We present a dynamic functional connectivity (DFC)-based classification analysis of functional magnetic resonance imaging (fMRI) data from veterans with a type of post-traumatic stress disorder (PTSD) and from matched normal control (NC) veterans. Whole-brain resting-state fMRI (rsfMRI) data which were scanned from 23 PTSD (mean age 49) and 30 NC (mean age 50) veterans were used for analyses. A computational method using statistics of DFC and support-vector machine (SVM) classifier were used to correctly classify PTSD vs NC with up to 98

Segmentation of Multiple Sclerosis (MS) lesions from Magnetic Resonance Imaging (MRI) data presents a significant challenge due to the necessity for large volumes of training data and a sophisticated training process. Traditional MRI datasets often lack the extensive sample sizes required for effective training, necessitating the exploration of alternative methods for accurate segmentation. This study proposes a robust machine learning algorithm designed to identify MS lesions using both single-modal and multi-modal MRI data. The proposed algorithm employs Convolutional Neural Networks (CNNs) in the form of U-Net architecture, a renowned model for biomedical image segmentation. To address the issue of insufficient training data, data augmentation techniques have been implemented, enhancing the diversity and volume of the training set. The dataset for this study was created from MRI data of 20 subjects. The algorithm's effectiveness was evaluated using the DSC score, a statistical tool that measures the similarity between two samples. The model achieved a DSC score of 0.7960 in the training set and 0.7912 in the test set, demonstrating its effectiveness in performing segmentation of MS from multi-modal MRI data. The predicted locations of MS lesions were compared with the corresponding layers of white matter, gray matter, and cerebrospinal fluid within the brain. This innovative approach aims to enhance the accuracy and efficiency of MS lesion segmentation, contributing to advancements in precision medicine and the overall understanding of MS.

Travelers are increasingly looking for customized itineraries that are suited to their specific interests. In response to this demand in this work, an automated travel planning system is offered to cater to personalized itinerary needs. It employs conversational Artificial Intelligence (AI), Chat Generative Pre-Trained Transformer (ChatGPT) to comprehend the user’s preferences and recommend appropriate destinations and activities. Playwright, a web scraper, is utilized to access real-time airline and hotel info from Google Travel. This data is used to train a content-based recommendation engine, which matches possible outcomes to the traveler’s budget and priorities. The algorithm generates personalized recommendations by taking into account essential parameters such as pricing, hotel ratings and fight layover. The precision score analysis validates the recommendations’ accuracy. The results demonstrate that the system provided accurate tailored recommendations with 86 percent precision. Based on user studies, comparative assessments revealed enhanced performance over existing technologies. User satisfaction was represented through customised suggestions that met their specific preferences and limits. The solution simplifies and improves the travel planning experience using AI-powered personalization that caters to individual preferences and limits.

Infrared imagery, like almost any other two-dimensional (2-D) imagery, have been traditionally sampled and acquired using a rectangular grid. Algebraic fixed-pattern noise/ nonuniformity correction (NUC) algorithms work on this traditional rectangular grid mitigating the most dominant, bias/offset portion of the nonuniformity. On the other hand, it is well-known that hexagonal sampling grid captures more information in sampled data/imagery when compared to traditional rectangular sampling, and a hexagonal addressing scheme for hexagonally-sampled imagery, namely array set addressing scheme, was recently developed. In this work, we derive the bilinear interpolation equations between two image frames for hexagonally-sampled infrared imagery with bias/offset nonuniformity under the 2-D global motion of the scene or the camera, and apply the 2-D algebraic NUC algorithm to hexagonally-sampled imagery. We present a simulation of mid-wave infrared imagery with hexagonally-sampled pixel array with bias nonuniformity under simulated global translational motion, and we test the efficiency of the NUC algorithm on the simulated infrared imagery and compare the performance of the hexagonally-sampled pixel array imagery NUC results to those of the traditional rectangularly-sampled pixel array imagery.

We present a conceptual design for an in-situ methane gas sensor that could be deployed rapidly to suspected sites of spurious methane emission. Based on the remote detection of suspected methane leaks now possible with a class of satellites currently in orbit or soon to be launched, in-situ sensors would be deployed to the location of the detection, and accurate measurements of methane leak rates would be reported. Our design is very high level at this time, but incorporates a spectral capability that allows switching on and off the wavelengths of peak methane emission to facilitate detection. Unlike the case for the satellite-based sensing of the gas, our design will lead to the quantification of smallest levels of methane leaks. Two approaches will be considered: ground sensors detecting methane in emission against a cooler sky background, and aerial sensors detecting the gas in absorption against the ground scene as a background source. Sensitivity plays a key role, with the infrared detector working in the 2.4-micron region and operating with near-theoretical sensitivity, with the limiting noise sources set by the background levels for the ground sensor. Consideration of levels of detector dark current, based on the background signal level, and required detector operation temperatures will be derived. The paper reports on the conceptual design, with details on the electronics approach needed to realize the needed levels of sensitivity. Performance quantification will be accomplished through simulation using accurate noise models.

Emotion recognition by artificial intelligence (AI) is a challenging task. A wide variety of research has been done, which demonstrated the utility of audio, imagery, and electroencephalography (EEG) data for automatic emotion recognition. This paper presents a new automated emotion recognition framework, which utilizes electroencephalography (EEG) signals. The proposed method is lightweight, and it consists of four major phases, which include: a reprocessing phase, a feature extraction phase, a feature dimension reduction phase, and a classification phase. A discrete wavelet transforms (DWT) based noise reduction method, which is hereby named multi scale principal component analysis (MSPCA), is utilized during the pre-processing phase, where a Symlets-4 filter is utilized for noise reduction. A tunable Q wavelet transform (TQWT) is utilized as feature extractor. Six different statistical methods are used for dimension reduction. In the classification step, rotation forest ensemble (RFE) classifier is utilized with different classification algorithms such as k-Nearest Neighbor (k-NN), support vector machine (SVM), artificial neural network (ANN), random forest (RF), and four different types of the decision tree (DT) algorithms. The proposed framework achieves over 93 % classification accuracy with RFE + SVM. The results clearly show that the proposed TQWT and RFE based emotion recognition framework is an effective approach for emotion recognition using EEG signals.

Infrared imagery, like almost any other two-dimensional (2D) imagery, have been traditionally sampled and acquired using a traditional rectangular grid. Therefore, nonuniformity correction (NUC) algorithms for infrared imaging systems which mitigate the most dominant, bias/offset portion of the nonuniformity were developed on the rectangular grid. However, it is well-known that hexagonal sampling grid captures more information in sampled data/imagery when compared to traditional rectangular sampling, and a hexagonal addressing scheme (HAS) for hexagonally-sampled imagery to convert imagery between the two different coordinate systems was developed. In this work, we build on prior work by Sakoglu et al. who developed bilinear interpolation equations between two image frames under the 2-D global motion of the scene or the camera, and apply this 2D algebraic NUC algorithm to hexagonally-sampled imagery directly in the HAS domain by utilizing simulated hexagonal sampling of real IR images.

A diversified infrared technology base has been developed over the recent decades for various civilian and military sensing needs. The technology has been optimized to balance performance and affordability constraints for a variety of end-use goals. Simplistically, these goals might involve the detection and measurement of nearby, bright sources that fill even the largest angular fields-of- view of pixels in simple, low-magnification systems for which abundant signal makes possible infrared detection and measurement with less-sensitive, uncooled sensor arrays. At another extreme are ultra-cryogenically-cooled systems operating below thermoelectric cooler capabilities and which enable the detection and measurement of much fainter sources that underfill even the tiny angular pixel fields of view set by the diffraction limit of large, high magnification optical systems. Our emphasis is closer to the latter for the applications described here. As one example of the environmental monitoring capabilities made possible in the infrared, gas leak detection in transmission pipelines is vitally important for safe operation and for protecting the environment by accounting for and assessing the impact of leaks that adversely affect climate change. Gas leak detection in the infrared spectrum is facilitated by the distinctive spectral fingerprints of fundamental molecular vibrational modes which can be exploited for the detection of the gas. Sensitivity becomes paramount for many applications requiring faint signal detection, and large sensor array formats facilitate surveillance coverage. Many climate change assessments are expected to involve wide-area coverage of Earth scenes with revisit times sufficiently short to capture important transitory events. Shorter term monitoring of containment compliance requires detecting sufficiently small gas leak flows over broad expanses of the Earth's surface with high detection sensitivities. In this paper we described supporting technologies in the areas of sensor arrays and optical sub-systems, with an emphasis on dispersive spectrometers. There are a plethora of applications involving the stewardship of a range of biological assets, both in the ocean and on land environments, as well as large-scale sensing of atmospheric properties, including concentrations of greenhouse gases.

In this work, we develop an adaptive, near-optimal, 3-Dimensional (3D) to 1D ordering methodology for brain magnetic resonance imaging (MRI) data, using a space-filling curve (SFC) trajectory, which is adaptive to brain's shape as captured by MRI. We present the pseudocode of the heuristics for developing the SFC trajectory. We apply this trajectory to functional MRI brain activation maps from a schizophrenia study, compress the data, obtain features, and perform classification of schizophrenia patients vs. normal controls. We compare the classification results with those of a linear ordering trajectory, which has been the traditional method for ordering 3D MRI data to 1D. We report that the adaptive SFC trajectory-based classification performance is superior than the linear ordering trajectory-based classification.

Previously, Ratliff et al. and Sakoglu et al. developed algebraic nonuniformity correction (NUC) algorithms (the latter developed a matrix-based version with regularization capabilities) which mitigate fixed-pattern nonuniformity (noise) that is notoriously present in infrared image sequences/videos, by utilizing global translational motion of the scene or the imaging camera system. Infrared imagery, like almost any other two-dimensional (2-D) imagery, have been traditionally sampled and acquired using a rectangular grid, therefore the developed NUC algorithms work on this traditional rectangular grid mitigating the most dominant, bias/offset portion of the nonuniformity. On the other hand, it is well-known that hexagonal sampling grid captures more information in sampled data/imagery when compared to traditional rectangular sampling, and a hexagonal addressing scheme for hexagonally-sampled imagery, namely array set addressing scheme, was recently developed by Rummelt et al. in order to be able to convert imagery between the two different coordinate systems and to perform various mathematical and image processing operations. In this work, we derive the bilinear interpolation equations between two image frames for hexagonally-sampled infrared imagery with bias/offset nonuniformity under the 2-D global motion of the scene or the camera, and apply the 2-D algebraic NUC algorithm to hexagonally-sampled imagery. We present a simulation of MWIR infrared imagery with hexagonally-sampled pixel array, with global motion of the scene and with bias/offset nonuniformity, and we test the efficiency of the NUC algorithm on the simulated infrared imagery (based on real MWIR infrared imagery) and compare the performance of the hexagonally-sampled pixel array imagery NUC results to those of the traditional rectangularly-sampled pixel array imagery.

BackgroundMultiple sclerosis (MS) is an immune-mediated inflammatory disease that attacks myelinated axons in the central nervous system, destroying myelin and axons to variable degrees and, resulting in significant physical disability. Magnetic resonance imaging (MRI) is useful in the diagnosis of MS, surpassing all other imaging techniques in terms of prediction accuracy. Depending on the number and location of lesions, however, the success of MR can vary significantly in terms of sensitivity and specificity in the diagnosis of MS. Adverse effects of various intensity and residual artifacts in the MRI data make it challenging to compute MS lesion volume to assess the progression of MS. Therefore, the development of robust and automated MS lesion detection methods has been a challenge.ObjectivesThis study aims to develop a novel, robust, and simple image segmentation method to perform quantitative analysis of MS lesions from multimodal MRI data.MethodsAn algorithm based on a supervised minimum Euclidean distance-based clustering method employing three 2D MRI modalities, T1-weighted (T1w), fluid-attenuated inversion recovery (FLAIR), and T2-weighted (T2w) MRI was developed for classification of significant brain tissues and MS lesions. The developed method was applied to an MRI dataset from six MS patients.ResultsThe developed method classifies various brain tissues and detects MS lesions with over 90% accuracy and specificity, and 62%–65% sensitivity, on average.ConclusionsSegmentation of different brain tissues using our proposed algorithm results in superior MS lesion-detection accuracy, comparable with the recent deep-learning classification results in the literature.

The purpose of this project was to analyze which image pre-processing technique was most beneficial in improving the performance of Facial Expression Recognition through Deep Learning and High-Performance Computing. Contrary to our expectations, the results obtained in this work showed that deep learning does not significantly benefit from various commonly used image pre-processing techniques such as resizing, smoothing, or edge detection. The results confirm previous findings that an increase in accuracy is obtained by increasing the size of the training dataset. This study proceeds to show that the increase in training data size can easily be handled by the High-Performance Computing (HPC) cluster provided by the Pittsburg Supercomputing Center (PSC) through XSEDE.

Around 200,000 veterans (up to 32% of those deployed) of the 1991 Gulf War (GW) suffer from GW illness (GWI), which is characterized by multiple deficits in cognitive, affective, sensory and nociception domains. In this study we employed resting state fMRI (rsfMRI) to map impairments in brain function in GWI with advanced network analysis. RsfMRI data was obtained from 60 GWI veterans and 30 age-matched military controls. Group independent component analysis (GICA) was conducted to probe the functional connectivity networks in all 90 subjects. GICA revealed impaired functional connectivity (FC) in GWI veterans between a number of brain function networks consistent with their self-reported symptoms. GWI veterans exhibited impaired FC between language networks, and sensory input networks of all modalities as well as motor output networks. GWI veterans also exhibited impaired FC between different sensory perception and motor networks, and between different networks in the sensorimotor domain. These FC impairments provide putative mechanism of central nervous system dysfunction in GWI.

Static functional connectivity (FC) analyses based on functional magnetic resonance imaging (fMRI) data have been extensively explored for studying various psychiatric conditions in the brain, including cocaine addiction. A recently emerging, more powerful technique, dynamic functional connectivity (DFC), studies how the FC dynamics change during the course of the fMRI experiments. The aim in this paper was to develop a computational approach, using a machine learning framework, to determine if DFC features were more successful than FC features in the classification of cocaine-dependent patients and healthy controls. fMRI data were obtained from of 25 healthy and 58 cocaine-dependent participants while performing a motor response inhibition task, stop signal task. Group independent component analysis was carried out on all participant data to compute spatially independent components (ICs). Eight ICs were selected manually as relevant brain networks, which were used to classify healthy versus cocaine-dependent participants. FC and DFC measures of the chosen IC pairs were used as features for the classification algorithm. Support Vector Machines were used for both feature selection/reduction and participant classification. Based on DFC with only seven IC pairs, participants were successfully classified with 95% accuracy (and with 90% accuracy with three IC pairs), whereas static FC yielded only 81% accuracy. Visual, sensorimotor, default mode, and executive control networks, amygdala, and insula played the most significant role in the DFC-based classification. These findings support the use of DFC-based classification of fMRI data as a potential biomarker for the identification of cocaine dependence.

Neuroimaging studies have yielded significant advances in the understanding of neural processes relevant to the development and persistence of addiction. However, these advances have not explored extensively for diagnostic accuracy in human subjects. The aim of this study was to develop a statistical approach, using a machine learning framework, to correctly classify brain images of cocaine-dependent participants and healthy controls. In this study, a framework suitable for educing potential brain regions that differed between the two groups was developed and implemented. Single Photon Emission Computerized Tomography (SPECT) images obtained during rest or a saline infusion in three cohorts of 2–4 week abstinent cocaine-dependent participants (n = 93) and healthy controls (n = 69) were used to develop a classification model. An information theoretic-based feature selection algorithm was first conducted to reduce the number of voxels. A density-based clustering algorithm was then used to form spatially connected voxel clouds in three-dimensional space. A statistical classifier, Support Vectors Machine (SVM), was then used for participant classification. Statistically insignificant voxels of spatially connected brain regions were removed iteratively and classification accuracy was reported through the iterations.

Intense computations in engineering and science, especially bioinformatics have been made practical by the recent advances in Graphical Processing Unit (GPU) computing technology. In this study, implementation and performance evaluations for a GPU-accelerated dynamic functional connectivity (DFC) analysis, which is an analysis method for investigating dynamic interactions among different brain networks, is presented. Open Computing Library (OpenCL), which provides a general framework for GPU computing, is utilized, and it is shown to reduce the DFC analysis computation time. The parallel implementation with OpenCL provides up to 10x speed-up over sequential implementation.

In this paper we introduce a method to develop knowledge bases for medical decision support systems, with a focus on evaluating such knowledge bases. Departing from earlier efforts with concept maps, we developed an ontological-semantic knowledge base and evaluated its information content using the metrics we have developed, and then compared the results to the UMLS backbone knowledge base. The evaluation method developed uses information entropy of concepts, but in contrast to previous approaches normalizes it against the number of relations to evaluate the information density of knowledge bases of varying sizes. A detailed description of the knowledge base development and evaluation is discussed using the underlying algorithms, and the results of experimentation of the methods are explained. The main evaluation results show that the normalized metric provides a balanced method for assessment and that our knowledge base is strong, despite having fewer relationships, is more information-dense, and hence more useful. The key contributions in the area of developing expert systems detailed in this paper include: (a) introduction of a normalized entropy-based evaluation technique to evaluate knowledge bases using graph theory, (b) results of the experimentation of the use of this technique on existing knowledge bases. (C) 2015 Elsevier Ltd. All rights reserved.

In this paper, a MATLAB-based graphical user interface (GUI) software tool for general biomedical signal processing and analysis of functional neuroimaging data is introduced. Specifically, electroencephalography (EEG) and electrocardiography (ECG) signals can be processed and analyzed by the developed tool, which incorporates commonly used temporal and frequency analysis methods. In addition to common methods, the tool also provides non-linear chaos analysis with Lyapunov exponents and entropies; multivariate analysis with principal and independent component analyses; and pattern classification with discriminant analysis. This tool can also be utilized for training in biomedical engineering education. This easy-to-use and easy-to-learn, intuitive tool is described in detail in this paper.