Primary domain: Clinical & Healthcare Innovation (Clinical Decision Support)
Secondary domain: Artificial Intelligence & Machine Learning (Computer Vision)
Article type: Innovation Spotlight
Technology: Image-based AI blood loss estimation
Innovation maturity: Emerging
Mission: Clinical, Research
Business impact: Faster, more consistent surgical measurements and safer item counting
Audience: Clinical, Research, Business/Leadership
Publication date: November 2026
Author / Contributor: Innovation team
Related technologies: ESTIMATA-EX, AI for Thai platform, LANTA supercomputer, NECTEC, Computer vision
In many operating rooms, one of the most important numbers is still an educated guess: how much blood has the patient lost? Chiang Mai University in Thailand says it has built an AI system to replace that guess with a measurement. In a university announcement dated 2 November 2026, the team introduced ESTIMATA-EX, which estimates blood loss and counts blood-soaked surgical materials from photos taken on a smartphone or tablet. This article covers the announcement as news and is not medical advice.
The work comes from a research team in the Department of Anesthesiology, Faculty of Medicine, led by Nurse Natthikarn Charoenrattanadechakul, in collaboration with Thailand’s National Electronics and Computer Technology Center (NECTEC).
The problem with eyeballing blood loss
According to the university, visual estimation of blood loss varies with the experience of the person doing it. That inconsistency matters because surgical teams use the estimate to make decisions during an operation. A quantitative tool that gives the same kind of answer regardless of who is in the room could support those decisions and improve patient safety, which is the stated aim of the project.
How ESTIMATA-EX works
The system analyses images of gauze and surgical swabs and returns quantitative estimates. It was developed on the AI for Thai platform and the LANTA supercomputer under the Medical AI Consortium. Because it relies on ordinary smartphone or tablet cameras, it does not require a dedicated scale or bespoke hardware at the operating table, at least as described in the announcement.
It also counts the blood-soaked surgical materials it sees. The university says this supports accurate counting of surgical items, which helps prevent retained surgical items, a long-recognised patient safety concern.
What the numbers show
The university reports R² values of 0.95 for gauze and 0.97 for swabs, which indicates close agreement between the system’s estimates and reference measurements. It says the tool achieved approximately twice the clinical accuracy of visual assessment and delivered results up to 20 times faster. The announcement does not describe the full test protocol, so these should be read as the team’s reported results rather than independent validation.
The system has reached Technology Readiness Level 6 to 7 and has undergone prototype testing under real operating-room conditions at Maharaj Nakorn Chiang Mai Hospital. There, it recorded a clinical satisfaction and acceptance rate of 96.5%. The work has also drawn recognition, including publication in Discover Artificial Intelligence (Springer Nature), awards at the MED CMU Quality Fair and Nursing Council, second runner-up at the 26th HA National Forum, and a 2026 Public Sector Excellence Award (Good Level) in the Service Innovation category.
What it means
The next steps show how far there is to go. The team plans multi-center validation and is preparing to register ESTIMATA-EX as Software as a Medical Device (SaMD) with the Thai Food and Drug Administration. Roughly 30 public and private hospitals have expressed interest in testing it.
For hospital leaders, the project is a good example of a narrow, practical AI use case: one measurement, one workflow, with commodity devices. It also shows a development path worth noting, starting with a university clinical team, a national technology agency and shared national compute, then moving through prototype testing in a live hospital before seeking regulatory registration.
For clinicians and IT teams, the open questions are the usual ones for any clinical AI: how it performs across multiple sites, lighting conditions and surgical types, how it fits into existing counting and documentation routines, and what regulators require. Multi-center validation should begin to answer them.
Conclusion
ESTIMATA-EX turns a subjective judgment into a number using a device already in every operating room. The reported accuracy and speed are promising, but regulatory registration and multi-center results will decide how far it travels beyond Chiang Mai.