Create Jira Ticket
This idea cannot be sent to Jira.
MEASURE it to CURE it: AI Powered Clinical Innovation
MAESTRO Health Innovators brings together clinical, AI, and patient-advocacy expertise in invisible illness: Tomer Tal, PhD (AI); Mady Hornig, MA, MD (Feinstein Institutes); Nasa Sinnott-Armstrong, PhD (Fred Hutchinson Cancer Center); Tyler Myers, PhD (AI research, Complex Disorders Alliance); Paige Hansen Colburn (MIT); Amy Rochlin (CEO, Complex Disorders Alliance); and Michal Caspi Tal, PhD (MIT). Through the Complex Disorders Alliance, our work is shaped directly by the patient community.
People living with Long COVID, ME/CFS, POTS, and related conditions often wait years for a diagnosis because their illness is hard to measure. Objective measurement devices exist, from wearables to nailfold capillaroscopy, but there is no systematic way to know which measures actually separate these conditions from health, or where deploying a new device would help the most patients.
Measure it to Cure it is an AI engine that connects novel measurement devices to the U.S. open data ecosystem. A researcher or device developer asks a plain-language question; the engine resolves conditions, devices, and places through a medical ontology layer, grades the evidence for each measurement, estimates need in every U.S. county, and returns ranked, traceable deployment opportunities with nearby clinics, trials, and funded labs. The same engine works in reverse: measurements already present in public data, such as wearable activity, can be tested for how well they separate a condition from controls.
Other tools find signals in health records. Ours links those signals to objective, affordable measurement devices and shows where to deploy them.
Our real-world example is CAPRIO, a clinical study at MIT launched September 14, 2026 and designed to become the largest nailfold capillaroscopy dataset. CAPRIO shows how a new device moves from "unknown" in our evidence table to measured, validated, and deployed where it is needed most.
Solution Overview (MVP Demonstration)
Interactive Visualization Too
Live demo, prototype, or working product
GitHub repository (Docker container): https://github.com/tydymy/measure-it-public/t...shboard_visuals
Access instructions (if needed)
Clone the repository and run the container with docker compose up -d --build, then open http://localhost:8501 in a browser to access the interactive dashboard. The System page summarizes the pipeline; Ask accepts plain-language questions; Map shows the county deployment heatmap; Embedding shows how public cohort records relate; Evidence shows how well each measurement separates conditions from controls. Outputs are candidate deployment opportunities for pilot evaluation, not diagnoses, and no individual is ever placed on the map. More instructions can be found at: https://talresearchgroup.mit.edu/maestro-health-innovators
Results for the American People
Our users are invisible illness clinicians and researchers, and the device developers who build new ways to measure these conditions. Our user research drew on the MIT MAESTRO clinical study at MIT and further sessions with patients, clinicians, and researchers. Clinicians described a clear need for objective, point-of-care measures. Patients described years of being dismissed; for many MAESTRO participants, seeing their own capillaries was the first time they felt seen by the medical system. Nurses described the need for low cost, quick, non-invasive testing. Parents mentioned how capillaroscopy use could be easily performed on their children — harmlessly, quickly, and not panic-inducing.
To exemplify this, we took a trip to America’s heartland at the University of Nebraska Medical center in Omaha, Nebraska. When showcasing this pipeline that identified the clinics, followed by the usage of capillaroscopy, people were wowed. Clinicians wanted to see it in their work places. Medical students were curious about the diagnostic potential. Everyday folk were excited to peer noninvasively into their health.
Measure it to Cure it fits into a real workflow. Researchers with new devices can ask a question. Within minutes, it returns locations with high need for their device, hospital and clinician names to reach out to, and evidence behind each measurement, allowing technology to be quickly deployed where it will be the most useful.
Data, Tech & Evidence Driven Approach
The engine connects the dots across 37 public sources organized into 160 tables, including ClinicalTrials.gov, NIH RePORTER, CDC surveys, and NHANES exam, laboratory, questionnaire, and wrist-activity data. Together these cover 329,019 facilities, 70,295 molecular evidence rows, and all 3,144 U.S. counties.
An ontology layer (Mondo, HPO, and ICD-10-CM) joins concepts rather than people, resolving 14 conditions from plain language. A phenotype module learns what each device sees in public cohorts; for example, daily steps separate Long COVID or ME/CFS from healthy or surgical controls with an AUROC of 0.831. A record embedding of 1,500 NHANES adults shows functional limitation as a continuum rather than discrete clusters. An evidence module reports each measurement's performance, our own analyses first, and labels untested measures "unknown" rather than estimating them. A burden model grades need A to D by county, and an opportunity model ranks 2,395 counties with components and uncertainty.
The pipeline runs 75 reproducible offline steps with 18 tools and 1,428 automated tests, and ships as a Docker container. Device-paired research cohorts such as CAPRIO plug in as new measurement modules, and All of Us patient-level EHR and multi-omic data will strengthen phenotype validation once access is granted.
Execution, Efficiency & Delivery:
Today, choosing where to pilot a new diagnostic device depends largely on personal networks. Measure it to Cure it replaces that with a transparent, ranked answer drawn from public data. One question returns the evidence for a measurement, the counties with the greatest unmet need, and the clinics, trials, and funded labs nearby – information that would otherwise require assembling dozens of public datasets by hand.
Every value is traceable to its source, and burden values are graded for quality, so decision-makers can see why a county ranks highly and how certain the estimate is. The evidence table also tells developers and funders which measurements are already validated and which remain unknown, directing studies like CAPRIO to the gaps that matter most.
The MVP works end-to-end today: 37 sources, 2,395 counties ranked, and 1,428 automated tests passing, packaged as a Docker container so research groups can run it on their own hardware without new infrastructure. CAPRIO launched on Sept 14th at the Project Lyme Gala to give people access to their own capillaroscopy results and initiate a new pipeline of non-invasive measures of microvascular health. Twenty clinics will be newly integrating capillaroscopy as part of the launch of CAPRIO clinics on Nov 14th at The Alex Manfull Symposium in New Hampshire. Measure it to Cure it will help identify key clinics where capillaroscopy could enable faster diagnosis and inform better clinical care.
Scalability, Reuse & Societal Impact:
The engine is modular by design. A new measurement device, whether capillaroscopy, retinal imaging, or a wearable, is added as a module without changing the pipeline, and the ontology layer extends to new conditions. It already codes 14 conditions, including Long COVID, ME/CFS, and POTS, and the same approach applies to Lyme disease and other tickborne illnesses. Its county-level need estimates can also inform cost-of-illness analyses.
Because it is open source and containerized, health systems, universities, and agencies can deploy it locally and connect their own data. Next, we will add FHIR-based outputs aligned with USCDI+ so results can flow into existing clinical and research systems.
Key assumptions: public data reflect relative need across counties, and device performance measured in research cohorts carries over to the clinics where devices are deployed. CAPRIO is our first test of both.
Next steps: integrate CAPRIO capillaroscopy results as the first device-paired module, leverage the All of Us data findings for patient-level validation, run pilots in top-ranked counties, and publish the evidence table so researchers and funders nationwide can see where objective measurement is still missing.
Watch our MVP! https://www.youtube.com/watch?v=53zp-OtNWsA
Reply 4
so exciting!!
Reply 2
There is such a need for this important work! - Alessandra Angelino MD, Pediatrician and Adolescent Medicine Specialist, Long Covid Families
Reply 4
Thrilled to work with this stellar team to help make invisible illness visible!
Reply 1
I vote Tal. Always.
Reply 2
Labeling untested measures “unknown” instead of estimating them is a choice I wish more tools made, and the evidence table that shows where objective measurement is still missing looks like a valuable output on its own.
I work on turning published Lyme and Long COVID papers into runnable simulators. Simulator 14 in my library covers an objective-measurement study from Germany (Tomaskovic et al., Sports Med Open 2026): 92 people with severe post-COVID illness assessed for work disability. Exercise testing found objective work disability in 66.3%, while 75% were on medical sick leave and only 13% were working (chart attached). It’s one clinic and a referred group of severely ill patients, so it describes that group and not everyone with Long COVID, but it shows what an objective measure can tie to.
My entry is here, in case it’s useful: https://invisibleillness.crowdicity.com/post/896844
Thank you for sharing the work, and good luck with the CAPRIO clinics in November.
Reply 0
thanks for the important work
Reply 1
🫶🫶🫶
Reply 1
Are you aware and connecting with the Neuroimmune.org researchers? They are a consortium of researchers, many are Stanford pediatricians whose researcb focus is infections causing autism and mental illnesses. For ex: 3Bs babesia, Bartonella, borrelia, PANS.
More broadly, they have the UNHIDEnow.org site for measurable devices for over 30 neuropsychiatric disorders. Christy Jagdfeld leads it and is acrltive on LinkedIn. I have many lecturers and advocacy notes on my LinkedIn posts and profile reversing chronic diseases.
2026 need a patient rights to ethical care to require "test don't guess" reversible causes before drigs, reserch or expensive, intesse ABA behavior therapy:vin aitism. Test for: 1. Nutrients/genomics 2. Infections 3. Toxins as it alreafy reverses neuops6ch disorders as in the 2025 EVANTHEA trial. Not testing can skew research results or make it useless as ABCD trial imaging 12 000 brains without testing Lyme, B12/MTHFR, toxins so no help for autism in 20 years of data collection by IACC. Shoukd also test binocular vision, bone and air hearing nd brain region connections, cross hemispheric ability. These are in Fit to Learn per Charlotte Davies founder, author and takes 15 min. Has helped ,most, including a boy with "autism" go to PhD.
Reply 0
Michal, really impressive work! I especially like how you're using AI and public health data to identify gaps in diagnosing conditions that are often difficult to recognize or measure.
I've been working on LymeSignal AI by Ace for the Lyme Innovation track, and I see some interesting similarities in how we're both trying to turn fragmented medical evidence into something more useful and actionable. While your focus is on identifying and validating measurement tools, mine is more on helping make sense of existing Lyme-related evidence.
I'm curious, do you see your platform eventually connecting these new measurements directly with patient-facing decision-support tools? I think there's some interesting potential for the two approaches to complement each other. Great work!
Reply 0