Pulse.

a daily field guide to health research that matters

◆ Console

‹ Fri · 3 Jul 2026
Promising but preliminary

Deep multi-modal features based spatio-temporal video regression for non-invasive hemoglobin estimation.

Deep learning estimates blood haemoglobin from facial video, potentially enabling portable anaemia screening in resource-limited settings.

Gunes et al. developed a deep learning framework combining multimodal spatial and temporal video features to estimate blood haemoglobin concentration non-invasively from facial video recordings, achieving mean absolute errors approaching clinical laboratory thresholds for anaemia screening. Published in Medical and Biological Engineering and Computing, this model represents a step toward portable contactless haematological monitoring applicable in resource-limited screening settings.

What the study was

Study design
Technical validation study (deep learning on facial video)
Population
Participants in haemoglobin estimation validation study using facial video
Category
Diagnostics
Maturity
Exploratory
Journal
Med Biol Eng Comput

Why it surfaced

Novel non-invasive haemoglobin estimation via video deep learning bridges T4 AI diagnostics; potential anaemia screening tool for resource-limited settings.

A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.