Research database

Stroke is a leading cause of death and disability in the European Union. The use of microwave imaging (MWI) for brain stroke diagnosis has attracted growing interest in the past decade thanks to its non-invasiveness and low costs. Furthermore, thanks to its portability, brain MWI holds significant promise for widespread adoption in-ambulance, dramatically reducing intervention time. Data calibration still holds a strong limit to this application, where the need for a physical phantom of the patient head is the main bottleneck. I2I-StrokeCal will develop and validate a novel method to perform data calibration relying solely on the patient measurement, thus without additional phantom measurements. This step-change will be attained through 3 specific objectives: (O1) development of the first dataset of brain MWI data in the literature; (O2) training and test of an image-to-image translation (I2I) tool which calibrates synthetically generated MWI data; (O3) training and test of the final I2I model which calibrates experimental MWI data. The proposed methodology applies cutting-edge computer vision techniques (I2I) to address a MWI problem (data calibration) applied to the medical field (stroke diagnosis). The main host will be POLITO, home to the most advanced brain MWI prototype (TRL-4) within academia. The fellowship will also include two 4-months secondments. The first will be at the University of Manitoba (Canada), where the Researcher will be trained on the integration between machine learning and inverse scattering. The second will be at the École Polytechnique (France), to gain experience on machine learning and computer vision. The results will then be exploited during a 6-months non-academic placemt at UBT S.r.l, an Italian startup that is conducting clinical trials with a TRL-6 brain MWI prototype. The acquired skills will shape the Researcher into a mature, independent scientist with interdisciplinary expertise and international recognition in medical MWI.