SCH: Machine LEarning & MicrofluiDics for Multimodal Sensing of TiCk-bOrne Diseases(MEDICO) - PROJECT SUMMARY (See instructions): The incidence of tick-borne diseases (TbD) is increasing, with ~500k reported cases during 2004-2016 from the National Notifiable Diseases Surveillance System of the Centers for Disease Control (CDC), and is expected to increase due to global climate change. Some of the most common TbD in the U.S. include anaplasmosis, babesiosis, and Lyme disease transmitted by the same tick vector, I. scapularis. Lyme disease is especially pernicious because it is difficult to diagnose early, often misdiagnosed, and is difficult to treat. Because current diagnostic methods are insufficiently sensitive or non-existent, we propose a novel approach for diagnosis via a microfluidic platform with an integrated multimodal sensing system and machine learning (ML) algorithm. Based on our preliminary and published data, we hypothesize that we can detect TbD and their coinfections from whole blood. Central to our vision is a system designed for minimal user intervention to detect and measure complex cell data using ML. The diagnosis of TbD will be achieved through three objectives: 1) design a dielectrophoresis (DEP) based platform for detecting TbDs from whole blood; 2) design 3D sensors and a readout integrated circuit (ROIG) for sensitive in-vitro detection of cells; and 3) develop an ML algorithm to diagnose early-stage Lyme disease, babesiosis, and anaplasmosis. An extremely sensitive (<0.1aF), low-voltage (1.8V for core & 3.3V for 10), low-power (<10mW), multi-channel (~16-ch), high-speed (~5MSample/sec per channel) readout integrated circuit (ROIG) will be integrated to detect single-cell behavior at high resolution (~10-bit resolution). The novel ML algorithm(s) will use cell data to determine the type of infection intelligently, reducing >96Gbit data during each diagnostic cycle (~2min) into as simple as 1-byte diagnostic information (i.e., 1: positive, 0: negative) per tested population of the cells or disease. Further, an open-source program to diagnose TbD will be created and made available freely from a public software repository. This developed diagnostic technology will simultaneously meet high-sensitivity, low-power/voltage, high-purity, and high-viability performance metrics by cutting the diagnosis time from 4-6 weeks (late stages) to <30 min (early stages). The proposed research will lead to faster diagnosis, reducing the hospitalization, morbidity, and mortality associated with TbD and their coinfections.
