This work presents volumetric color medical image compression for the Picture Archiving and Communication Systems (PACS) using AT-SPIHT algorithm. PACS has inclined the medical information system by integrating RIS, HIS and all the imaging modalities and scanning devices to PACS so that doctor's use the PACS efficiently for diagnosis and decision making in petite time. PACS has the potential usage in telemedicine applications which is the future in medical field. With the above technology, enormous digital images are generated and stored in PACS for the doctor's diagnostic purpose where volumetric color medical image compression is imperative and to utilize the PACS competently. The volumetric color medical images are a three-dimensional (3-D) image data set, and can be considered as a sequence of two- dimensional (2-D) slices. AT-SPIHT applied to the 3-D color Computed Tomography (CT) medical image is an efficient progressive coding technique. AT-SPIHT algorithm utilizes a more efficient spatial orientation tree structure. This novel technique enables the PACS implementation and deployment in medical information system.

View full abstract

In this paper, we have proposed a Gabor filter technique for Computed Tomography image which is used for early diagnosis of human brain infarct. This technique aims to help the doctor in clinical diagnosis of stroke. The process involves design of a Gabor filter tuned to different spatial-frequencies and orientations to cover the spatial-frequency space and histogram based thresholding. The results from this technique are validated with the conventional diagnosis procedures and with radiologists.

View full abstract

In this paper, we present a proposed method for the real-time quantification of mental workload using the ECG signal. One unique feature of the proposed method is the so-called "relative" measure-a measure relative to individuals and tasks which the individuals perform. Another relatively new feature with the method is the application of the wavelet transform technique to extract ECG signals. Further, the artificial neural network technique, especially the competitive neural network architecture together with the clustering technique, is employed. The clustering serves for two purposes: (1) to determine the relative measure the number of levels or scales of the measure and (2) to be a part of the competitive neural network. The vehicle driving is used as an example to illustrate and validate the method.

View full abstract

The transcription technique of the acoustic structure has been studied, by modifying the conventional sampling methodology. The maximum and minimum data of the acoustic waveform play a key role in the reproduction of sound, maintaining the pitch, loudness and timbre. The acoustic structure has been simply modeled, in which ten partial sums of the Fourier series representing the audible sound. Because one of the partial sums seems like a zigzag curve, a virtual acoustic particle with energy has been assumed between the two successive extrema, hypothesizing that the monotonous curve between the two was not perceptible. The model has been evaluated by both the sum squared error and the reduction of the data size by the removal operation of a series of acoustic particles with trivial energy, without an apriori knowledge or the technique of the provisional threshold. Consequently, the extremal sampling technique at the passing bandwidth of 80 to 5,120Hz is approximately equivalent to the conventional sampling with less than 1% of the sum squared error, and has an advantage of the analysis of noise structure with 3.6% reduction of the speech data size and 1.3% for even more complex speech accompanying music.

View full abstract

Recently, little attention has been paid to EEG signal for emotion recognition when compared to other physiological signals. This paper proposes an emotion recognition system from EEG (Electroencephalogram) signals. We designed an efficient acquisition protocol for acquiring the EEG signals under audio-visual induction environment for participants. Totally, 6 healthy subjects with an age group of 21-27 using 63 biosensors are used for registering the EEG signal for various emotions. After preprocessing the signals, discrete wavelet transform is employed to extract the EEG parameters. These extracted features are classified into discrete emotions using Fuzzy C-Means (FCM) clustering. Results confirm the possibility of using different wavelet transform based feature extraction for assessing the human emotions from EEG signal.

View full abstract

Segmentation technique has fast becoming popular for signal and image processing in the last few decades. This study investigates the application of conventional seed based region growing (SBRG) as a segmentation algorithm. Conventional SBRG algorithm offers several advantages over other segmentation techniques. However, segmentation process by using SBRG algorithm need to be repeated according to the number of region presence in an image. Therefore, a modified SBRG algorithm to automatically segment all regions of interest in an image is proposed. Firstly, the proposed technique employed clustering algorithm to determine the threshold value. Then, the process was continued with moments calculations to find a point having (x,y) coordinate in every region of interest in an image. To start the segmentation process, these coordinates were applied as position of seed pixel in the image. The capability of the modified algorithm was tested on the ThinPrep^[○!R] images. The proposed algorithm yields a promising result where the seed pixels were successfully placed within the nucleus region in every region of interests (cervical cells) and all cervical cells presence in ThinPrep^[○!R] image were successfully classified and distinguished from the background area.

View full abstract

Motor imagery is the mental simulation of a motor act that includes preparation for movement, passive observations of action and mental operations of motor representations implicitly or explicitly. The ability of an individual to control his EEG through imaginary mental tasks enables him to control devices through a brain machine interface. Brain machine interfaces are used to rehabilitate people suffering from neuromuscular disorders as a means of communication or control. A brain machine interface design using PSO Elman Neural Network (PSOENN-BMI) is proposed to discriminate EEG signals acquired during motor imagery for left and right hand movements. EEG is recorded at the C3 and C4 locations using noninvasive scalp electrodes placed over the motor cortex. The performance of the three state PSOENN-BMI is tested with two feature sets namely band power (BP) and principal component analysis (PCA) features. From the results it is observed that the performance of the PSOENN-BMI is better when the PCA features are used with an average efficiency range of 74.85% to 84.96%.

View full abstract

Nowadays voice disorders are increasing dramatically due to the modern way of life. Most of the voice disorders cause changes in the voice signal. Acoustic analysis on the speech signal could be a useful tool for diagnosing voice disorders. This paper applies Mel-scaled wavelet packet transform (Mel-scaled WPT) based features to perform accurate diagnosis of voice disorders. A Functional Link Neural Network (FLNN) is developed to test the usefulness of the suggested features. Two simple modifications are newly proposed in the FLNN architecture to improve the classification accuracy. In the first architecture, a hidden layer is newly introduced in a FLNN and trained by Back Propagation (BP) procedure. In the second architecture, the Integral and Derivative controller concepts are introduced to the neurons in the hidden layer and the network is trained by BP procedure. The performance is compared with conventional neural network model. The results prove that the proposed FLNN gives very promising classification accuracy and suggested features can be employed clinically to diagnose the voice disorders.

View full abstract

We present an application level framework which makes use of Wavelet Packet Analysis (WPA) for improved target detection in oddball paradigm, which are being researched for a brain biometric system. The novelty lies in the usage of both P300 (delta and theta band) and gamma band features from a wavelet perspective using just forty trials. The features were extracted using WPA analysis for target detection, wherein Daubechies (Db4) and Coiflet (Coif3) wavelets are used respectively to extract the P300 and Gamma band energy features. A comparison on the classification accuracy is also presented when the P300 features are used with and without Gamma band features. This work also discusses a new dynamic backward referencing technique which seems to enhance the features (delta, theta and gamma band) from eight channels. A Radial Basis Function (RBF) classifier is used to classes the obtained features as target and non-target for both the paradigms. Initial results on these lines from four subjects show motivating results for further time frequency research.

View full abstract

In this paper we present a novel approach based on discrete wavelet transform for the removal of slow baseline drift, power line interference and white noise of electrocardiographic (ECG) signals. These unwanted extrinsic components are efficiently removed by zeroing scaling and wavelet coefficients of the discrete wavelet transform (hard thresholding). This pre-processing approach can filter the unwanted components without introducing distortions in the ECG waveform. In addition, it can easily be combined with other wavelet-based techniques for analysis and classification of ECG signals. First, we will describe in detail the design of a three-channel redundant wavelet decomposition (comprised of a low-pass as well as a band-pass and a high-pass filter) with linear phase properties that outperform critically sampled non-linear phase wavelets by generating fewer artifacts. Then we will present several examples of baseline drift removal, power line interference and white noise reduction for different levels of baseline drift and noise contamination in the ECG signals.

View full abstract

Multiple classifier combination is a technique that combines the decisions of different classifiers. Combination can reduce the execution time of classification, variance of estimation errors, thereby improving the overall classification accuracy. This paper introduces a genetic algorithm able to combine three different classifiers, fuzzy, MLP, K-NN. The use of a genetic algorithm is motivated by the fact that the combination phase is based on the optimization of vote strategy. The method has been applied to classification of The Pima Indians Diabetes database, results show a significant improvement of recognition accuracy using the genetic algorithm combination strategy compared with the recognition accuracy of each single classifier.

View full abstract

Dysarthria is a neuromotor impairment of speech that affects one or more of the speech sub-systems. It is reflected in the acoustic characteristics of the phonemes as deviations from their healthy counterparts. To capture these deviations, in this work a continuous speech, an isolated-style monophone-based, and a triphone-based speech recognition systems are developed. These speech recognition systems are trained with the TIMIT speech corpus and tested with the Nemours database of dysarthric speech. The correlation coefficient between the performance of the speech recognition systems and the Frenchay dysarthria assessment (FDA) scores is computed for the assessment of articulatory sub-systems. It is observed that triphone-based system after necessary phoneme grouping based on place of articulation correlates well with the FDA scores. It is further observed that apart from the articulatory problems, some of the speakers are affected with velopharyngeal incompetence also. It is analyzed with group delay function-based acoustic measure for the detection of hypernasality on dysarthric speech and found that 4 out of 10 dysarthric speakers in the Nemours database are hypernasal.

View full abstract

Drivers' drowsiness is one of the main contributing factors in fatal collisions. Early detection of drivers' drowsiness accurately will have a high potential to improve traffic safety. Electroencephalography (EEG), known to reflect the brain activity, has the potential of being an effective indicator of drowsiness. This paper presents an experimental study on the sensitivity of EEG signals to drowsiness in driving applications and on the influence of the probe or scalp location for the EEG signal acquisition. The study recorded participants' EEGs in alert states or drowsy states for a driving task. Features of EEG such as Theta wave, Alpha wave, Beta wave, and Gamma wave were extracted from the EEGs and then analyzed for their power of discriminating the alert state and drowsy state. The study concluded that (1) these features of EEG can be used to find a drowsy state and (2) the location of the scalp will significantly affect the measurement accuracy.

View full abstract

One important aspect of dentofacial image processing is the utilization of comprehensive oral and maxillofacial imaging modalities that aid in the assessment, treatment planning and evaluation of oral diseases. Also, it plays a vital role in identifying the pathological conditions, and it improves the practice management of the doctors for diagnosis. In this paper Radial Basis Function (RBF) Neural Network for automated diagnosis of cysts in dental X-ray Images is proposed. The severity of the cysts is then evaluated based on gray level properties, circularity and area. The results obtained provide the complete information about severity of the cysts by increasing the diagnostic ease of the dental surgeon.

View full abstract

In this study, we demonstrate that we can use non-invasive physiology sensing to detect stress level and lying, within the context of structured poker game settings. In our study, we instrument players with non-invasive physiological sensors that can measure heart rate, temperature, skin conductance, movement, and audio (voice/speech) during live play in real-money, no-limit Hold'em poker tournaments. We show how simply derived physiological features such as voice pitch variation, skin conductance peaks, and heart rate variability are correlated to a number of high-stress situations found in these type of tournaments. Using these features, we can develop simple classifiers that can accurately identify stress and bluffing in these settings.

View full abstract