Nomogram of Naive Bayesian Model for Recurrence Prediction of Breast Cancer / 대한의료정보학회지

OBJECTIVES: Breast cancer has a high rate of recurrence, resulting in the need for aggressive treatment and close follow-up. However, previously established classification guidelines, based on expert panels or regression models, are controversial. Prediction models based on machine learning show excellent performance, but they are not widely used because they cannot explain their decisions and cannot be presented on paper in the way that knowledge is customarily represented in the clinical world. The principal objective of this study was to develop a nomogram based on a naïve Bayesian model for the prediction of breast cancer recurrence within 5 years after breast cancer surgery. METHODS: The nomogram can provide a visual explanation of the predicted probabilities on a sheet of paper. We used a data set from a Korean tertiary teaching hospital of 679 patients who had undergone breast cancer surgery between 1994 and 2002. Seven prognostic factors were selected as independent variables for the model. RESULTS: The accuracy was 80%, and the area under the receiver operating characteristics curve (AUC) of the model was 0.81. CONCLUSIONS: The nomogram can be easily used in daily practice to aid physicians and patients in making appropriate treatment decisions after breast cancer surgery.

New Alert Override Codes for the Drug Utilization Review System Derived from Outpatient Prescription Data from a Tertiary Teaching Hospital in Korea / 대한의료정보학회지

OBJECTIVES: This paper proposes new alert override reason codes that are improvements on existing Drug Utilization Review (DUR) codes based on an analysis of DUR alert override cases in a tertiary medical institution. METHODS: Data were obtained from a tertiary teaching hospital covering the period from April 1, 2012 to January 15, 2013. We analyzed cases in which doctors had used the 11 overlapping prescription codes provided by the Health Insurance Review and Assessment Service (HIRA) or had provided free-text reasons. RESULTS: We identified 27,955 alert override cases. Among these, 7,772 (27.8%) utilized the HIRA codes, and 20,183 (72.2%) utilized free-text reasons. According to the free-text content analysis, 8,646 cases (42.8%) could be classified using the 11 HIRA codes, and 11,537 (57.2%) could not. In the unclassifiable cases, we identified the need for codes for "prescription relating to operation" and "emergency situations." Two overlapping prescription codes required removal because they were not used. Codes A, C, F, H, I, and J (for drug non-administration cases) explained surrounding situations in too much detail, making differentiation between them difficult. These 6 codes were merged into code J4: "patient was not taking/will not take the medications involved in the DDI." Of the 11 HIRA codes, 6 were merged into a single code, 2 were removed, and 2 were added, yielding 6 alert override codes. We could codify 23,550 (84.2%) alert override cases using these codes. CONCLUSIONS: These new codes will facilitate the use of the drug-drug interactions alert override in the current DUR system. For further study, an appropriate evaluation should be conducted with prescribing clinicians.

Development of Novel Breast Cancer Recurrence Prediction Model Using Support Vector Machine / 한국유방암학회지

PURPOSE: The prediction of breast cancer recurrence is a crucial factor for successful treatment and follow-up planning. The principal objective of this study was to construct a novel prognostic model based on support vector machine (SVM) for the prediction of breast cancer recurrence within 5 years after breast cancer surgery in the Korean population, and to compare the predictive performance of the model with the previously established models. METHODS: Data on 679 patients, who underwent breast cancer surgery between 1994 and 2002, were collected retrospectively from a Korean tertiary teaching hospital. The following variables were selected as independent variables for the prognostic model, by using the established medical knowledge and univariate analysis: histological grade, tumor size, number of metastatic lymph node, estrogen receptor, lymphovascular invasion, local invasion of tumor, and number of tumors. Three prediction algorithms, with each using SVM, artificial neural network and Cox-proportional hazard regression model, were constructed and compared with one another. The resultant and most effective model based on SVM was compared with previously established prognostic models, which included Adjuvant! Online, Nottingham prognostic index (NPI), and St. Gallen guidelines. RESULTS: The SVM-based prediction model, named 'breast cancer recurrence prediction based on SVM (BCRSVM),' proposed herein outperformed other prognostic models (area under the curve=0.85, 0.71, 0.70, respectively for the BCRSVM, Adjuvant! Online, and NPI). The BCRSVM evidenced substantially high sensitivity (0.89), specificity (0.73), positive predictive values (0.75), and negative predictive values (0.89). CONCLUSION: As the selected prognostic factors can be easily obtained in clinical practice, the proposed model might prove useful in the prediction of breast cancer recurrence. The prediction model is freely available in the website (http://ami.ajou.ac.kr/bcr/).

Differential Diagnosis in Idiopathic Granulomatous Mastitis and Tuberculous Mastitis / 한국유방암학회지

PURPOSE: Idiopathic granulomatous mastitis (IGM) is a rare chronic inflammatory disease of unknown etiology. The diagnosis of IGM requires that other granulomatous lesions in the breast be excluded. Tuberculous mastitis (TM) is also an uncommon disease that is often difficult to differentiate from IGM. The purpose of this study is to develop a new algorithm for the differential diagnosis and treatment of IGM and TM. METHODS: Medical records of 68 patients (58 with IGM and 10 with TM) between July 1999 and February 2009 were retrospectively reviewed. RESULTS: The mean age of the patients was 33.5 (IGM) and 40 (TM) years (p=0.018). The median follow-up was 84 months. Of the total 10 patients with TM, 5 patients had a history of pulmonary tuberculosis. The most common symptoms of the diseases were breast lump and pain. However, axillary lymphadenopathy was more seen in TM (50%) compared to IGM (20.6%) (p=0.048). TM showed more cancer-mimicking findings on radiologic study (p=0.028). In IGM, 48 patients (82.7%) underwent surgical wide excision and 21 patients (36.2%) were managed with corticosteroid therapy and antibiotics. All of the TM patients received anti-tuberculosis medications and 9 patients (90%) underwent wide excision. The mean treatment duration was 2.8 months in IGM and 8.4 months in TM. Recurrence developed in 5 patients (8.6%) in IGM and 1 patient (10%) in TM. CONCLUSION: This study shows different characteristics between IGM and TM. The IGM patients were younger and had more mastalgia symptoms than the TM patients. Axillary lymphadenopathy was seen more often in TM patients. Half of the TM patients had pulmonary tuberculosis or tuberculosis lymphadenitis. Surgical wide excision might be both therapeutic and useful for providing an exact diagnosis.

An Automated Measurement of Ciliary Beating Frequency using a Combined Optical Flow and Peak Detection / 대한의료정보학회지

OBJECTIVES: The mucociliary transport system is a major defense mechanism of the respiratory tract. The performance of mucous transportation in the nasal cavity can be represented by a ciliary beating frequency (CBF). This study proposes a novel method to measure CBF by using optical flow. METHODS: To obtain objective estimates of CBF from video images, an automated computer-based image processing technique is developed. This study proposes a new method based on optical flow for image processing and peak detection for signal processing. We compare the measuring accuracy of the method in various combinations of image processing (optical flow versus difference image) and signal processing (fast Fourier transform [FFT] vs. peak detection [PD]). The digital high-speed video method with a manual count of CBF in slow motion video play, is the gold-standard in CBF measurement. We obtained a total of fifty recorded ciliated sinonasal epithelium images to measure CBF from the Department of Otolaryngology. The ciliated sinonasal epithelium images were recorded at 50-100 frames per second using a charge coupled device camera with an inverted microscope at a magnification of x1,000. RESULTS: The mean square errors and variance for each method were 1.24, 0.84 Hz; 11.8, 2.63 Hz; 3.22, 1.46 Hz; and 3.82, 1.53 Hz for optical flow (OF) + PD, OF + FFT, difference image [DI] + PD, and DI + FFT, respectively. Of the four methods, PD using optical flow showed the best performance for measuring the CBF of nasal mucosa. CONCLUSIONS: The proposed method was able to measure CBF more objectively and efficiently than what is currently possible.

Analysis of Relationship between Levofloxacin and Corrected QT Prolongation Using a Clinical Data Warehouse / 대한의료정보학회지

OBJECTIVE: The aim of this study was to examine whether or not levofloxacin has any relationship with QT prolongation in a real clinical setting by analyzing a clinical data warehouse of data collected from different hospital information systems. METHODS: Electronic prescription data and medical charts from 3 different hospitals spanning the past 9 years were reviewed, and a clinical data warehouse was constructed. Patients who were both administrated levofloxacin and given electrocardiograms (ECG) were selected. The correlations between various patient characteristics, concomitant drugs, corrected QT (QTc) prolongation, and the interval difference in QTc before and after levofloxacin administration were analyzed. RESULTS: A total of 2,176 patients from 3 different hospitals were included in the study. QTc prolongation was found in 364 patients (16.7%). The study revealed that age (OR 1.026, p < 0.001), gender (OR 0.676, p = 0.007), body temperature (OR 1.267, p = 0.024), and cigarette smoking (OR 1.641, p = 0.022) were related with QTc prolongation. After adjusting for related factors, 12 drugs concomitant with levofloxacin were associated with QTc prolongation. For patients who took ECGs before and after administration of levofloxacin during their hospitalization (n = 112), there was no significant difference in QTc prolongation. CONCLUSIONS: The age, gender, body temperature, cigarette smoking and various concomitant drugs might be related with QTc prolongation. However, there was no definite causal relationship or interaction between levofloxacin and QTc prolongation. Alternative surveillance methods utilizing the massive accumulation of electronic medical data seem to be essential to adverse drug reaction surveillance in future.

A Comparison of Intensive Care Unit Mortality Prediction Models through the Use of Data Mining Techniques / 대한의료정보학회지

OBJECTIVES: The intensive care environment generates a wealth of critical care data suited to developing a well-calibrated prediction tool. This study was done to develop an intensive care unit (ICU) mortality prediction model built on University of Kentucky Hospital (UKH)'s data and to assess whether the performance of various data mining techniques, such as the artificial neural network (ANN), support vector machine (SVM) and decision trees (DT), outperform the conventional logistic regression (LR) statistical model. METHODS: The models were built on ICU data collected regarding 38,474 admissions to the UKH between January 1998 and September 2007. The first 24 hours of the ICU admission data were used, including patient demographics, admission information, physiology data, chronic health items, and outcome information. RESULTS: Only 15 study variables were identified as significant for inclusion in the model development. The DT algorithm slightly outperformed (AUC, 0.892) the other data mining techniques, followed by the ANN (AUC, 0.874), and SVM (AUC, 0.876), compared to that of the APACHE III performance (AUC, 0.871). CONCLUSIONS: With fewer variables needed, the machine learning algorithms that we developed were proven to be as good as the conventional APACHE III prediction.