Extensible framework for fast deformable matching of very large imaging data
Active Grant Awarded
- Award ID
- R01EB038985
- Recipient
-
The Trustees Of The University Of Pennsylvania UEI GM1XX56LEP58
- Awarding agency
-
National Institute of Biomedical Imaging and Bioenginering
National Institutes of Health, Department of Health and Human Services
- Obligated
- $697,439
Award facts
- Assistance listing
- 93.286 Discovery and Applied Research for Technological Innovations to Improve Human Health
- Completion date 5% complete
- Place of performance
- Philadelphia, PA 19104, USA
Award details come from USAspending.gov.
About this award
The Department of Health and Human Services' National Institutes of Health, through its National Institute of Biomedical Imaging and Bioengineering, awarded The Trustees of the University of Pennsylvania (doing business as University of Pennsylvania) $697,439 on September 9, 2026, under Assistance Listing 93.286, Discovery and Applied Research for Technological Innovations to Improve Human Health. The award funds Grant R01EB038985, which proposes FIREANTS, an extensible, scalable framework for deformable image registration. The framework combines the robustness of classical optimization with the speed of deep learning to register heterogeneous, high-resolution datasets spanning MRI, histology, and spatial multi-omics. The proposed work includes mixed-precision fused CUDA kernels. The source excerpt ends before describing the remaining technical approach.
The obligated amount is $697,439, and the ultimate completion date is May 31, 2030. The place of performance is Philadelphia, PA 19104. The source does not name sub-recipients for this award. The program supports hypothesis-, design-, technology-, or device-driven research on biomedical imaging and bioengineering technologies.
Description
- Abstract
- Image registration is a fundamental task in quantitative biomedical image analyses, enabling researchers to compare, integrate, and interpret imaging data across subjects, time points, modalities, and experimental conditions. Imaging data continues to grow both in scale and in the breadth of different technologies for their acquisition, as clinical and biomedical research increasingly embraces high-resolution imaging and large-scale data collection studies, from population-level magnetic resonance imaging to cellular-resolution spatial omics modalities. Although current image registration methodologies continue to play an essential role in processing these data, many of the existing frameworks were not originally optimized for leveraging the full potential and complexity of modern acquisitions. As a result, off-the-shelf tools are often inadequate when applied to heterogeneous, high-resolution datasets, where biological variability, imaging artifacts, and extreme differences in local morphology stress the limits of current image registration paradigms. Simultaneously, multi-modal analyses which integrate MRI, histology, and spatial multi-omics require registration frameworks that can operate effectively across vast differences in spatial resolution, signal contrast, and data structure. Existing methods struggle to meet these demands as optimization-based tools are often too slow, while deep-learning models require costly training, depend heavily on training data distribution, and lack the flexibility to generalize across tasks without retraining. This project directly addresses these critical gaps by proposing fireants, an extensible, scalable framework for deformable image registration that combines the robustness and adaptability of classical optimization with the speed and representational power of modern deep learning. Specifically, we propose mixed-precision fused cuda kernels and a distributed implementation to achieve memory-efficient and scalable optimization; incorporate advanced adaptive optimization techniques for riemannian manifolds to handle high-dimensional diffeomorphic deformations; enable end-to-end differentiable pipelines to learn multi-scale, registration-aware feature representations and deep metrics that improve alignment accuracy under challenging heterogeneity; and evaluate these developments through their application to real-world use cases and their practical integration with major community resources. These innovations will enable the biomedical research community to move beyond existing alignment pipelines, accelerate large-scale analysis workflows, and unlock quantitative insights across a broad and varied range of biological and medical applications.
Quick answers
Who received award R01EB038985?
R01EB038985 was awarded to The Trustees Of The University Of Pennsylvania by National Institute of Biomedical Imaging and Bioenginering (National Institutes of Health) on September 9, 2026.
How much is award R01EB038985 worth?
$697,439 has been obligated on R01EB038985.
On GovTribe
See this award's full record
- Every transaction and subaward
- The recipient's other awards and teaming partners
- Similar grant awards
- Ask GovTribe AI about this award