2017-057 - Interpretable Deep Learning Framework for Predictive Modeling

Description:
  • Predictive modeling
  • Data mining
  • Personalized Healthcare

Abstract

USC researchers have developed a novel knowledge-distillation approach called Interpretable Mimic Learning, to learn interpretable features for making robust prediction while mimicking the performance of deep learning models. This approach provides interpretable models that help primary care providers, physicians and clinical experts in monitoring and decision-making for patient care. The technology can be successfully applied not only to healthcare, but also to other applications such as speech processing, computer vision, finance or marketing.

Benefit

  • Achieves or exceeds state-of-the-art prediction performance

  • Provides interpretable features and decision rules

  • Phenotype discovery for clinical decision making

  • Better quality of patient care

  • Faster adoptability among clinical staff

Market Application

Exponential growth in Electronic Healthcare Records (EHR) has resulted in the urgent need for discovery of meaningful data-driven representations and patterns of diseases in Computational Phenotyping research. Deep Learning models have shown superior performance for robust prediction in computational phenotyping tasks, but suffer from the issue of model interpretability which is crucial for clinicians involved in decision-making.

Publications

Distilling Knowledge from Deep Networks with Applications to Healthcare Domain

Other

  • Experimentally validated

  • Available for exclusive and non-exclusive license

Patent Information:

  • Title: INTERPRETABLE DEEP LEARNING FRAMEWORK FOR MINING AND PREDICTIVE MODELING OF HEALTH CARE DATA
  • App Type: Utility
  • Country: United States
  • Serial No.: 15/829,768
  • Patent No.: 11,144,825
  • File Date: 12/1/2017
  • Issued Date: 10/12/2021
  • Expire Date: 8/12/2040
  • Patent Status: Patent Issued