Brain Metastases: How Artificial Intelligence Chooses the Most Effective Therapies

More encouraging news on the use of artificial intelligence in medicine. This time it comes from a clinical study conducted by the Istituto Clinico Humanitas Irccs in Rozzano (Milan) and the University Hospital of Tubinga (Germany), coordinated by Letterio Politi and Marta Scorsetti , respectively responsible for neuroradiology and radiotherapy and radiosurgery at Humanitas. In their work, published in the journal Neuro Oncology , the researchers “taught” artificial intelligence to analyze brain metastases after radiotherapy treatment and to distinguish between lesions due to tumor progression and those caused by the treatment itself.

The results obtained are important, satisfactory and necessary, because at the moment conventional diagnostic techniques, including magnetic resonance imaging, often fail the objective because the characteristics of the two types of tissue are very similar. The instrument developed by clinicians, however, has shown excellent reliability in the operation of discerning between radionecrosis and tumor progression.

Promising results

In particular, the researchers isolated 124 brain lesions in patients undergoing stereotactic radiosurgery, a high-precision radiotherapy technique. The cause of these lesions was already known and the information was fed to an algorithm based on deep learning models and artificial neural networks. In this way, the system “learned” to distinguish the two types of lesions and, subsequently, tested on an external group of patients, it showed that it could do so with high accuracy, which allowed this approach to be validated and the results to be generalized.

“Stereotactic radiosurgery is an effective technique for treating brain metastases,” Scorsetti said. “However, distinguishing between radiation necrosis, a side effect of treatment, and tumor progression can be difficult. Our research shows that artificial intelligence has the potential to provide a more accurate diagnostic tool, potentially reducing the need for biopsies or exploratory surgeries.”

Towards broader studies

Now, say the authors of the work, it is time to take the next step, repeating the experiment with a larger number of patients. “Although further research will be needed to validate these models in larger populations, the results obtained are promising and suggest that artificial intelligence could become a valuable tool to support the work of neuroradiologists of today and tomorrow - explains Politi, who also directs the master's degree program in Data analytics and artificial intelligence in health sciences (Daihs), born from the collaboration between Humanitas and Bocconi University to train future professionals on these cutting-edge topics -. The research represents an important step forward in the field of neuro-oncology and a further demonstration of how advanced technologies and multidisciplinary work can combine to improve our diagnostic and treatment capabilities, moving ever closer to precision medicine approaches”.