Making an Impact on International Women’s Day
At 41 percent, less than half of the world’s researchers are women. At the MIT Jameel Clinic we are proud to support women who are creating a healthier world through groundbreaking research. This academic year, there are 12 women in total whose research projects in AI and health are being supported by the Jameel Clinic. Below is a list of their most cited work, some of which eventually led them to conduct their current research at the intersection of in AI and health.
Regina Barzilay
A Deep Learning Approach to Antibiotic Discovery (2020), Cell
“The researchers say the antibiotic, called halicin, is the first discovered with artificial intelligence (AI). Although AI has been used to aid parts of the antibiotic-discovery process before, they say that this is the first time it has identified completely new kinds of antibiotic from scratch, without using any previous human assumptions.” Nature News, 2020
Fotini Christia
Alliance Formation in Civil Wars (2012), Cambridge University Press
“If ethnicity, religion, and other markers of identity didn’t matter to warlords, Christia asked, what did? It turns out the answer was simple: power. After studying the cases of Afghanistan, Bosnia, and Iraq in intricate detail, Christia built a database of 53 conflicts to test whether her theory applied more widely. She ran regression analyses and showed that it did: Warlords adjusted their loyalties opportunistically, always angling for the best slice of the future government.” Boston Globe, 2013
Marzyeh Ghassemi
“Across the world’s coronavirus disease 2019 (COVID-19) hot spots, the need to streamline patient diagnosis and management has become more pressing than ever. As one of the main imaging tools, chest X-rays (CXRs) are common, fast, non-invasive, relatively cheap, and potentially bedside to monitor the progression of the disease. This paper describes the first public COVID-19 image data collection as well as a preliminary exploration of possible use cases for the data.” MELBA Journal, 2020
Polina Golland
This paper co-authored by Golland presents the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS), which evaluated 20 state-of-the-art tumor segmentation algorithms using 130 MR scans, including both real and simulated glioma cases. The results highlighted significant variability among human raters, underscoring the challenge of tumor sub-region segmentation. While no single algorithm performed best across all sub-regions, combining multiple strong algorithms via a hierarchical majority vote consistently outperformed individual methods. The BRATS dataset remains publicly available as a benchmarking resource for ongoing advancements in brain tumor segmentation.
Dina Katabi
XORs in the air: practical wireless network coding (2006), SIGCOMM
“Network coding is a new communication paradigm which allows network nodes not only to forward packets but also to combine them together to reduce the number of transmissions. COPE has been the first network coding scheme designed for unicast traffic in wireless networks.” International Conference on Advanced Technologies for Communications, 2015
Muriel Medard
An algebraic approach to network coding (2003), IEEE/ACM Transactions on Networking
This work demonstrates that network coding is essential to achieving network capacity. Extending previous research on multicast networks, the study develops a broader framework for arbitrary networks and robust communication. Medard’s paper establishes the necessary and sufficient conditions for linear network coding feasibility and introducing coding strategies that enable maximally robust networks, even in the face of link failures, without requiring network adaptation.
Georgia Perakis
“Governments often offer subsidies to consumers for clean-technology products, from home solar panels to electric vehicles. But what are the right levels of subsidy, and how should they be calculated? As a new paper co-authored by MIT researchers shows, governments can easily make subsidies too low when they ignore a basic problem: Consumer demand for these products is usually highly uncertain.” MIT News, 2016
Rosalind Picard
Affective Computing (2000), MIT Press
“Technology has become more physically and psychologically intimate, which has created a demand for new technologies that can infer emotional states from humans. The term ‘affective computing’ was coined in 1995 by Professor Rosalind Picard, founder and director of the affective computing research group…She recognized the extent to which emotions governed our lives and decided to drive forward the concept of ‘engineering emotion.’” IEEE Spectrum, 2019
Loza Tadesse
Nanophotonic Platforms for Chiral Sensing and Separation (2020), Accounts of Chemical Research
Chirality plays a crucial role in molecular interactions, impacting fields from medicine to agriculture, yet detecting and separating enantiomers remains challenging. Recent advances in nanophotonics offer promising solutions by leveraging nanoscale chiral light–matter interactions to enhance chiral sensing, spectroscopy, and enantioselective photochemistry. This research highlights how engineered nanostructures, such as plasmonic and dielectric nanoparticles, can amplify optical chirality density, enabling more efficient enantiomer detection, manipulation, and separation—advancements that could revolutionize disease diagnostics, pharmaceuticals, and agrochemicals.
Caroline Uhler
This study, co-authored by Uhler, extracted the mitochondrial DNA from bones in stringently sanitized clean rooms before applying statistical techniques to weed out any remaining genetic contaminants. After these methods were applied to three Neanderthal bones discovered across Europe, the entire Neanderthal genome was successfully sequenced. Live Science, 2022
Ashia Wilson
The Marginal Value of Adaptive Gradient Methods in Machine Learning (2017), NeurIPS
This paper examines the impact of adaptive optimization methods like AdaGrad, RMSProp, and Adam on deep neural network training, revealing that they often lead to solutions that generalize worse than those found by gradient descent (GD) or stochastic gradient descent (SGD). Through both theoretical analysis and empirical evaluations, the researchers demonstrate that while adaptive methods may achieve better training performance, they can significantly underperform in test accuracy, suggesting a need for caution when using them in practice.
Sixian You
“A new microscope system can image living tissue in real time and in molecular detail, without any chemicals or dyes…The system uses precisely tailored pulses of light to simultaneously image with multiple wavelengths. This enables the researchers to study concurrent processes within cells and tissue, and could give cancer researchers a new tool for tracking tumor progression and physicians new technology for tissue pathology and diagnostics.” Phys.org, 2018
