Professors Stephen Bach, Stefanie Tellex, and Ugur Çetintemel of Brown CS, along with their Brown University collaborators below, have just received Seed Awards from Brown's Office of the Vice President for Research (OVPR) to help them compete more successfully for large-scale, interdisciplinary, multi-investigator grants. They join numerous previous Brown CS recipients of OVPR Seed Awards, including (most recently) Carsten Eickhoff, Daniel Ritchie, Stefanie Tellex, and James Tompkin.
Stephen's research is a collaboration with Primary Investigator Professor Stephon Alexander (Physics) as well as Professor Ian Dell’Antonio (Physics), Professor Richard Gaitskell (Physics), and Assistant Professor Jonathan Pober (Physics).
"As more and more analyses depend on machine learning," they write, "physicists need to understand the biases induced in their models by the training data they use. As outlined in the Proposal, in many situations they face a tradeoff between two types of data: (1) abundant but low-resolution, synthetic, or otherwise approximate data, and (2) scarce but high-resolution, high-fidelity, or otherwise more realistic data. The ideal would be to correct the biases in the abundant data so that it is more like the scarce data in the ways that affect the predictions of the machine learning models trained on that data. If done, then scientists would have access to higher quality, abundant data. The problem is that, even though it is easy to tell that models trained on each type of data disagree in their predictions, it is hard to tell which aspects of the input are leading to that disagreement. Our proposal is to create a software framework for identifying these key causes of differences."
Ugur's research is a collaboration with Assistant Professor Harrison Bai (Diagnostic Imaging).
"Stroke is a leading cause of long-term disability," they write, "and outcome in regaining functionality in areas supplied by anterior circulation large vessel is directly related to timely endovascular therapy (EVT). However, not all patients benefit from rapid intervention. CT perfusion is widely recognized as the selection tool to identify patients who will most likely benefit from reperfusion based on stroke core and penumbra size estimation as well as mismatch quantification. However, it is not routinely performed at many institutions in the United States and around the world. In this proposal, we propose to develop a fully automated artificial intelligence (AI) pipeline that identifies the images/series of interest, detect emergent large vessel occlusion and predicts immediate (e.g. the Thrombolysis in Cerebral Infarction [TICI] score) and functional (e.g. modified Rankin score [mRS]) outcomes from EVT based on pre-procedure CT angiography. We will establish an end-to-end AI platform that interfaces with the Rhode Island Hospital (RIH) Picture Archiving and Communications Systems for real-time clinical use. The ability to predict immediate outcomes of EVT will affect management because proceduralists will be able to anticipate different reperfusion based on these predictions and adjust their treatment approach accordingly, while prediction of functional outcome assists in patient selection. We anticipate that the proposed project will further collaboration between the Department of Computer Science and the Department of Diagnostic Imaging, which is crucial in advancing Brown University's position in research on AI, machine learning and computer vision applied to the healthcare system and medical imaging."
Stefanie's research is a collaboration with Senior Lecturer Diane H. Silva Pimentel (Education).
"The aim of this proposal," they write, "is to test the hypothesis that we can prepare high school teachers to teach students about autonomous aerial robots on their own, at scale by 1) providing a project-based curriculum targeted at the high school level on an open-source low-cost autonomous robot with few infrastructure requirements; 2) providing remote professional development workshops for teachers; and 3) pairing teachers with Brown students familiar with the curriculum who will provide help and technical support. We will study the interactions between teachers and curriculum materials, tools, and Brown students who facilitate the learners’ conceptual development, and what characteristics our online PD and remote support give rise to these interactions. Our work will assess each of these three interventions by assessing teacher content knowledge as well as self-efficacy. We will also assess the effectiveness with which we can engage students in both urban and suburban districts through hands-on remote learning curricula that emphasize physical hardware, in the hands of the students as well as via remote laboratories. This work has the potential to directly benefit students in Rhode Island, consistent with President Paxson’s commitment to Providence public schools. Moreover this funding will provide critical preliminary work enabling us to apply for follow-on funding for larger expansions from NSF and industry resources to grow our project to a nationwide and international effort to teach students about autonomous robotics."
A full list of awardees is available here.
For more information, please click the link that follows to contact Brown CS Communication Outreach Specialist Jesse C. Polhemus.