2.Maximal Diagnostic Accuracy in Virtual Telepathology System according to Input Device and Video Signal.
Rae Woong PARK ; Hee Jae JOO ; Hyunee YIM ; Yoon Mi JIN ; Kyi Beom LEE
Korean Journal of Pathology 1999;33(12):1191-1198
Varieties of telepathology system had been developed and in use, but their functional capability and diagnostic accuracy are considered to be inferior to those of conventional optical microscope. This study is intended to find out: 1) the diagnostic accuracy and reproducibility rate according to the input devices and the video signals; 2) any potential technical problems of the telepathology system; 3) any possible physical and psychological impacts. We devised a virtual telepathology system using our existing microscope equipped with CCD camera unit that has no restriction of network speed. Total fifty-five surgical pathology cases from 11 different organs were selected. Three pathologists were involved in making diagnoses. The resulting diagnostic accuracies were: 1 CCD camera with composite video signal was 86.2%; 3 CCD camera with composite video signal was 93.1%; 3 CCD camera with component video signal was 95.0%. The 3 CCD camera with component video signal resulted in 95.0% diagnostic accuracy and was superior to 1 CCD camera with composite video signal. Some technical problems noted during this study were: the visual field of the virtual telepathology system was smaller by 43% than that of microscope; the difference of cell sizes between microscope and monitor; low resolution of image. Some physical and psychological symptoms were noted.
Cell Size
;
Diagnosis
;
Pathology, Surgical
;
Telepathology*
;
Visual Fields
4.A clinical research strategy using longitudinal observational data in the post-electronic health records era.
Journal of the Korean Medical Association 2012;55(8):711-719
Adoption of electronic health records (EHRs) is increasing worldwide. The worldwide EHR adoption rate is estimated to be around 9% to 12%. Thus, the accumulation of medical records in electronic form is also sharply increasing and is expected to be a precious asset for clinical research. Longitudinal observational studies based on EHRs are also increasing. Observational studies covering more than a million people are not rare at present. However, much of the current EHR data are equivalent in form to those of paper records, but are just stored in electronic stor-age devices, rather than as electronic data that can be transferred and shared without loss of clinical semantics. Current EHR systems must be improved in many ways to be used for anal-yses to yield important clinical knowledge. These improvements, which are addressed in this review, include the adoption of clinical data warehouses, use of controlled vocabulary, avoidance of personal/departmental research databases, a standardized interface of many diagnostic devices with the EHR system, control of time-stamp granularity, preparedness for whole-genome sequencing of every patient, confederation or consolidation of multi-institutional EHR data, protection of privacy and confidentiality, and an education system for clinical informaticians.
Adoption
;
Confidentiality
;
Electronic Health Records
;
Electronics
;
Electrons
;
Humans
;
Medical Informatics
;
Medical Records
;
Privacy
;
Semantics
;
Vocabulary, Controlled
5.Interpreting Epidemiologic Evidence: Strategy for Study Design and Analysis.
Healthcare Informatics Research 2011;17(3):196-197
No abstract available.
6.Two Cases of Median Raphe Cyst of Male External Genitalia.
Yong Sun HEO ; Jae Il KWON ; Rae Woong PARK ; Young Soo KIM
Korean Journal of Urology 2000;41(2):349-351
No abstract available.
Genitalia*
;
Humans
;
Male*
7.Basic Concepts and Principles of Data Mining in Clinical Practice.
Journal of Korean Society of Medical Informatics 2009;15(2):175-189
Recently, many hospitals have been adopting clinical data warehouses (CDW) as well as electronic medical records. These new hospital information systems are inevitably introducing very large amounts of clinical data that might be useful for further analysis. However, the electronic clinical data in the CDW are usually byproducts of clinical practice rather than the product of research. Therefore, they include inconsistent and sometimes erroneous information that might not have the specific context of the clinical situations. Data miners usually have various academic backgrounds such as electronics, informatics, statistics, biomedicine, and public health. If the complex situations surrounding the clinical data are not well understood, investigators performing data mining in clinical fields may have problems assessing the information they are confronted with. Here, we would like to introduce some basic concepts on the principles of data mining in clinical fields including legal and ethical considerations as well as technical concerns.
Machine Learning
;
Data Mining
;
Electronic Health Records
;
Electronics
;
Electrons
;
Hospital Information Systems
;
Humans
;
Informatics
;
Public Health
;
Research Personnel
8.The Distributed Research Network, Observational Health Data Sciences and Informatics, and the South Korean Research Network
Korean Journal of Medicine 2019;94(4):309-314
No abstract available.
Informatics
10.Analysis of Adverse Drug Reactions Identified in Nursing Notes Using Reinforcement Learning
Eunjoo JEON ; Youngsam KIM ; Hojun PARK ; Rae Woong PARK ; Hyopil SHIN ; Hyeoun-Ae PARK
Healthcare Informatics Research 2020;26(2):104-111
Electronic Health Records (EHRs)-based surveillance systems are being actively developed for detecting adverse drug reactions (ADRs), but this is being hindered by the difficulty of extracting data from unstructured records. This study performed the analysis of ADRs from nursing notes for drug safety surveillance using the temporal difference method in reinforcement learning (TD learning). Nursing notes of 8,316 patients (4,158 ADR and 4,158 non-ADR cases) admitted to Ajou University Hospital were used for the ADR classification task. A TD(λ) model was used to estimate state values for indicating the ADR risk. For the TD learning, each nursing phrase was encoded into one of seven states, and the state values estimated during training were employed for the subsequent testing phase. We applied logistic regression to the state values from the TD(λ) model for the classification task. The overall accuracy of TD-based logistic regression of 0.63 was comparable to that of two machine-learning methods (0.64 for a naïve Bayes classifier and 0.63 for a support vector machine), while it outperformed two deep learning-based methods (0.58 for a text convolutional neural network and 0.61 for a long short-term memory neural network). Most importantly, it was found that the TD-based method can estimate state values according to the context of nursing phrases. TD learning is a promising approach because it can exploit contextual, time-dependent aspects of the available data and provide an analysis of the severity of ADRs in a fully incremental manner.