MACHINE LEARNING ON ICLINIC DATA
DIGITAL MEDICINE & WEARABLE TECHNOLOGY
Author: Nathaniel Bischoff
Coauthor(s): Spyro Mousses PhD, William Feaster MD, Anthony Chang
Status: Work in Progress
The iClinic is a philosophy of leveraging emerging technologies to help create efficiencies, improve workflow and the continuity of care for patients by bringing CHOC expertise to patients’ homes. There are five key components to the iClinic: instantaneous, intermittent, individual therapy, intelligent data-driven medicine, and intuitive interactions. The iClinic will use virtual visits with home monitoring to create more regular data. To create a more precision medicine approach, the patient’s genome will be sequenced and corresponding pharmacogenomics will be determined to find medicine and treatment best suited for the patient.
All of the iClinic data from inpatient and outpatient visits, genomic data, and home monitoring data will be compiled, formatted and analyzed. We will be creating a digital safe through deep geno- and pheno- typing. We will be assembling a training set from previous patients to begin the model building process. The use of deep neural networks in Python using TensorFlow and Keras will be our machine learning approach of choice because of our structured data. The outcome would be to create a system that will allow for real time prediction of appropriate treatment for patients in the iClinic. Because we are selecting a subset of the data and not the full population, we will be introducing some bias into our system. At first, we will be selecting just CHOC patients with cardiopulmonary diseases to build our models. Our models will not only track patients but prescribe actionable options for seasoned clinicians to reason with, and decide the best course of action.
Putting intelligent programs in the hands of the physicians will allow them to make better decisions without adding the burden of paperwork and finding knowledge. This is the clinic of the future. The ultimate goal is improving quality care and positive outcomes for patients.