Viet Dung Nguyen

An efficient camera-based surveillance for fall detection of elderly people

Falls and consequent injuries among the elderly population are major public health problems that often require medical attention. Every year, the world has more than 30% of people aged over 65 fall and most of them could not stand up without assistance promptly. The long laying makes the injury even more seriously. A system of monitoring the status of the elderly based on image processing technology can help mitigate this situation. This paper proposes an efficient camera-based fall detection system for senior people. We present our preliminary results in our effort to create an emergency alarming system for fall detection aiming at simple, real-time response and high performance. Images from the camera are processed by computer. Status of elderly persons are detected based on combination of two features: aspect ratio and fall angle. Remarkable accuracy rate of the system is as much as 92.5% .

Combination of block difference inverse probability features and support vector machine to reduce false positives in computer-aided detection for massive lesions in mammographic images

A new false positive reduction approach in computer-aided mammographic mass detection has been proposed in this paper. The goal is to discriminate true recognized masses from the normal parenchyma ones. To describe masses, Block Difference Inverse Probability (BDIP) features are utilized. Once the descriptors are extracted, we use Support Vector Machine (SVM) to classify the detected masses. Evaluation on about 2700 suspicious regions detected from Mini-MIAS database gives the discrimination result of 0.91. It indicates that using BDIP features is effective and efficient for reducing false positives.

Fusion of Random Projection, Multi-Resolution Features and Distance Weighted K Nearest Neighbor for Masses Detection in Mammographic Images

With the rapid growth of communications via the Internet, the need for an effective firewall system which has not badly affect the overall network performances has been increased. In this paper, a Field Programmable Gate Array (FPGA) -based firewall system with high performance has been implemented using Network FPGA (NetFPGA) with Xilinx Kintex-7 XC7K325T FPGA. Based on NetFPGA reference router project, a NetFPGA-based firewall system was implemented. The hardware module performs rule matching operation using content addressable memory (CAM) for higher speed data processing. To evaluate system performance, throughput, latency, and memory utilization were measured for different cases using different tools, also the number of rules that an incoming packet is subjected to was varied to get more readings using both software and hardware features. The results showed that the designed firewall system provides better performance than traditional firewalls. System throughput was doubled times of the one with Linux-Iptables firewalls.

Kinect-based virtual training system for rehabilitation

In this paper, we propose a Kinect-based virtual training system for rehabilitation system. The system is aimed to assist patients after stroke or with spiral cord injuries who consequently have movement disorders to perform rehabilitation exercises at home. Movements of patients is capture and their skeletons are tracked by means of Kinect sensor. Patients movements are then compared to the prerecorded movements of virtual coach/trainer.

Design System to Remotely Monitor Patients with Parkinson’s Disease

Parkinson’s disease does not cause immediate danger but it should be monitored continuously because not at any time symptoms can easily observed. Aimed at saving time and costs for patients, as well as the patients cannot always go to the hospital for checking health, in this paper, we present architecture of a system to remotely monitor patients with Parkinson’s disease. The system could help tracking the status of Parkinson patients at home. Movement information of patient’s tremor hand is acquired then sent to central unit via GPRS. At the central unit, the data is collected, stored in 24-hour format and displayed in a tabular or continuous graph. By that mean, doctors can monitor and diagnose the status of the patients.

Detection of tumor in mammographic images by hierarchy of block's features

In the world, breast cancer in female population has increased significantly in recent years. In this paper, we present a new method for circumscribed masses detection in mammograms. First of all, the original mammogram is preprocessed to remove unwanted regions such as label, pectoral muscle and small bright spots similar to mass. Next, we divide the mammogram into equal blocks and calculate statistical characteristics or features for each block. Then, each block is classified as abnormal or normal block based on hierarchy of its features. Adjacent abnormal blocks are then merged into suspicious region. A sensitivity of 95.3% with only 0.48 false alarms per image is observed when evaluating the proposed method on mammographic images from Mini-MIAS database.

Design Current Stabilization Circuit for Transcranial Impulse Electro Stimulation

For more than 30 years, researchers and doctors from Russian, the U.S, France, Israel and other countries have studied methods of transcrebral or transcranial electrotherapy stimulation (TES). Transcranial impulse electrotherapy is a specific type of transcranial electrotherapy that uses low power impulse current to stimulate the antinoceptive system of the encephalon to provide procedures of electrodream and transcranial electroanagelsia. The stimulation current must be kept constant during treatment session. Traditionally, it can be done by employing close-loop control that continuously supervises the stimulation current. The fact that the impulse current for TES is low power can cause difficulties in measuring and controlling the current. Therefore, in this paper, we will introduce detailed design of current stabilization circuit that produces constant impulse current but doesn’t require closed loop control.

Application of Neural Network in Classifying Massive Lesions in Mammogram

Breast cancer is one of the most frequently leading causes of cancer deaths in middle-aged women. Until now, mammography is still the most effective procedure for early diagnosis of the breast cancer. Computer-aided detection (CAD) system can be very helpful for radiologists in identification abnormalities earlier and faster than traditional screening program. Therefore, efforts have been made to develop CAD systems for mammographic interpretation. The identification of massive lesions in mammograms is still challenging because of their irregular shapes and ill-defined margins and faint contrast. Regions-of-Interest (ROIs) that high probably contain massive lesions are first detected. They are then classified as mass region or non-mass region using their extracted features. In this paper we will investigate the application of neural network in classifying massive lesions in digital mammograms.

Mass Detection and Classification on Digitalized Mammograms

In this paper we will present a computer-aided detection (CAD) system for mass detection and classification. This CAD system performs mass detection on regions of interest (ROIs) and performs the normal-abnormal classification on detected masses. To enhance the ROIs, histogram equalization and average filtering were employed. ROIs or suspicious lesions were detected by an edge-base segmentation technique. Each ROI was represented by 12 textural features calculated from spatial gray-level co-occurrence matrix (GLCM). Finally, a back propagation neural network was used to classify a ROI to normal or abnormal one. Performance of the proposed system was analyzed in mini-MIAS data set by means of receiver operating characteristics curve (ROC curve) and free-response ROC curve (FROC curve). We archived area under the ROC curve AZ = 0,815. This value lied in the range from 0.7 to 0.9 so that our CAD system could be considered accurate enough.