Amirhessam Tahmassebi

An evolutionary approach for fMRI big data classification

Resting-state function magnetic resonance imaging (fMRI) images allow us to see the level of activity in a patients brain. We consider fMRI of patients before and after they underwent a smoking cessation treatment. Two classes of patients have been studied here, that one took the drug N-acetylcysteine and the ones took a placebo. Our goal was to classify the relapse in nicotine-dependent patients as treatment or non-treatment based on their fMRI scans. The image slices of brain are used as the variable and as results here we deal with a big data problem with about 240,000 inputs. To handle this problem, the data had to be reduced and the first process in doing that was to create a mask to apply to all images. The mask was created by averaging the before images for all patients and selecting the top 40% of voxels from that average. This mask was then applied to all fMRI images for all patients. The average of the difference in the before treatment and after fMRI images for each patient were found and these were flattened to one dimension. Then a matrix was made by stacking these 1D arrays on top of each other and a data reduction algorithm was applied on it. Lastly, this matrix was fed into some machine learning and Genetic Programming algorithms and leave-one-out cross-validation was used to test the accuracy. Out of all the data reduction machine learning algorithms used, the best accuracy was obtained using Principal Component Analysis along with Genetic Programming classifier. This gave an accuracy of 74%, which we consider significant enough to suggest that there is a difference in the resting-state fMRI images of a smoker that undergoes this smoking cessation treatment compared to a smoker that receives a placebo.

Dynamical graph theory networks techniques for the analysis of sparse connectivity networks in dementia

Graph network models in dementia have become an important computational technique in neuroscience to study fundamental organizational principles of brain structure and function of neurodegenerative diseases such as dementia. The graph connectivity is reflected in the connectome, the complete set of structural and functional connections of the graph network, which is mostly based on simple Pearson correlation links. In contrast to simple Pearson correlation networks, the partial correlations (PC) only identify direct correlations while indirect associations are eliminated. In addition to this, the state-of-the-art techniques in brain research are based on static graph theory, which is unable to capture the dynamic behavior of the brain connectivity, as it alters with disease evolution. We propose a new research avenue in neuroimaging connectomics based on combining dynamic graph network theory and modeling strategies at different time scales. We present the theoretical framework for area aggregation and time-scale modeling in brain networks as they pertain to disease evolution in dementia. This novel paradigm is extremely powerful, since we can derive both static parameters pertaining to node and area parameters, as well as dynamic parameters, such as system’s eigenvalues. By implementing and analyzing dynamically both disease driven PC-networks and regular concentration networks, we reveal differences in the structure of these network that play an important role in the temporal evolution of this disease. The described research is key to advance biomedical research on novel disease prediction trajectories and dementia therapies.

High Performance GP-Based Approach for fMRI Big Data Classification

We consider resting-state Functional Magnetic Resonance Imaging (fMRI) of two classes of patients: one that took the drug N-acetylcysteine (NAC) and the other one a placebo before and after a smoking cessation treatment. Our goal was to classify the relapse in nicotine-dependent patients as treatment or non-treatment based on their fMRI scans. 80% accuracy was obtained using Independent Component Analysis (ICA) along with Genetic Programming (GP) classifier using High Performance Computing (HPC) which we consider significant enough to suggest that there is a difference in the resting-state fMRI images of a smoker that undergoes this smoking cessation treatment compared to a smoker that receives a placebo.

iDeepLe: deep learning in a flash

Emerging as one of the most contemporary machine learning techniques, deep learning has shown success in areas such as image classification, speech recognition, and even playing games through the use of hierarchical architecture which includes many layers of non-linear information. In this paper, a powerful deep learning pipeline, intelligent deep learning (iDeepLe) is proposed for both regression and classification tasks. iDeepLe is written in Python with the help of various API libraries such as Keras, TensorFlow, and Scikit-Learn. The core idea of the pipeline is inspired by the sequential modeling with considering numerous layers of neurons to build the deep architecture. Each layer in the sequential deep model can perform independently as a module with minimum finitudes and does not limit the performance of the other layers. iDeepLe has the ability of employing grid search, random search, and Bayesian optimization to tune the most significant predictor input variables and hyper-parameters in the deep model via adaptive learning rate optimization algorithms for both accuracy and complexity, while simultaneously solving the unknown parameters of the regression or the classification model. The parallel pipeline of iDeepLe has the capacity to handle big data problems using Apache Spark, Apache Arrow, High Performance Computing (HPC) and GPU-enabled machines as well. In this paper, to show the importance of the optimization in deep learning, an exhaustive study of the impact of hyper-parameters in a simple and a deep model using optimization algorithms with adaptive learning rate was carried out.

Genetic Programming Based on Error Decomposition: A Big Data Approach

An investigation of the deviations of error and correlation for different stages of the multi-stage genetic programming (MSGP) algorithm in multivariate nonlinear problems is presented. The MSGP algorithm consists of two main stages: (1) incorporating the individual effect of the predictor variables, (2) incorporating the interactions among the predictor variables. The MSGP algorithm formulates these two terms in an efficient procedure to optimize the error among the predicted and the actual values. In addition to this, the proposed pipeline of the MSGP algorithm is implemented with a combination of parallel processing algorithms to run multiple jobs at the same time. To demonstrate the capabilities of the MSGP, its performance is compared with standard GP in modeling a regression problem. The results illustrate that the MSGP algorithm outperforms standard GP in terms of accuracy, efficiency, and computational cost.

Determining disease evolution driver nodes in dementia networks

Imaging connectomics emerged as an important field in modern neuroimaging to represent the interaction of structural and functional brain areas. Static graph networks are the mathematical structure to capture these interactions modeled by Pearson correlations between the representative area signals. Dynamical functional resting state networks seen in most fMRI experiments can not be represented by the classic correlation graph network. The changes in brain connectivity observed in many neuro-degenerative diseases in longitudinal data series suggest that more sophisticated graph networks to capture the dynamical properties of the brain networks are required. Furthermore, certain brain areas seem to act as ”disease epicenters” being responsible for the spread of neuro-degenerative diseases. To mathematically describe these aspects, we propose a novel framework based on pinning controllability applied to dynamic graphs and seek to determine the changes in the ”driver nodes” during the course of the disease. In contrast to other current research in pinning controllability, we aim to identify the best driver nodes describing disease evolution with respect to connectivity changes and location of the best driver nodes in functional 18F-Fluorodeoxyglucose Positron Emission Tomography (18FDG-PET) and structural Magnetic Resonance Imaging (MRI) connectivity graphs in healthy controls (CN), and patients with mild cognitive impairment (MCI), and Alzheimer’s disease (AD). We present the theoretical framework for determining the best driver nodes in connectivity graphs and their relation to disease evolution in dementia. We revolutionize the current graph analysis in brain networks and apply the concept of dynamic graph theory in connection with pinning controllability to reveal differences in the location of ”disease epicenters” that play an important role in the temporal evolution of dementia. The described research will constitute a leap in biomedical research related to novel disease prediction trajectories and precision dementia therapies.

Optimized Naive-Bayes and Decision Tree Approaches for fMRI Smoking Cessation Classification

This paper aims at developing new theory-driven biomarkers by implementing and evaluating novel techniques from resting-state scans that can be used in relapse prediction for nicotine-dependent patients and future treatment efficacy. Two classes of patients were studied. One class took the drug N-acetylcysteine and the other class took a placebo. Then, the patients underwent a double-blind smoking cessation treatment and the resting-state fMRI scans of their brains before and after treatment were recorded. The scientific research goal of this study was to interpret the fMRI connectivity maps based on machine learning algorithms to predict the patient who will relapse and the one who will not. In this regard, the feature matrix was extracted from the image slices of brain employing voxel selection schemes and data reduction algorithms. Then, the feature matrix was fed into the machine learning classifiers including optimized CART decision tree and Naive-Bayes classifier with standard and optimized implementation employing 10-fold cross-validation. Out of all the data reduction techniques and the machine learning algorithms employed, the best accuracy was obtained using the singular value decomposition along with the optimized Naive-Bayes classifier. This gave an accuracy of 93% with sensitivity-specificity of 99% which suggests that the relapse in nicotine-dependent patients can be predicted based on the resting-state fMRI images. The use of these approaches may result in clinical applications in the future.

Wearable biosignal acquisition system for decision aid

This article presents a prototype of a wearable instrument for oxygen saturation and ECG monitoring. The proposed measuring system is based on the variability of the light reflection of a LED emission placed on the subject’s temple. Besides, the system has the capacity to incorporate electrodes to obtain ECG measurements. The activity of the user can be monitored through an accelerometer. All measurements are stored and transmitted to a mobile device (tablet or smartphone) through a Bluetooth link where the information is treated and shown to the user.

Multi-level analysis of spatio-temporal features in non-mass enhancing breast tumors

Diagnostically challenging breast tumors and Non-Mass-Enhancing (NME) lesions are often characterized by spatial and temporal heterogeneity, thus difficult to detect and classify. Differently from mass enhancing tumors they have an atypical temporal enhancement behavior that does not enable a straight-forward lesion classification into benign or malignant. The poorly defined margins do not support a concise shape description thus impacting morphological characterizations. A multi-level analysis strategy capturing the features of Non-Mass- Like-Enhancing (NMLEs) is shown to be superior to other methods relying only on morphological and kinetic information. In addition to this, the NMLE features such as NMLE distribution types and NMLE enhancement pattern, can be employed in radomics analysis to make robust models in the early prediction of the response to neo-adjuvant chemotherapy in breast cancer. Therefore, this could predict treatment response early in therapy to identify women who do not benefit from cytotoxic therapy.

Reconfigurable instrument for measuring variations of capacitor's dielectric: an application to olive oil quality monitoring

The current method for the extraction of olive oil consists on the use of a decanter to split it by centrifugation. During this process, different olive oil samples are analyzed in a chemical laboratory in order to determine moisture levels in the oil, which is a decisive factor in olive oil quality. However, these analyses are usually both costly and slow. The developed prototype is the foundation of an instrument for real-time monitoring of moisture in olive oil. Using the olive oil as the dielectric of a parallel-plate capacitor, a model to relate the moisture in olive oil and capacitance has been created. One of the challenges for this application is the moisture range, which is usually between 1 and 2%, thus requiring the detection of pF-order variations in capacitance. This capacitance also depends on plate size and the distance between plates. The presented prototype, which is based on a PSoC (Programmable System-on-Chip), includes a reconfigurable digital and analog subsystem, which makes the determination of moisture independent of the capacitor. Finally, the measure is also sent to a smartphone via Bluetooth.