Deep Learning Approach for Classifying Bacteria types using Morphology of Bacterial Colony.

The significant bottlenecks in determining bacterial species are much more time-consuming and the biology specialist's long-term experience requirements. Specifically, it takes more than half a day to cultivate a bacterium, and then a skilled microbiologist and a costly specialized machine are utilized to analyze the genes and classify the bacterium according to its nucleotide sequence. To overcome these issues as well as get higher recognition accuracy, we proposed applying convolutional neural networks (CNNs) architectures to automatically classify bacterial species based on some key characteristics of bacterial colonies. Our experiment confirmed that the classification of three bacterial colonies could be performed with the highest accuracy (97.19%) using a training set of 5000 augmented images derived from the 40 original photos taken in the Hanoi Medical University laboratory in Vietnam.

Bacteria Shape Classification using Small-Scale Depthwise Separable CNNs.

Fast detection and classification of bacteria species play a crucial role in modern clinical microbiology systems. These processes are often performed manually by medical biologists using different shapes and morphological characteristics of bacteria species. However, it is clear that the manual taxonomy of bacteria types from microscopy images takes time and effort and is a great challenge for even experienced experts. A new revolution has been inaugurating with the development of machine learning methods to identify bacteria automatically from digital electron microscopy. In this paper, we introduce an automated model of bacteria shape classification based on Depthwise Separable Convolution Neural Networks (DS-CNNs). This architecture has excellent advantages with lower computational costs and reliable recognition accuracy. The experiment results indicate that after training with 1669 images, the proposed architecture can reach 97% validation accuracy and work well to classify three main shapes of bacteria.

High Accuracy Heartbeat Detection from CW-Doppler Radar Using Singular Value Decomposition and Matched Filter.

Heart rate measurement using a continuous wave Doppler radar sensor (CW-DRS) has been applied to cases where non-contact detection is required, such as the monitoring of vital signs in home healthcare. However, as a CW-DRS measures the speed of movement of the chest surface, which comprises cardiac and respiratory signals by body motion, extracting cardiac information from the superimposed signal is difficult. Therefore, it is challenging to extract cardiac information from superimposed signals. Herein, we propose a novel method based on a matched filter to solve this problem. The method comprises two processes: adaptive generation of a template via singular value decomposition of a trajectory matrix formed from the measurement signals, and reconstruction by convolution of the generated template and measurement signals. The method is validated using a dataset obtained in two different experiments, i.e., experiments involving supine and seated subject postures. Absolute errors in heart rate and standard deviation of heartbeat interval with references were calculated as 1.93±1.76bpm and 57.0±28.1s for the lying posture, and 9.72±7.86bpm and 81.3±24.3s for the sitting posture.

Contactless Heartbeat Detection from CW-Doppler Radar using Windowed-Singular Spectrum Analysis.

The continuous-wave Doppler radar measures the movement of a chest surface including of cardiac and breathing signals and the body movement. The challenges associated with extracting cardiac information in the presence of respiration and body movement have not been addressed thus far. This paper presents a novel method based on the windowed-singular spectrum analysis (WSSA) for solving this issue. The algorithm consists of two processes: signal decomposition via WSSA followed by the reconstruction of decomposed heartbeat signals through convolution. An experiment was conducted to collect chest signals in 212 people by Doppler radar. In order to confirm the effect of reducing the large noise by the proposed method, we evaluated 136 signals that were considered to contain respiration body movements from the collected signals. When comparing to the performance of a band-pass filter, the proposed analysis achieves improved beat count accuracy. The results indicate its applicability to contactless heartbeat estimation under involving respiration and body movements.

Non-Contact Blood Pressure Measurement Scheme Using Doppler Radar.

A continuous cuffless non-contact blood pressure (BP) measurement scheme using Doppler radar is proposed. This non-contact BP estimation scheme uses the periods in which the heart beats and periods in which the heart contracts. These periods are obtained using Doppler radar signals. Diastolic BP (DBP) was estimated using the period in which the heart contracts. Pulse pressure (PP) was estimated using one period in which the heart beats and one period in which the heart contracts. Systolic BP (SBP) was obtained by the sum of the estimated DBP and PP. The correlation of the estimated BP and the BP acquired by the BP monitor was calculated. The correlation coefficients were 0.79 for SBP, 0.88 for DBP, and 0.81 for PP. The BP was successfully measured in a contactless manner.

Precise Heart Rate Measurement Using Non-contact Doppler Radar Assisted by Machine-Learning-Based Sleep Posture Estimation.

Non-contact and continuous heart rate measurement using Doppler radar is important for various healthcare applications. In this paper, we propose a precise heart rate measurement method assisted by machine learning based sleep posture estimation. Machine learning is used for processing time-domain signal of the Doppler radar. Doppler radar has attracted much attention due to its non-contact to the subject feature. Moreover, it will not encroach into the privacy of the subject compared to image sensors. The method proposed in this paper automatically removes the data from the raw signal while the patient is moving or is not staying on the bed. This method based on machine learning uses simple features to reduce the computational cost thereby enabling real-time application. The sleeping posture was detected with an accuracy of 88.5%, and the error ratios of heart rate estimation were reduced by 15.2% in a sleep laboratory testing on 6 subjects.

A non-contact infection screening system using medical radar and Linux-embedded FPGA: Implementation and preliminary validation.

OBJECTIVES: In this study, an infection screening system was developed to detect patients suffering from infectious diseases. In addition, the system was also designed to deal with the variability in age and gender, which would affect the accuracy of the detection. Furthermore, to enable a low-cost, non-contact and embedded system, multiple vital signs from a medical radar were measured and all algorithms were implemented on a Field Programmable Gate Array, named PYNQ-Z1. METHODS: The system consisted of two main stages: digital signal processing and data classification. In the former stage, Butterworth filters, with flexible cut-off frequencies depending on age and gender, and a time-domain peak detection algorithm were deployed to compute three vital signs, namely heart rate, respiratory rate, and standard deviation of heart beat-to-beat interval. For the classification problem, two machine learning models, Support Vector Machine and Quadratic Discriminant Analysis, were implemented. RESULTS: The Student's t-test showed that our proposed digital signal processing algorithms coped well with the variability of human cases in age and gender. Meanwhile, the f1-score of roughly 98.0% represented the high sensitivity and specificity of our proposed machine learning methods. CONCLUSION: This study outlines the implementation of an infection screening system, which achieved competent performance. The system might be beneficial for fast screening of infected patients at public health centers in underdeveloped areas, where people have little access to healthcare.

Dengue Fever Detecting System Using Peak-detection of Data from Contactless Doppler Radar.

Infectious diseases, such as dengue fever and Middle East respiratory syndrome, have become prevalent worldwide in recent times. To conduct highly accurate and effective infection screening, we are working on the development of a contactless infection screening system using Doppler radar and thermography. In our previous work, three parameters (face temperature, heartbeat rate, and respiration rate) were used to judge whether a subject was infected. However, facial temperature measurements may be vastly different from temperatures measured at the axilla owing to influence from the immediate environment. In this study, heartbeat rate (HR), respiration rate (RR), and standard deviation of heartbeat interval (SDHI) were used to quantify the infection screening system without using facial temperature as a parameter. We found that respiratory sinus arrhythmia (RSA) diminished in patients who had dengue fever. We gathered data from 47 patients with dengue fever using a 10-GHz Doppler radar system at the National Hospital of Tropical Diseases (NHTD) in Hanoi, Vietnam. To evaluate the accuracy, the data of these patients were compared to that of 23 unaffected subjects. We observed that a linear discriminant analysis (LDA) was effective at detecting the dengue fever conditions, and the detection accuracy was approximately 97.6%.

Non-contact acquisition of respiration and heart rates using Doppler radar with time domain peak-detection algorithm.

The non-contact measurement of the respiration rate (RR) and heart rate (HR) using a Doppler radar has attracted more attention in the field of home healthcare monitoring, due to the extremely low burden on patients, unconsciousness and unconstraint. Most of the previous studies have performed the frequency-domain analysis of radar signals to detect the respiration and heartbeat frequency. However, these procedures required long period time (approximately 30 s) windows to obtain a high-resolution spectrum. In this study, we propose a time-domain peak detection algorithm for the fast acquisition of the RR and HR within a breathing cycle (approximately 5 s), including inhalation and exhalation. Signal pre-processing using an analog band-pass filter (BPF) that extracts respiration and heartbeat signals was performed. Thereafter, the HR and RR were calculated using a peak position detection method, which was carried out via LABVIEW. To evaluate the measurement accuracy, we measured the HR and RR of seven subjects in the laboratory. As a reference of HR and RR, the persons wore contact sensors i.e., an electrocardiograph (ECG) and a respiration band. The time domain peak-detection algorithm, based on the Doppler radar, exhibited a significant correlation coefficient of HR of 0.92 and a correlation coefficient of RR of 0.99, between the ECG and respiration band, respectively.