A comprehensive study was conducted to differentiate cardiovascular disease (CVD) subjects from non-CVD subjects using short recording electrocardiogram (ECG) of 244 Malaysian adults in The MalaysianCohort project. An...A comprehensive study was conducted to differentiate cardiovascular disease (CVD) subjects from non-CVD subjects using short recording electrocardiogram (ECG) of 244 Malaysian adults in The MalaysianCohort project. An automated peak detection algorithm to detect nine fiducialpoints of electrocardiogram (ECG) was developed. Forty-eight features wereextracted in both time and frequency domains, including statistical featuresobtained from heart rate variability and Poincare plot analysis. These includefive new features derived from spectrum counts of five different frequencyranges. Feature selection was then made based on p-value and correlationmatrix. Selected features were used as input for five classifiers of artificialneural network (ANN), k-nearest neighbors (kNN), support vector machine(SVM), discriminant analysis (DA), and decision tree (DT). Results showedthat six features related to T wave were statistically significant in distinguishingCVD and non-CVD groups. ANN had performed the best with 94.44% specificity and 86.3% accuracy, followed by kNN with 80.56% specificity, 86.49%sensitivity and 83.56% accuracy. The novelties of this study were in providingalternative solutions to detect P-onset, P-offset, T-offset as well as QRS-onsetpoints using discrete wavelet transform method. Additionally, two out of thefive newly proposed spectral features were significant in differentiating bothgroups, at frequency ranges of 1–10 Hz and 5–10 Hz. The prediction outcomeswere also comparable to previous related studies and significantly importantin using ECG to predict cardiac-related events among CVD and non-CVDsubjects in the Malaysian population.展开更多
Fiducial marker detection algorithms in kilovoltage x-ray images using physical characteristics of transmission x-ray have been proposed. It, however, has been suggested recently that factors besides transmission x-ra...Fiducial marker detection algorithms in kilovoltage x-ray images using physical characteristics of transmission x-ray have been proposed. It, however, has been suggested recently that factors besides transmission x-ray affect x-ray images. The purpose of this study was to develop a new fiducial detection algorithm using fiducial intensity estimation based on physical characteristics of x-ray images with gold fiducials. First, x-ray images of a fiducial on a water-equivalent phantom were acquired. It was observed that the ratio of background to fiducial intensity in the images decreased as phantom thickness increased. Based on the negative correlation, we identified a function for estimating fiducial intensity that consists of background intensity and the amount of scattered radiation by the other x-ray source of an orthogonal imaging system and a treatment beam. Then, we developed an algorithm that extracts fiducial candidates using the estimation function. Its performance was measured using x-ray images which had 3824 fiducials altogether. The average number of false-positive detection of the proposed algorithm in single image was one-tenth of an algorithm considering only transmission x-ray. The proposed algorithm detected 99.5% of all fiducials under an error of 1.0 mm, while the other algorithm detected 94.7% or less (Clinical trial number: UMIN000005324).展开更多
基金This study was supported by the Ministry of Education Malaysia’s Fundamental Research Grant Scheme FRGS/1/2019/TK04/UKM/02/4TMC research was funded by a top-down grant from the Ministry of Education Malaysia(Grant Number PDE48).
文摘A comprehensive study was conducted to differentiate cardiovascular disease (CVD) subjects from non-CVD subjects using short recording electrocardiogram (ECG) of 244 Malaysian adults in The MalaysianCohort project. An automated peak detection algorithm to detect nine fiducialpoints of electrocardiogram (ECG) was developed. Forty-eight features wereextracted in both time and frequency domains, including statistical featuresobtained from heart rate variability and Poincare plot analysis. These includefive new features derived from spectrum counts of five different frequencyranges. Feature selection was then made based on p-value and correlationmatrix. Selected features were used as input for five classifiers of artificialneural network (ANN), k-nearest neighbors (kNN), support vector machine(SVM), discriminant analysis (DA), and decision tree (DT). Results showedthat six features related to T wave were statistically significant in distinguishingCVD and non-CVD groups. ANN had performed the best with 94.44% specificity and 86.3% accuracy, followed by kNN with 80.56% specificity, 86.49%sensitivity and 83.56% accuracy. The novelties of this study were in providingalternative solutions to detect P-onset, P-offset, T-offset as well as QRS-onsetpoints using discrete wavelet transform method. Additionally, two out of thefive newly proposed spectral features were significant in differentiating bothgroups, at frequency ranges of 1–10 Hz and 5–10 Hz. The prediction outcomeswere also comparable to previous related studies and significantly importantin using ECG to predict cardiac-related events among CVD and non-CVDsubjects in the Malaysian population.
文摘Fiducial marker detection algorithms in kilovoltage x-ray images using physical characteristics of transmission x-ray have been proposed. It, however, has been suggested recently that factors besides transmission x-ray affect x-ray images. The purpose of this study was to develop a new fiducial detection algorithm using fiducial intensity estimation based on physical characteristics of x-ray images with gold fiducials. First, x-ray images of a fiducial on a water-equivalent phantom were acquired. It was observed that the ratio of background to fiducial intensity in the images decreased as phantom thickness increased. Based on the negative correlation, we identified a function for estimating fiducial intensity that consists of background intensity and the amount of scattered radiation by the other x-ray source of an orthogonal imaging system and a treatment beam. Then, we developed an algorithm that extracts fiducial candidates using the estimation function. Its performance was measured using x-ray images which had 3824 fiducials altogether. The average number of false-positive detection of the proposed algorithm in single image was one-tenth of an algorithm considering only transmission x-ray. The proposed algorithm detected 99.5% of all fiducials under an error of 1.0 mm, while the other algorithm detected 94.7% or less (Clinical trial number: UMIN000005324).
