Adaptive computing speeds up medical and astronomy image processing
Adaptive Computation with Conformal Certification
Machine Learning
Summary
Big scientific images like those from medical scans or telescopes are much larger and harder to process than regular photos. The authors found a way to speed up analyzing these huge images by focusing more computing power on important areas, like tumors or stars, while using less on unimportant background. Their method also knows how sure it is about its results, so it won’t miss key details. They tested it on different types of scientific data and showed it can be much faster without losing accuracy.
What this means in practice
- •For medical imaging teams: Accelerate segmentation of large 3D medical scans by focusing computation on critical areas with certified uncertainty control.
- •For astronomy data analysts: Enable faster prediction of astronomical signals like galaxy brightness by adaptively computing dense regions and certifying prediction accuracy.
Authors
Shehtab Zaman, Aneesh Rangnekar, Nancy Lee, Harini Veeraraghavan
Abstract
Scientific imaging data involving medical (e.g., 3D radiographic, microscopic images) and astronomical sensors pose different challenges from natural images. One key challenge is that these datasets are often orders of magnitude larger than natural images. Efficient training and inference on such data is challenging due to the increased hardware requirements, often needing bespoke solutions. Moreover, the relevant objects such as tumors and astronomical phenomena are sparsely distributed with respect to the background information. Standard efficiency techniques like quantization and pruning apply uniform scaling of all regions, potentially compromising the precision/fidelity of predictions. Thus, efficient methods that reduce computational requirements without losing resolution for relevant parts of the data and with uncertainty estimates on predictions are essential. We present \textit{Conformal Adaptation} with risk guarantees for uncertainty-aware sparse computations. Our formalism is general, spanning dense pixel-level classification and regression tasks where hierarchical models predict at low resolution, certify the signal using conformal risk control, and refine the certified predictions. We benchmark our method on multiple tasks such as 3D medical image segmentation, 2D galaxy flux prediction, and 1D calcium spike rate prediction, achieving $2.5$--$72\times$ wall-clock speedup across the various tasks, including on commodity CPUs.