AI Model Forecasts 1,000 Diseases

FEATURED RESEARCH

AI Model Predicts Risk for Over 1,000 Diseases Decades in Advance

I’ve featured “predict-the-future” AI tools before, and while most predict one disease, this one maps health as a timeline. A team from EMBL, DKFZ, and the University of Copenhagen trained an AI model, Delphi-2M, on 400,000 UK Biobank participants and validated it on 1.9 million Danish patients.

It estimates the chance of developing more than 1,000 conditions over the next 20 years, including cancers, heart attacks, diabetes, and sepsis.

What the model actually does: It reads a person’s history as a sequence of events: diagnoses, body mass, smoking and alcohol use, plus long quiet stretches.

From that, it estimates what is likely to happen next. It can also generate “synthetic health futures,” letting researchers test ideas without exposing real patient data.

Why this matters for care and planning: Chronic illness is rising. In England, about 700,000 more working-age adults are expected to live with a major disease by 2040.

A model that flags rising risk years earlier can advance screening schedules, target prevention to the right people at the right time, and help health systems plan clinics, staffing, and budgets. For individuals, it highlights windows of opportunity when lifestyle changes or medications are most likely to be effective.

Important limits: This is not a clinical tool yet. Training data skew toward ages 40-60 and largely European ancestry, so performance may drop in younger people and underrepresented groups.

Mental health and pregnancy-related outcomes were harder to forecast because life events drive much of that risk.

As a research instrument, though, Delphi-2M gives a clearer map of how illnesses cluster and when risk begins to climb.

For more details: Full Article 

Brain Booster

Select the right answer! (See explanation below and source)

What Caught My Eye

PRODUCT PIPELINE

AI Reconstructs Fetal Shape and Pose with Millimeter Accuracy for Better Diagnosis

MIT CSAIL researchers, with Boston Children’s Hospital and Harvard Medical School, have created Fetal SMPL, a machine-learning model that turns fetal MRI scans into detailed 3D reconstructions.

Unlike standard MRIs, which can be hard to interpret, Fetal SMPL models the fetus’s shape and pose with millimeter-level accuracy, small enough to measure the size of the head or abdomen and compare them with healthy norms.

The system was trained on 20,000 MRI volumes and uses a “kinematic tree” with 23 joints to represent fetal motion realistically. In tests on pregnancies between 24 and 37 weeks, it consistently aligned with unseen MRI data within about 3.1 mm.

Doctors could use these models to detect growth abnormalities earlier and guide clinical decisions.

The team plans to expand testing across more cases and develop volumetric models that capture internal anatomy, potentially making fetal MRI a much more powerful diagnostic tool.

For more details: Full Article

Top Funded Startups

Byte-Sized Break

📢 Other Happenings in Healthcare AI

• AlterEgo is a non-invasive device that turns silent speech signals into words, now being tested to help ALS and MS patients communicate. [Link]

• Melbourne researchers are commercialising Baby Moves VIEW, an AI app that screens infants for cerebral palsy from home, aiming to cut diagnosis age from 19 to 3 months. [Link]

• Nolla Health launches its AI-powered Acne Care app in 40+ U.S. states, providing instant AI skin scans, clinician-reviewed plans, and home-delivered meds to bypass long dermatologist wait times. [Link]

Have a Great Weekend!

Trivia Answer: A. Ovum (egg cell)

The human ovum is the largest cell in the body—visible to the naked eye! It's much bigger than other cells because it needs to store nutrients to support early development after fertilization. [Source]