Skip to main content

QUantitative Imaging Biomarkers Medicine

Article Category

Article available in the folowing languages:

Artificial intelligence in radiology imaging: more than meets the eye

Algorithms based on machine learning are increasingly finding their way into many medical applications. European scientists have generated a novel platform that enables the quantitative analysis of radiological images.

Digital Economy

The EU-funded QUIBIM Precision project developed an innovative processing technique for radiological images to assess changes in the body. The project applied artificial intelligence and advanced computational models to associate phenotypic changes with diseases, lesions or pharmacological treatments, offering quantitative information on the qualitative nature of radiological imaging.

Combining algorithms and machine learning in a single analysis platform

“QUIBIM Precision® enables any physician to make more accurate diagnosis by providing additional information extracted from the same imaging sample,″ explains project coordinator and QUIBIM CEO and founder, Dr Angel Alberich-Bayarri. The technology can be used on a variety of modalities (MRI, CT, X-rays, PET-CT) and the goal is to help radiologists more effectively detect and track the progression of diseases. QUIBIM Precision® relies on machine learning and image processing algorithms to scout the image and compare it to similar diagnosed images in given databases. The diagnosis is based on patterns not obvious to a human eye. Machine learning is implemented in two stages, first during the creation of the algorithms and then at the patient level for associating imaging features and patients. This enables the discovery of new imaging features as potential biomarkers of a specific disease. The platform can be used both in hospital environments and by pharmaceutical companies in clinical trials. In hospitals, the platform is connected to the imaging repository, and it constantly looks to apply different algorithms to the new examinations uploaded by the different machines of the hospital. When it finds a match, it applies a specific algorithm and undertakes the corresponding analysis. Take for example, the detection of abnormalities in chest X-rays or the detection of brain iron deposits in an MR exam. Analysis is entirely automated and the clinician alongside the radiology image receives the QUIBIM quantitative information. In clinical trials, QUIBIM Precision® centralises all data (imaging and non-imaging) from the different centres involved in a clinical study, making the evaluation of response to treatment more objective.

QUIBIM Precision® merits

QUIBIM Precision® is the first cloud platform that allows the automated analysis of imaging biomarkers with high sensitivity and specificity. It has been medically certified to support decision making and is open to physicians everywhere in the world. Moreover, it is a cost-effective approach that minimises misdiagnosis due to human error and the need for re-examination. The platform has already demonstrated added value in suspicious cases of prostate cancer with the prostate cancer MR algorithm among other pathologies. So far, the platform has a total of 24 algorithms and has analysed more than 5 million images with customers from 22 countries. Applications include the detection of demyelinating brain lesions in multiple sclerosis, changes in brain morphometry as an indication of Alzheimer’s disease, cancer lesions as well as osteoporosis and osteoarthritis. According to Alberich-Bayarri “the inclusion of algorithms within our platform with patented artificial intelligence technologies, and the introduction of new visualisation tools have been key to the company going global.″ The platform is already being used in 70 hospitals both in clinical routine and as part of clinical trials. “Obtaining clearance by the FDA is paramount to taking the platform to the US market,″ concludes Alberich-Bayarri.


QUIBIM Precision, algorithm, imaging, machine learning, artificial intelligence, radiology, biomarkers

Discover other articles in the same domain of application