Wednesday, 30 June 2021

Artificial intelligence for the diagnosis of head and neck cancer

Versione stampabile

The University of Trento and the Provincial Health Care Services -APSS continue to collaborate, this time on the diagnosis of serious conditions. In particular, researchers are investigating the use of artificial intelligence in the diagnosis of head (gliomas and meningiomas) and neck tumors. The goal, as usual, is to use the latest technology to make the diagnostic process more precise, faster, less invasive, to plan therapies in the best possible way and, consequently, to improve the health and life expectancy of patients.

This new area of collaboration involves, on one side, the University of Trento, with a multidisciplinary group of researchers from different departments and, on the other, the Provincial Health Care Services with the staff of the Neuroradiology department. 

Artificial intelligence in cancer diagnosis
Artificial intelligence has an advantage over conventional diagnostic procedures. With AI based analysis techniques, in particular, researchers have access quickly and without invasive procedures to quantitative and qualitative information about the tumor that they could not obtain using visual observation; algorithms convert medical images obtained from MRI, CT and PET scans into numerical data. The amount of data that must be analyzed is huge: that is why this is the ideal environment to apply AI statistical techniques. Using IT, the researchers obtain information on the morphological characteristics of the tumor (shape, volume, tissue) which can be associated with its molecular and genomic characteristics. With this information, they will be able to better assess the clinical risk and the aggressiveness of the tumor and, as a consequence, to choose the most suitable treatment.

The research work on AI diagnostics will initially involve some brain (gliomas and meningiomas) and neck tumors. Early diagnosis of these types of cancer is crucial: if they are treated in the early stages, when they are small, there are good chances of recovery. But if the tumor is large the outlook for patients is worse, and the treatment will be longer and more invasive. AI thus helps neuroradiologists in the diagnosis, and oncologists and neurosurgeons in the planning of surgeries. The same technology will then be used to follow up on patients in the later stages through non-invasive devices that monitor vital parameters.

The UniTrento-APSS collaboration
The test phase involves the Neuroradiology Unit of APSS, which in recent months has been working on the early characterization of molecular patterns in gliomas through the analysis of magnetic resonance images and radiomics applications. The project is coordinated by Paola Feraco, neuroradiologist at APSS, and also involves the University of Bologna. 

The analysis of data is conducted by a research team of the University of Trento, coordinated by Massimo Donelli, professor of Electromagnetic fields at the Department of Information Engineering and Computer Science, and Giuseppe Espa, full professor of Economic statistics at the Department of Economics and Management. Latest generation software will be used, which allow the complete management and control of the whole process even by users who don't have advanced statistical skills.

"The analysis of this type of data may also be useful to train new medical students as well as future postgraduates in radiology - Feraco explains. Through a more modern approach, it will allow them to have a global and multidisciplinary vision of oncological pathologies, but also of other conditions. And this will lead to the development of an approach that takes into account the characteristics of patients and treatments tailored to their needs".

It is the newly established Center of Security and Crime Sciences (CSSC) of the University of Trento, which has just started to be operational, to come up with idea of applying new technologies to diagnostics. "The collaboration with the Provincial Health Care Services is based on teamwork and different skills. Alongside state-of-the-art techniques and tools in the area of statistics, there are other techniques and tools in biomedical imaging analysis, skills in augmented reality, virtual reality and information technologies", comments Giuseppe Espa. "Recently, we have seen a lot of interest and investments in innovation in medicine. The University of Trento is investing a lot in research on health and medical risks, because understanding and managing these risks requires a multidisciplinary approach that our University can ensure, with opportunities for further study and research coming from the collaboration with the University of Verona".

"The project started only a few months ago, but we are already analyzing a very large database, to the benefit of early diagnoses - adds Massimo Donelli. We use and process the collected to make predictions and better characterize a particular type of tumor. In this way, we are able to obtain and provide doctors with very accurate information, including on the evolution of the disease, and to suggest the best treatment options. The idea of applying machine learning for personalized medicine is the result of discussions with medical staff, when we put together our skills to solve new problems in the medical-health sector. In fact, at the moment there are no protocols for this type of diagnosis: our goal is to develop one thanks to the collaboration with Apss".
The work on the diagnosis of these tumors is just one of the collaborations that involve the University and Apss. They are working on the support to the diagnosis of type 2 diabetes through a galvanic skin resistance sensor.