Distributed Stratified Locality Sensitive Hashing for Critical Event Prediction in the Cloud
This work addresses the need for efficient tools in data-driven medicine, specifically for critical event prediction in healthcare, but it is incremental as it builds on existing hashing methods with optimizations for cloud environments.
The paper tackled the problem of fast similarity-based prediction on large medical waveform datasets by introducing a distributed system for Stratified Locality Sensitive Hashing, achieving a 21x speedup in comparisons with a 10% loss in Matthews correlation coefficient on a dataset of 1.37 million points.
The availability of massive healthcare data repositories calls for efficient tools for data-driven medicine. We introduce a distributed system for Stratified Locality Sensitive Hashing to perform fast similarity-based prediction on large medical waveform datasets. Our implementation, for an ICU use case, prioritizes latency over throughput and is targeted at a cloud environment. We demonstrate our system on Acute Hypotensive Episode prediction from Arterial Blood Pressure waveforms. On a dataset of $1.37$ million points, we show scaling up to $40$ processors and a $21\times$ speedup in number of comparisons to parallel exhaustive search at the price of a $10\%$ Matthews correlation coefficient (MCC) loss. Furthermore, if additional MCC loss can be tolerated, our system achieves speedups up to two orders of magnitude.