우울증과 조울증은 꾸준한 약물치료에도 자주 재발하는 질환으로 기존의 약물치료만으로는 치료에 있어 한계가 있기 때문에 새로운 치료에 대한 필요성이 대두된다. 특히 규칙적인 생활습관과 수면의 관리는 재발 예방에 있어서 매우 중요하다. 기존의 약물치료와 병행한 스마트밴드와 스마트폰을 이용한 디지털치료제(스마트폰앱)를 통해 생활습관 관리를 할 때 우울증, 조울증의 재발을 현저히 감소시킬 수 있다는 연구결과를 발표했다.
Research on digital healthcare technology to improve mental health
Depression and bipolar disorder are diseases that frequently recur despite steady drug treatment, and the need for new treatment is emerging because existing drug treatment alone has limitations in treatment. In particular, regular lifestyle and sleep management are very important in preventing recurrence. Research results have been announced that the recurrence of depression and bipolar disorder can be significantly reduced when lifestyle management is performed through a digital treatment (smartphone app) using a smart band and a smartphone in parallel with conventional drug treatment.
웨어러블 센서 디지털표현형 기반 정신적 불안정성 예측 및 위기상황 조기감지 모니터링 시스템 개발
연구진이 이미 기확보하고 있는 500여명의 정신과 환자로부터 수년간 수집된 웨어러블 기기 데이터를 통해 수 일간 매일 하루주기로 수집되는 데이터 기반 정신적 불안정성을 예측하는 알고리즘을 개발하며, 웨어러블에서 얻어지는 실시간 센서 데이타에서 즉각적인 자타해 위험성을 실시간으로 측정하는 알고리즘을 통해 위험신호 조기감지 모델을 개발하며, 기존에 이미 잘 알려져 있는 자타해 위험성이 높은 임상 특성(과거력, 설문결과, 현재증상, 치료진평가 등)을 같이 총체적, 다층적으로 분석하여 총체적 Stacking 앙상블 자‧타해 위기상황 통합 예측모델을 개발한다. 개발된 통합 예측모델에 기반한 시작품을 개발 및 데이터수집 및 운용에 관한 Pilot study를 시행하며, 그 결과 완성된 위기상황 조기감지 모니터링 시스템 시제품을 개발하여, 이를 종합병원 정신병동에 적용하여 feasibility를 확인하는 실증연구를 시행한다.
Development of a Wearable Sensor-Based Monitoring System for Early Detection and Prevention of Mental Instability and Suicidal/Violent Risk
We develop an algorithm to predict mental instability based on data collected every day for several days through wearable device from 500 psychiatric patients already secured by the research team. A risk signal early detection model is developed through an algorithm that measures the immediate risk of self-harm in real-time sensor data obtained from wearable. A comprehensive and multi-layered analysis of clinical characteristics (past history, survey results, current symptoms, treatment team evaluation, etc.) was studied to develop a comprehensive stacking ensemble self/other risk situation integrated prediction model. We develop a prototype based on the developed integrated prediction model and conduct a pilot study on data collection and operation. As a result, we develop an early detection monitoring system for a completed crisis situation and apply it to the psychiatric ward of a general hospital to conduct empirical research to confirm feasibility.
To facilitate software quality assurance, defect prediction metrics, such as source code metrics, change churns, and the number of previous defects, have been actively studied. Despite the common understanding that developer behavioral interaction patterns can affect software quality, these widely used defect prediction metrics do not consider developer behavior. We therefore propose micro interaction metrics (MIMs), which are metrics that leverage developer interaction information. The developer interactions, such as file editing and browsing events in task sessions, are captured and stored as information by Mylyn, an Eclipse plug-in. Our experimental evaluation demonstrates that MIMs significantly improve overall defect prediction accuracy when combined with existing software measures, perform well in a cost-effective manner, and provide intuitive feedback that enables developers to recognize their own inefficient behaviors during software development.
Adherence to coding conventions during the code production stage of software development is essential. Benefits include enabling programmers to quickly understand the context of shared code, communicate with one another in a consistent manner, and easily maintain the source code at low costs. In reality, however, programmers tend to doubt or ignore the degree to which the quality of their code is affected by adherence to these guidelines. This paper addresses research questions such as “Do violations of coding conventions affect the readability of the produced code?”, “What kinds of coding violations reduce code readability?”, and “How much do variable factors such as developer experience, project size, team size, and project maturity influence coding violations?” To respond to these research questions, we explored 210 open-source Java projects with 117 coding conventions from the Sun standard checklist. We believe our findings and the analysis approach used in the paper will encourage programmers and QA managers to develop their own customized and effective coding style guidelines.
Many existing warning prioritization techniques seek to reorder the static analysis warnings such that true positives are provided first. However, excessive amount of time is required therein to investigate and fix prioritized warnings because some are not actually true positives or are irrelevant to the code context and topic. In this paper, we propose a warning prioritization technique that reflects various latent topics from bug-related code blocks. Our main aim is to build a prioritization model that comprises separate warning priorities depending on the topic of the change sets to identify the number of true positive warnings. For the performance evaluation of the proposed model, we employ a performance metric called warning detection rate, widely used in many warning prioritization studies, and compare the proposed model with other competitive techniques. Additionally, the effectiveness of our model is verified via the application of our technique to eight industrial projects of a real global company.