This week's rounds
Can AI help scientists design cancer medicines?
A company says AI helped design IAM1363, an experimental pill for HER2-driven cancers that's built to reach the brain. In an early trial, tumours shrank in 19 of 53 people (12 of 53 confirmed on a follow-up scan), and brain tumours shrank in 9 of 22 (4 of 22 confirmed). Stomach upset was common.
- It's early: no comparison group, and not peer-reviewed. The clinician decides.
- The AI role is the company's claim.
- Iambic sponsored the study, and it listed on Nasdaq on Oct 15, 2026.
Poster: Dumbrava et al., ESMO 2026, FPN 1072P, Oct 23, 2026. Read the poster · Read the full story
An AI that estimates biological age from a face photo.
A computer looks at a face photo and estimates biological age, a measure of how the body is ageing. Researchers at Mass General Brigham describe their latest FaceAge model in a new paper. It was built and tested on existing photos and past records, and it isn't used to guide anyone's care.
Many older adults with early lung cancer have needs you can't see, but very few get a full check of how they're managing. That check can lead to help like nutrition advice or physical therapy before radiation begins.
That's a separate study. A small trial at Brigham and Women's Hospital is testing a short questionnaire for adults 65 and older having radiation for early lung cancer, to see whether more get referred for support. It also collects photos to study whether AI like this could someday help.
- It's retrospective research, not a screening tool. The clinician decides.
- The trial is separate from the paper. It looks at referrals to support services.
- An outside expert called it promising but says it needs testing in more hospitals and patient groups.
- Some authors, including senior authors Dr. Mak and Dr. Aerts, report consulting, advisory or research ties to drug and device companies. The study was funded by the European Research Council. Mass General Brigham has filed a provisional patent application on these tools; the first author and both senior authors declared it.
Study: Haugg, Lee, He et al., "Foundation artificial intelligence models for health recognition by use of face photographs (FAHR-Face): a model development and validation study," The Lancet Digital Health, Oct 19, 2026. Read the study · Read the full story
Hearing one voice in a noisy room.
In a blinded lab test with loud background chatter, every one of 30 hearing aid wearers understood more words with an AI hearing aid. The maker funded the study, and it was one controlled setup, not a real restaurant.