New AI-based method to diagnose breast cancer

A new artificial intelligence-based method to diagnose breast cancer is developed in an international partnership featuring IFIC

A team of researchers from the Instituto de Física Corpuscular (IFIC), has participated in the development of a system to help in the diagnosis of breast cancer. The development, which will be of great help in clinical practice, is capable of reducing the number of false positives. It has a detection accuracy of 90 %, one of the highest in this kind of screening systems. In addition to IFIC, which is a joint institute of the Sspanish Consejo Superior de Investigaciones Científicas (CSIC) and the Universitat de València, the Universitat Politècnica de València and research groups from seven other international institutions participate of this work.

As the researchers point out, current methods of diagnostic support employed by radiologists are most often limited to only observing and detecting potentially suspicious areas in tests obtained by imaging techniques. The developed device, for its part, is capable of reducing the number of false positives, and to give more reliable information on the likelihood of cancer presence. This is attained on basis of AI techniques as neural networks, and the use of predictive algorithms.

Breast cancer is one of the types of cancer with the highest incidence in developed countries. Mamographies are diagnostic techniques for its early detection which have proven their efficacy over the years. Complementarily, the new system will be able to decrease the number of false positives in all age ranges, and will minimise false alarms, avoiding the unnecessary use of more aggressive diagnostic tests. It allows, as well, for a reduction of clinical costs, which could help to include new risk populations into screening campaigns.

Several phases of lesion detection using the new diagnostic test
Several phases of lesion detection using the new diagnostic test

“In addition, if other clinical evidence hints the professional on the possibility of non-evident positive diagnostic, he or she can amplify image regions which are cause of higher concern, surpassing the capabilities of an expert human eye. Also, this could help to determine future biopsy locations.”, says Dr. Francisco Albiol, CSIC researcher at IFIC, a Severo Ochoa centre of excellence.

“For each year a developing breast cancer is diagnosed earlier, a 20 % increase in the 5-year survival rate results increase a 20 %. The algorithm we have developed can be a great tool towards early diagnosis of this type of cancer, offering clinicians an additional advanced diagnosis tool”, says Francisco Albiol.

The participants in this project currently study how to translate the method into clinical practice. “One of the simplest possibilities would be its application to reduce the fatigue of radiologists, by screening the easiest cases”, adds Alberto Albiol, researcher from the Universitat Politècnica de València.

Digital Mammography DREAM Challenge

DREAM Challenges pose problems relating to the fields of biology or medical research. Participant researchers present their projects, which have to positively impact society, and which are subject to evaluation.

The development is a result of the Digital Mammograpy DREAM Challenge, a world project driven by the main American institutions fighting against breast cancer, hand in hand with multinationals as IBM and Amazon. The aim is to improve the detection of breast cancer by the interpretation of mammographies by artificial intelligence. 120 multidisciplinary teams participated of this challenge, of which the Instituto de Física Corpuscular – Universitat Politècnica de Valencia team was the only Spanish participant.

In the study, data of patients provided by United States medical institutions were analysed. The results were presented in the Congress of the International Society for Computational Biology, which took place in New York.

“To be able to generalize large-scale use of these kinds of technologies, it is of utter importance to generate and maintain local collections of patient data representing the general ethnical, [social] and economic state of a public health system” stresses Francisco Albiol.


“Several phases of lesion detection using the new diagnostic test” by IFIC / CSIC / UV / UPV.