Researchers at the Vall d’Hebron Institute of Oncology (VHIO), in collaboration with radiologists from the Bellvitge University Hospital, have introduced an innovative tool based on artificial intelligence (AI) to enhance the diagnosis of brain cancer. This tool, named Diagnosis In Susceptibility Contrast Enhancing Regions for Neuroncology (DISCERN), has been developed through pattern learning from standard magnetic resonance imaging (MRI), and its outcomes, published in the journal Cell Reports Medicine, demonstrate superiority over conventional methods.

The DISCERN software relies on deep learning, an AI method that utilizes all spatial and temporal information from standard MRIs. This technique enables the identification of specific behavioral patterns in each tumor image, which is crucial as 70% of brain tumors are of different types, each requiring a distinct therapeutic approach.

Raquel Pérez-López, a researcher at VHIO, highlights that the non-invasive differential diagnosis of these tumors is currently based on MRI, but often neurosurgical intervention is necessary for a definitive diagnosis, compromising the patient’s quality of life. DISCERN automates pre-surgical diagnostic classification with high precision and offers a user-friendly interface for clinicians.

Alonso García-Ruiz, a VHIO researcher and the study’s lead author, explains that DISCERN has been trained with characteristics of brain tumors identified in MRI images of already diagnosed patients. Validation of the tool in over 500 additional cases showed that 78% of the diagnoses made by DISCERN were correct, a rate higher than that obtained with conventional methods.

This advancement represents a diagnostic support tool that could guide medical decisions regarding the necessity and type of surgery required to confirm the diagnosis, as commented by Carles Majós, a clinical neuroradiologist at Bellvitge. The published results pave the way for further development and validation of DISCERN with more patients for implementation in clinical practice.

Article written by SYNC