During the final year of my Master's, I had the opportunity to work as a student researcher at the SAMPL Lab (Signal Acquisition, Modeling, Processing, and Learning) under the direction of Professor Yonina C. Eldar. The lab specializes in signal processing, with a growing focus on integrating artificial intelligence across fields such as computer vision and signal processing using AI techniques. Over the course of a year, I initially joined as a student worker and later transitioned into a more formal student position. During my time at SAMPL, I was fortunate to work on a specific project that eventually became the basis for my Master's thesis. This experience allowed me to explore cutting-edge research in both signal processing and AI, significantly contributing to the development of my thesis work.
The Signal Acquisition, Modeling, Processing, and Learning (SAMPL) Lab, led by Prof. Yonina C. Eldar, focuses on developing cutting-edge technologies to enhance the extraction and processing of signals and information across various domains, including medical imaging, radar, communications, scientific and optical imaging, and biological inference. The lab is at the forefront of developing model-based artificial intelligence (AI) methods designed to maximize the amount of information extracted with minimal resources.
Signal processing, the core of SAMPL Lab’s research, is the science of generating, acquiring, transmitting, and analyzing signals and information through mathematical theory and methods. It plays a crucial role in modern technologies, enabling innovations like smartphones, autonomous vehicles, and medical and defense systems. The lab’s research addresses the challenge of improving signal sampling and processing efficiency, which is particularly critical in applications such as medical diagnostics, aviation, and automotive systems.
SAMPL Lab introduces a revolutionary approach by combining signal sampling and processing, a shift from traditional systems that treat these as separate stages. By designing them jointly, SAMPL Lab leverages the inherent properties of signals and tasks during the sampling process, enabling the acquisition and processing of only the necessary information. This innovation allows for significant reductions in sampling and processing rates, well below the Nyquist rate, while improving the resolution of signals in time, space, and frequency.
This innovative approach has the potential to transform technologies such as wireless ultrasound, portable imaging devices, fast MRI scans, high-resolution radar, efficient communication systems, and even super-resolution microscopy. SAMPL Lab’s research also supports the development of efficient and interpretable deep learning networks for medical imaging, radar-based medical sensing, and more.
To achieve these advances, SAMPL Lab blends theoretical research in mathematics, information theory, statistical signal processing, AI, and computer science with practical engineering research. The lab's state-of-the-art facilities enable the transition from theoretical models to prototype systems and clinical studies. Through collaborations with physicians, industry partners, and scientists in fields such as biology and physics, SAMPL Lab works to advance healthcare technologies, medical diagnostics, and scientific discovery.
Prof. Yonina C. Eldar is a leading figure in the field of signal processing and artificial intelligence, renowned for her groundbreaking work in medical imaging, communications, and radar systems. As a professor at the Weizmann Institute of Science, she has significantly advanced the development of novel algorithms for efficient signal acquisition, processing, and learning. Her research has led to innovations in areas like wireless ultrasound, MRI, and super-resolution imaging, transforming how we interact with and understand complex data. Prof. Eldar is also a prolific mentor, guiding students and researchers to achieve new heights in both theoretical and applied signal processing. Her work has had a profound impact on numerous industries, from healthcare to defense, making her a leading authority in her field.
During this year, I worked on two closely related projects. The first focuses on the quantification of liver fat using ultrasound channel data, and the second involves the detection of pleural effusion in the lungs using the same ultrasound channel data. In collaboration with NYU and Haemek Hospital, we gathered a small amount of raw ultrasound data that will be used for the project. Traditionally, these detections are performed using MRI, which is costly and risky, especially when repeated verifications are needed. The aim of the project is to explore three different approaches for detecting these two conditions using more affordable and non-invasive ultrasound technology. These projects combine AI applied to signal processing and computer vision, with the first part focusing on enhancing ultrasound-generated images through AI models to highlight relevant areas for detection. The other two parts explore direct approaches on the channel data, including feature extraction and learning directly from raw data using transformer and RNN models.
During my work at the SAMPL Lab at the Weizmann Institute of Science, alongside my thesis, I had the opportunity to lead a project focused on automated organ segmentation in ultrasound images of mouse fetuses. The goal was to develop models capable of automatically outlining hearts, livers, and placentas in ultrasound scans, generating precise Region of Interest (ROI) files to assist researchers in their analyses. Using a U-Net-based architecture, the models achieved high accuracy, particularly for hearts and livers, while addressing challenges such as variable organ visibility and complex shapes. This project significantly reduces manual annotation efforts and provides a reliable tool for advancing developmental biology research.