed 33 days ago

https://docs.aireadi.org/

Documentation for the AI-READI dataset, designed to boost AI-driven research in Type 2 Diabetes. It provides technical details on data domains including clinical lab tests, vision and physical assessments, Montreal Cognitive Assessment (MoCA), and wearable monitors (Dexcom G6 glucose and activity monitors). The docs cover retinal imaging (OCT, OCTA, and photography), data processing workflows, DICOM standards, OMOP Common Data Model adherence, and instructions for accessing data via Azure Storage.

Tokens
106K
Snippets
277
Records
647
Agent score
60%

What's inside AI-READI

  1. Overview of AIREADI Wearable Activity Monitor Data

    The AIREADI wearable activity monitor data domain (v3.0.0) collects health metrics using the Garmin Vivosmart 5. Data is provided in .JSON format and is categorized as Open Source. The following variables are tracked:

    • Heart Rate (bpm)
    • Oxygen Saturation (pulse ox)
    • Physical Activity (steps)
    • Physical Activity (calories)
    • Respiratory Rate (bpm)
    • Sleep data (hrs)
    • Stress (stress level)

  2. Overview of the AI-READI Dataset

    The Artificial Intelligence Ready and Equitable Atlas for Diabetes Insights (AI-READI) is a dataset containing harmonized data from individuals with and without Type 2 Diabetes Mellitus (T2DM) across three collection sites. It is specifically designed for AI/ML research, with recruitment sampling aimed at achieving an equal distribution of participants across sex, race, and diabetes severity to better understand salutogenesis (the pathway from disease to health) in T2DM.

  3. Overview of Montreal Cognitive Assessment (MoCA) data

    The MoCA data domain contains screening results used to assess cognitive function (short-term memory, visuospatial abilities, executive functions, attention, concentration, working memory, language, and orientation). For the AI-READI project, data is captured via the MoCAtest Duo App on tablets by trained Clinical Research Coordinators.

    Scoring Logic:

    • Maximum score: 30 points.
    • Normal: 26 points and above.
    • Cognitive Impairment: Below 24 points.
    • Education Adjustment: Participants with 12 or fewer years of education receive 1 additional point.

  4. Overview of Wearable Activity Monitor data collection

    The AI-READI research program uses the Garmin Vivosmart 5 fitness tracker to monitor participant activity over a 10-day period. The device is worn on the non-dominant hand (though participants can choose their preferred wrist) and records physical activity and sleep data continuously. Note that data gaps exist due to battery charging requirements every 2-3 days.

  5. Overview of Montreal Cognitive Assessment (MoCA) data collection

    The Montreal Cognitive Assessment (MoCA) is used to screen for cognitive impairment, targeting domains such as short-term memory, visuospatial abilities, executive functions, attention, concentration, working memory, language, and orientation. In the AI-READI project, data is captured electronically via the MoCAtest Duo App on tablets and administered by trained Clinical Research Coordinators. The maximum score is 30; a score below 24 indicates impairment, while 26+ is considered normal. Participants with 12 or fewer years of education receive one additional point.

  6. Overview of Vision Assessment Data Domain

    The vision assessment data domain tracks retinal health in Type 2 diabetes patients, focusing on the impact of hyperglycemia and blood glucose fluctuations on refractive power, visual acuity, and contrast sensitivity. The domain captures three primary measurements: Autorefraction, Best Corrected Visual Acuity (BCVA), and Mars Letter Contrast Sensitivity.

  7. Overview of AI-READI Retinal Photography Imaging Types

    The AI-READI dataset contains three primary types of retinal photography used to visualize the posterior segment (fundus) of the eye:

    • Color Fundus Photography: Direct color visualization of the fundus. Used to identify abnormalities like microaneurysms, retinal hemorrhages, cotton wool spots, and neovascularization.
    • Fundus Autofluorescence (FAF): A lipofuscin density map generated by the intrinsic fluorescence of the retinal pigment epithelium (RPE). Brighter zones indicate higher lipofuscin concentration. Used for evaluating RPE alterations in diseases like age-related macular degeneration.
    • Fundus Infrared Photography: Provides deeper penetration into subretinal layers and better vessel-to-background contrast. It is effective for detecting pathologies even in the presence of media opacity (e.g., cataracts or hemorrhages) and is often used during OCT imaging.

  8. Overview of the AI-READI Dataset for Type 2 Diabetes

    The AI-READI dataset, titled 'Flagship Dataset of Type 2 Diabetes from the AI-READI Project', is a multimodal, AI-ready dataset developed under the NIH-funded Bridge2AI Program. Its primary objective is to provide data for studying salutogenesis in Type 2 Diabetes Mellitus.

  9. Overview of the AI-READI Type 2 Diabetes Dataset

    The AI-READI dataset (Flagship Dataset of Type 2 Diabetes) contains harmonized data from individuals with and without Type 2 Diabetes Mellitus (T2DM) across three collection sites. It is specifically designed for AI/ML research, featuring recruitment sampling for balanced diabetes severity and a multi-domain data acquisition protocol (surveys, physical measurements, clinical data, imaging, and wearables). The primary goal is to study salutogenesis (the pathway from disease to health) in T2DM.

  10. Overview of the AI-READI Flagship Dataset for Type 2 Diabetes

    The AI-READI dataset, titled 'Flagship Dataset of Type 2 Diabetes from the AI-READI Project', is a multimodal, AI-ready dataset designed for studying salutogenesis in Type 2 Diabetes Mellitus. It is part of the National Institutes of Health (NIH) funded Bridge2AI Program.

  11. Overview of Clinical Lab Tests Data Domain

    The Clinical Lab Tests data domain (Version 3.0.0) provides diagnostic data from blood, urine, tissue, and other bodily substances. It consists of five key elements: Test Name and Description, Interpretation (analysis of results relative to patient history), Measurement Units (e.g., mg/dL or cells per microliter), Reference Ranges (norms based on age, sex, and population), and Method of Collection (e.g., blood draw, biopsy). Reference ranges are provided for tests performed by the UW NORC lab and CBC tests from local labs.

  12. Overview of Wearable Activity Monitor Data Domain

    The Wearable Activity Monitor data domain (v1.0.0) collects health metrics using the Garmin Vivosmart 5 device. All data is stored in .JSON format and is categorized as Open Source. The domain includes the following variables:

    • Heart Rate (bpm)
    • Oxygen saturation (pulse ox)
    • Physical Activity (steps)
    • Physical Activity (calorie)
    • Respiratory Rate (bpm)
    • Sleep data (hrs)
    • Stress (stress level)