The AI ensures that patients can answer all relevant questions that apply to...
The AI ensures that patients can answer all relevant questions that apply to their individual diagnosis, which makes the job easier for the physician

Image source: Danderyds sjukhus

News • Smart diagnostic tool

CLEOS: The AI that listens to the patient

A new digital tool that tailors specific questions based on a patient's medical history could improve the safety of diagnosis and effectiveness of care, according to a new study at Danderyd Hospital and Karolinska Institutet.

"The AI ensures the patient is asked all relevant questions for that particular individual," says doctoral student Helge Brandberg, one of the developers behind the new digital tool called CLEOS. 

The study started as early as May 2017. By October this year, more than 1,200 patients with acute chest pain at Danderyd Hospital had participated in the study. The software was developed originally in the U.S. by Professor David Zakim. It has since been developed further at Danderyd's Cardiac Clinic in collaboration with the Department of Clinical Sciences, Danderyd Hospital, and the Department of Learning, Informatics, Management and Ethics (LIME) at Karolinska Institutet. The system has been adapted to Swedish conditions by Helge Brandberg and his supervisors, Tomas Kahan and Jonas Spaak, and KI's Carl Johan Sundberg and David Zakim. 

The study was now published in the journal BMJ Open.

An AI program can handle much more data than humans, which in turn enables a more accurate diagnosis. It does not get tired or stressed. It never has a 'bad day' or reacts emotionally

Helge Brandberg

"Today, we rely too much on results from blood and imaging tests and other expensive investigations," says Helge Brandberg. "Important diagnoses can be missed in the absence of a careful medical history. For example, patients with unstable angina can have similar symptoms to patients having a heart attack, and unstable angina and heart attack can be equally life-threatening."

The tool is intended for patients without life-threatening conditions who have the ability to take time to answer CLEOS' questions while waiting for the doctor. Additional research is needed before the tool can be used in clinical practice. "An AI program can handle much more data than humans, which in turn enables a more accurate diagnosis," says Helge Brandberg. "In addition, an AI program does not get tired or stressed. It never has a 'bad day' or reacts emotionally. It is a safe complement to other tests and the doctor's assessment. But what it is all about is saving lives. With safer and faster diagnoses that lead to the right intervention at the right time, we can create the best conditions to do so."


Source: Danderyds sjukhus/Karolinska Institutet

23.01.2020

Related articles

Photo

News • Patient-written symptom descriptions

ChatGPT struggles when medical questions are put in layman's terms

Enter symptoms into ChatGPT, receive an accurate diagnosis? Research reveals that LLM AI models are not quite there yet, struggling to identify genetic conditions from patient-written descriptions.

Photo

Article • Digital PET imaging

Digital Photon Counting improves diagnostic accuracy

Built as the first commercially available scanner to deliver truly digital PET, the Vereos PET/CT, from Philips, offers revolutionary Digital Photon Counting technology. The science behind this…

Photo

News • Misleading medical analyses

AI “predicts” beer drinking based on knee X-rays – why this is not only wrong, but dangerous

Can an AI determine whether or not a person drinks beer by looking at their knee X-rays? It can't – but the claim shows why “shortcut learning” is such a dangerous mechanism in medical AI.

Related products

Subscribe to Newsletter