Using the Lumme system to improve real-time blood alcohol detection

点击翻译按钮获取中文摘要

PROJECT SUMMARY There is a critical need to develop non-invasive technologies to monitor blood alcohol levels in real time. Such technology would advance research assessing drinking in the natural environment, clinical platforms intervening in real-time, and justice applications. Currently developed technology, transdermal alcohol sensors (TAS), passively detect alcohol intoxication by continuously testing for alcohol excreted in sweat. While such technology provides continuous monitoring of alcohol consumption, there are several significant shortcomings that limit its utility in research and clinical contexts. Limitations include lack of temporal resolution (i.e., delay between drinking and sweat detection), poor sensitivity of drinking episodes that do not meet binge criteria, and lack of information about drinking topography. In response to PAR-16-410, we propose to pair our Lumme Inc behavioral tracking platform with a TAS device to improve the detection of real-time drinking behavior. The Lumme system combines the use of existing commercial hardware (smartwatch and smartphone) to assess drinking behavior in real time. The Lumme system uses inertial measurement units to record the movement of a person’s arm in three dimensional space with accelerometers and gyroscope contained in a smartwatch. Arm movements associated with drinking are distinctive, and can be accurately captured with a machine learning algorithm. The combined Lumme/TAS system complements one another’s strengths by providing detailed drinking data that can be linked to subsequent increases in blood alcohol content (BAC). The combined system will be able to accurately track drinking variables of interest to researchers and clinicians (e.g., start/end of drinking, length of drinking episode, rate of drinking, activity while drinking, social networks while drinking, location of drinking, and resulting BAC). Additionally, the gesture data can remotely capture low level drinking episodes that are undetected by existing TAS systems. We currently have a funded SBIR from NCI to develop and validate a smoking cessation platform using the Lumme system. For the current application, we will propose a Phase I study to establish the feasibility and initial validity of using the combined Lumme/TAS system for tracking alcohol use behavior in real time. We plan to video record ad-libitum drinking behavior in heavy drinkers fitted with both a TAS system (BACtrack Skyn) and our Lumme system. The data collected in the clinical setting will act as ground truth which will be used to train our machine learning algorithm that will be sensitive to (i) how drinking gestures may change as blood alcohol levels rise (ii) how gestures may be affected by type of alcohol and the drinking vessel. To our knowledge, the Lumme system would be the first system to remotely detect drinking gestures using a machine learning algorithm. In addition to improving real-time BAC detection technology the Lumme system could be also developed to deliver just-in-time interventions for problem drinking.