AI and Tech for Medicine
Artificial Intelligence for Medicine
One of our main foci of interest is to harness innovative AI methods, in combination with unique signal processing tools, to address real-world health challenges. Through close collaboration with radiologists and clinicians both in Israel and abroad, we identify unmet clinical needs and engage in clinical projects designed to enhance early detection of diseases, reduce diagnostic errors, support physicians in their decision-making, and create imaging devices with improved quality.
Our research topics include
- Multimodal deep learning: combining the information obtained from different imaging modalities in the interest of better diagnosis. For example, physicians use various imaging modalities (mammography, ultrasound, MRI) for the diagnosis of breast cancer. In our research, we try to integrate the information obtained from these different modalities in order to improve the diagnosis.
- Use of AI methods for the analysis of ultrasound “channel data” (the pre-beamformed RF data received at the ultrasound machine): in attempt to extract important tissue properties that would support disease diagnosis and assessment, e.g., help determine whether lesions are benign or malignant or help quantify liver fat.
- Use of AI for conversion between imaging modalities: e.g., using deep learning to convert ultrasound images to CT images synthetically.
- AI-guided ultrasound image acquisition: with the goal of overcoming ultrasound’s operator-dependency, enabling even non-expert sonographers to perform high-quality scans.
- Deep learning for super-resolution vascular ultrasound imaging: applying model-based deep learning methods for the processing of data received from contrast-enhanced ultrasound, in order to create vascular reconstructions with high resolution.
- Use of AI for Covid-19 diagnosis and prediction of outcome: in order to support the fight against the global pandemic.
- D. Keidar et. al, “COVID-19 Classification of X-ray Images Using Deep Neural Networks”, European Radiology, pp. 1-10, May 2021.
- O. Frank et. al, "Integrating Domain Knowledge into Deep Networks for Lung Ultrasound with Applications to COVID-19", IEEE Transaction on Medical Imaging, vol. 41, issue 3, pp. 571-581, March 2022.
- A movie illustrating our work in the Covid-19 domain.
O. Bar-Shira, A. Grubstein, Y. Rapson, D. Suhami, E. Atar , K. Peri-Hanania, R. Rosen, Y. C. Eldar, "Learned Super Resolution Ultrasound for Improved Breast Lesion Characterization", MICCAI 2021.
B. Luijten, R. Cohen, F. J. de Bruijn, H. A. W. Schmeitz, M. Mischi, Y. C. Eldar, and R. J. G. Van Sloun, "Adaptive Ultrasound Beamforming Using Deep Learning", IEEE Transactions on Medical Imaging, vol. 39, issue 12, pp. 3967-3978, December 2020.
R. J. G. van Sloun, R. Cohen, Y. C. Eldar, "Deep Learning in Ultrasound Imaging", Proceedings of the IEEE, vol. 108, issue 1, pp. 11-29, January 2020.
Radar for Medical Applications
In the last decade, small, sophisticated, and robust millimeter-wavelength (mm-Wave) radar systems have evolved, which opens the door to emerging new and exciting applications. The healthcare sector is one of the biggest beneficiaries of this development. An essential application of these radar systems is remote monitoring of vital signs, without the need for wired connections that produce discomfort or irritations, which are affected by the manner of contact and can be easily detached. Today, monitoring devices in clinics and hospitals are connected to patients by medical staff, by an interaction that consumes valuable time from both staff and patients and increases the risk of transmission of infectious diseases, especially in times of pandemics, such as COVID-19. This technology is ideal in these situations since radar systems do not require users to wear, carry, or interact with any additional electronic device. We develop efficient methods for heart rate and respiration monitoring using a minimal number of samples and bits. The purpose of our research is to allow efficient remote vital sign monitoring of multiple people simultaneously, for real-time implementation.