A groundbreaking AI-based diagnostic tool, named Diagnosis in Susceptibility Contrast Enhancing Regions for Neuroncology (DISCERN), has been developed by the Vall d’Hebron Institute of Oncology’s (VHIO) Radiomics Group in collaboration with the Bellvitge University Hospital’s Neuroradiology Unit. This non-invasive tool utilizes deep learning technology to provide precise brain tumor diagnoses, surpassing current methods.
The DISCERN tool is based on training patterns using artificial intelligence models from standard magnetic resonance imaging (MRI) data. Published in Cell Reports Medicine, a study led by VHIO demonstrates the efficacy and accuracy of DISCERN in diagnosing brain tumors from perfusion MRI, outperforming traditional diagnostic techniques. Glioblastoma multiforme, brain metastasis from solid tumors, and primary central nervous system lymphoma are responsible for a significant portion of malignant brain cancers, highlighting the importance of accurate differential diagnosis among these malignancies due to varying treatment approaches.
Currently, the non-invasive differential diagnosis of brain tumors relies on assessing MRI scans before and after contrast administration. However, definitive diagnoses often necessitate invasive neurosurgical procedures, impacting patient quality of life. Raquel Perez-Lopez, Head of VHIO’s Radiomics Group and corresponding author of the study, emphasized the critical clinical need for more advanced diagnostic tools to avoid unnecessary interventions.
Over five years of research focused on identifying innovative MRI perfusion imaging biomarkers have culminated in the development of DISCERN. This tool leverages insights from previous AI research projects to automate presurgical diagnostic classification with high precision. Albert Pons-Escoda, a Clinical Neuroradiologist and Investigator at Bellvitge University Hospital, highlighted the user-friendly interface of DISCERN, making it applicable for clinical use.
Deep learning algorithms within DISCERN analyze spatial and temporal information from MRI scans to identify unique imaging patterns specific to each tumor type. By training the tool on data from previously diagnosed patients, DISCERN can accurately distinguish between different brain malignancies. Alonso García-Ruiz, a Ph.D. Student of VHIO’s Radiomics Group and the study’s first author, explained that the system learns distinctive features of tumors, similar to identifying characteristics of different animal species, enabling accurate tumor classification.
The study validated DISCERN on over 500 additional cases, achieving an impressive 78% accuracy rate in tumor classification, surpassing conventional methods. Carlos Majós, Clinical Neuroradiologist and Investigator at Bellvitge University Hospital, highlighted that DISCERN supports medical decision-making by providing essential information on the need for surgery to confirm diagnoses.
The user-friendly DISCERN application, developed as part of the study in collaboration with Bellvitge Biomedical Research Institute (IDIBELL) and Clínic Hospital in Barcelona, facilitates brain tumor classification and aims to accelerate its adoption in clinical settings. The innovative tool holds promise for enhancing the accuracy and efficiency of brain tumor diagnosis, ultimately improving patient care.
