TRACER: A Framework for Facilitating Accurate and Interpretable Analytics for High Stakes Applications

Project Reference :

AISG-100E-2018-007

Institution :

National University of Singapore (NUS)

Principal Investigator :

Professor Ooi Beng Chin

Technology Readiness :

5 (Technology validated in relevant environment)

Technology Categories :

AI- Predictive Analytics

Background/Problem Statement

Artificial intelligence in the healthcare market is projected to grow from USD 6.9 billion in 2021 to USD 67.4 billion by 2027, with a CAGR of 46.2% from 2021 to 2027.

The key factors fueling the growth of the market include market influx of large and complex healthcare datasets, growing need to reduce healthcare costs, improving computing power and declining hardware cost, rising number of partnerships and collaborations among different domains in healthcare sector, and surging need for improvised healthcare services due to imbalance between health workforce and patients. Additionally, growing potential of AI-based tools for elderly care, increasing focus on developing human-aware AI systems, and rising potential of AI technology in genomics, drug discovery, and imaging & diagnostics is expected to create a growth opportunity for artificial intelligence in the healthcare market.

In high stakes applications such as healthcare and finance analytics, the interpretability of predictive models is required and necessary for domain practitioners to trust the predictions. Traditional machine learning models e.g. logistic regression (LR) are easy to interpret in nature. However, many of these models aggregate time-series data without considering the temporal correlations and variations. Therefore, their performance cannot match up to recurrent neural network (RNN) based models. Although effective in modeling time-series data, the lack of interpretability tends to become an obstacle for deploying RNN-based models in healthcare analytics. An accurate and interpretable analytic model can help healthcare practitioners to make effective and responsive decisions on patient management and resource allocation.

Solution

InTerpRetAble Clinical dEcision SuppoRt (TRACER) is a novel and general framework to facilitate accurate and interpretable predictions which uses a novel model which captures both Time-Invariant and Time-Variant (TITV) feature importance. It is devised for healthcare analytics and other high stakes applications such as financial investment and risk management. 

Different from LR and other existing RNN-based models, TITV is designed to capture both the time-invariant and the time-variant feature importance. It uses a feature-wise transformation subnetwork and a self-attention subnetwork for the feature influence shared over the entire time series and the time-related importance respectively. Healthcare analytics is adopted as a driving use case by evaluating the accuracy of TRACER extensively in two real-world hospital datasets. 

TRACER supports doctor validation with accurate and interpretable clinical decisions in scenarios including real-time prediction & alert, patient-level interpretation and feature-level interpretation. Besides, TRACER is also validated in a critical financial application. The experimental results confirm that TRACER facilitates both accurate and interpretable analytics for high stakes applications. 

TRACER outperforms existing baseline methods such as LR and GBDT significantly due to the capability of modeling the time-series data and outperforms RETAIN by capturing both time-invariant and time-variant feature importance. 

Furthermore, TRACER achieves better interpretability and prediction performance than BIRNN and Dipole for both datasets.

Benefits

TRACER facilitates doctor validation with accurate and interpretable clinical decision support in the following ways:

  • Real-time prediction and alert – TRACER can send an alert to the doctor when the prediction for a patient exceeds a predefined risk threshold to enable the doctor to take preventive action to avoid further deterioration
  • Patient-level interpretation – TRACER improves the understanding of doctors on each patient, and helps identify the underlying reasons for developing certain diseases
  • Feature-level interpretation – TRACER assists doctors in understanding the characteristics of particular features in disease development and hence, contributes to the advancement of medical research

Potential Application(s)

TRACER can be applied in any high stake application that places importance on accurate and interpretable analytics such as healthcare and finance.

We welcome interest from the industry for collaboration/ co-development / customisation of the technology into a new product or service. If you have any enquiries or are keen to collaborate, please contact us.