Research Summary

The research team aims to develop novel statistical methods for analyzing and modeling modern data of complex and/or mixed modalities. The data modalities he studies encompass traditional types that are categorical-, ordinal- and continuous-scaled, as well as those that are highly structured and complex, such as high-dimensional, multi-dimensional, functional (e.g., curve and image) and multi-mode data. The newly developed statistical methods are designed to effectively leverage information from complex data produced by high-tech imaging or wearable technologies, with the overarching goal of facilitating discovery, evaluation and validation of novel non-invasive markers, improving prediction of health outcomes, and delineating complex pathophysiology of various diseases for their cure and prevention.

1. Development of statistical methods for non-invasive detection of anemia in very low birth weight infants using patient-sourced fingernail photos

This project is motivated by the need to develop non-invasive tools for monitoring anemia in very low birth weight (VLBW; birth weight < 1,500 grams) and reduce the number of routine painful, invasive blood sampling procedures (phlebotomy) that may alter infant neurodevelopment and behavior. The goal is to develop a new image analysis algorithm (IAA) that analyzes full structural information of color data from fingernail smartphone photos to enable accurate, non-invasive prediction of blood hemoglobin level and anemia risk among VLBW infants. Specifically, our aim is three-fold:

A. To develop a new image analysis algorithm (IAA) that produces non-invasive, accurate and stable prediction of hemoglobin level. The IAA will be based on a novel principal component analysis and partial least square methods that provide a non-parametric and parsimonious means to jointly model high-dimensional photos and image metadata, while fully leveraging their spatial structures and co-varying patterns.

B. To develop new unsupervised and supervised clustering methods to study sub-population structures of fingernail photos and image metadata and study their relationships with the underlying physiological mechanisms of anemia. This approach will allow us to formulate a non-invasive image-based screening tool by identifying clusters of VLBW infants with high anemia risk.

C. To develop data-driven tools that leverage longitudinal, patient-level clinical data and IAA predictions to achieve the overarching clinical goal of minimizing the number of blood draws in VLBW infants throughout the care continuum. We use the data of VLBW infants monitored at three level III neonatal intensive care units in Atlanta.

2. Development of statistical methods for non-invasive detection of kidney obstruction

This project is motivated by the urgent need to non-invasively diagnose kidney obstruction which is estimated to affect 27 million Americans. Renal scans are performed by the intravenous injection of a gamma emitting tracer (MAG3) that is rapidly removed from the blood by the kidneys and then travels down the ureters from the kidney to the bladder. The transit of the tracer through the kidneys can be modeled by a time activity curve (renogram curve), generated by placing a region of interest (ROI) over each kidney or specific area of a kidney and counting the photons detected in the kidney ROI at multiple intervals during an initial 20-24 min period of data acquisition and a subsequent 20 acquisition following the administration of a potent diuretic. Interpretation of renal scan is based on the analysis of images, renogram curve and a few selected points on a renogram curve. However, lack of training and limited experience in physicians coupled with the demands of interpreting a large variety of complex imaging studies at ever faster rates increases the error rate of the diagnosis.

To assist physicians in limiting their errors and arriving at the correct diagnosis in their interpretations of MAG3 renal scans, our aim is to develop novel statistical tools that can analyze renal scan data to non-invasively detect kidney obstruction, enhance diagnostic accuracy, direct therapy, and monitor the response to treatment.