Sajib Saha

Compressed sensing inspired rapid algebraic reconstruction technique for computed tomography

In this paper, we present an innovative compressive sensing based iterative algorithm for tomographic reconstruction. Back-projection has been customized to make it work even when the projections are not uniformly distributed, and thus ensures a better initial guess to start ART iterations. Contour information of the object has been used efficiently for faster and finer reconstruction. Aiming successful reconstruction with minimum number of iterations, conjugate gradient method that enjoys the full benefit of ART with good initial guess has been used instead of commonly used steepest descent method. Based on the experiments on simulated and real medical images it has been shown that the proposed modality is capable of producing much better reconstruction than the state-of-the-art methods.

A New Imaging Method for Real-time 3D X-ray Reconstruction

Existing Computed Tomography (CT) systems are vulnerable to internal organ movements. This drawback is compensated by extra exposures and digital processing. CT being a radiation dose intensive modality, it is imperative to limit the patient’s exposure to X-ray radiation, if only by removing the necessity to take extra exposures. A multiple pinhole camera, akin to optical lightfield imaging, to acquire simultaneously multiple X-ray projections is presented. This new method allows a single snapshot acquisition of all necessary projections for 3D reconstruction. It will also allow the real-time dynamic 3D X-ray reconstruction of moving organs, as it requires no scanning and no moving parts in its final implementation. A proof-of-concept apparatus that simulates the intended process was built and parallaxed images were obtained with minor processing. Synthetic 3D reconstruction tests are also presented.

Perceptual Dissimilarity: A Measure to Quantify the Degradation of Medical Images

This paper introduces a novel image quality metric that outperforms other existing and widely used metrics, namely the Root Mean Square Error (RMSE). While objective methods for assessing perceptual image quality to quantify the visibility of errors between a distorted image and a reference image have got utmost importance nowadays, none of the proposed metrics are completely satisfactory. Structural Similarity Index has been found promising in some cases, however for medical imaging (especially for CT and MRI images) it still remains immature. On that perspective, the Perceptual Dissimilarity (PD) measure proposed in this paper has been found satisfactory enough to work with medical imaging demands.

Multi-axial CT reconstruction from few view projections

This paper focuses on tomographic reconstruction from a smaller number of projections than usual. Whereas traditional CT scanner are based on sequential X-ray sources, the proposed methodology in this work is based on simultaneous x-ray sources on each projection. Simulations have shown that only four projections are needed to reconstruct a slice, which are captured simultaneously, offering drastic reduction of image capture time. Algebraic Reconstruction Technique (ART) has been used for reconstruction. Although ART has many advantages over the established methods, it remained unpopular due to its high computational cost, and most importantly due to the artefacts caused by the patients movement during image capture. The simultaneity of the projections helps to overcome this serious shortcoming of ART.

Evaluation of spatial resolution and noise sensitivity of sLORETA method for EEG source localization using low-density headsets

Electroencephalography (EEG) has enjoyed considerable attention over the past century and has been applied for diagnosis of epilepsy, stroke, traumatic brain injury and other disorders where 3D localization of electrical activity in the brain is potentially of great diagnostic value. In this study we evaluate the precision and accuracy of spatial localization of electrical activity in the brain delivered by a popular reconstruction technique sLORETA applied to EEG data collected by two commonly used low-density headsets with 14 and 19 measurement channels, respectively. Numerical experiments were performed for a realistic head model obtained by segmentation of MRI images. The EEG source localization study was conducted with a simulated single active dipole, as well as with two spatially separated simultaneously active dipoles, as a function of dipole positions across the neocortex, with several different noise levels in the EEG signals registered on the scalp. The results indicate that while the reconstruction accuracy and precision of the sLORETA method are consistently high in the case of a single active dipole, even with the low-density EEG configurations considered in the present study, successful localization is much more problematic in the case of two simultaneously active dipoles. The quantitative analysis of the width of the reconstructed distributions of the electrical activity allows us to specify the lower bound for the spatial resolution of the sLORETA-based 3D source localization in the considered cases.

Perceptual dissimilarity metric: A full reference objective image quality measure to quantify the degradation of perceptual image quality

This paper introduces a full reference objective image quality measure to quantify the degradation of perceptual image quality. Objective methods for assessing perceptual image quality are important for many image processing applications, such as monitoring and controlling image quality for quality control systems, benchmarking image processing systems and so on. The novel image quality metric proposed in this paper uses a relatively small number of pair-wise intensity comparisons to represent a patch as binary string, then compares corresponding patches using Hamming distances. It then calculates a dissimilarity value between images as an average of the Hamming distances computed between patches. The proposed metric is more consistent with human visual system and thus outperforms other existing and widely used metrics, namely the root mean square error (RMSE) and structural similarity index (SSIM). The computational cost of the proposed metric is also less compared to the state-of-the-art method.

Novel algebraic reconstruction technique for faster and finer CT reconstruction

n this paper, we present an innovative iterative algorithm for tomographic reconstruction. Algebraic reconstruction technique (ART) which is considered as the core of iterative approach has been enhanced to ensure much finer and faster reconstruction. Backprojection has been customized to make it work even when the projections are not uniformly distributed. Contour information of the object has been combined with customized backprojection to ensure a better initial guess to start ART iterations. Based on experiments with both simulated and real medical images it has been shown that the proposed modality is capable of computing more accurate reconstructions in addition with lower computational cost than traditional ART.

EEG source localization using a sparsity prior based on Brodmann areas

Localizing the sources of electrical activity in the brain from electroencephalographic (EEG) data is an important tool for noninvasive study of brain dynamics. Generally, the source localization process involves a high‐dimensional inverse problem that has an infinite number of solutions and thus requires additional constraints to be considered to have a unique solution. In this article, we propose a novel method for EEG source localization. The proposed method is based on dividing the cerebral cortex of the brain into a finite number of “functional zones” which correspond to unitary functional areas in the brain. To specify the sparsity profile of human brain activity more concisely, the proposed approach considers grouping of the electrical current dipoles inside each of the functional zones. In this article, we investigate the use of Brodmanns areas as the functional zones while sparse Bayesian learning is used to perform sparse approximation. Numerical experiments are conducted on a realistic head model obtained from segmentation of MRI images of the head and includes four major compartments namely scalp, skull, cerebrospinal fluid (CSF), and brain with relative conductivity values. Three different electrode setups are tested in the numerical experiments. The results demonstrate that the proposed approach is quite promising in solving the EEG source localization problem. In a noiseless environment with 71 electrodes, the proposed method was found to accurately locate up to 6 simultaneously active sources with accuracy >70%.

Evaluating the performance of BSBL methodology for EEG source localization on a realistic head model

In this paper, we evaluate the performance of block sparse Bayesian learning (BSBL) method for EEG source localization. By exploiting the internal block structure, the BSBL method solves the ill‐posed inverse problem more efficiently than other methods that do not consider block structure. Simulation experiments were conducted on a realistic head model obtained by segmentation of MRI images of the head. Two definitions of blocks were considered: Brodmann areas and automated anatomical labeling (AAL). The experiments were performed both with and without the presence of noise. Six different noise levels were considered having SNR values from 5 dB to 30 dB with 5dB increment. The evaluation reveals several potential findings—first, BSBL is more likely to produce better source localization than sparse Bayesian learning (SBL), however, this is true up until a limited number of simultaneously active areas only. Experimental results show that for 71‐channel electrodes setup BSBL outperforms SBL for up to three simultaneously active blocks. From four simultaneously active blocks SBL turns out to be marginally better and the difference between them is statistically insignificant. Second, different anatomical block structures such as Brodmann areas or AAL does not seem to produce any significant difference in EEG source localization relying on BSBL. Third, even when the block partitions are not known exactly BSBL ensures better localization than SBL as soon as block structure persists in the signal.

CT reconstruction from simultaneous projections: a step towards capturing CT in One Go

This paper focuses on minimising the time requirement for CT capture through an innovative simultaneous X-ray capture method. The state-of-the-art CT-imaging methodology captures a sequence of projections during which the internal organ movements may lead to poor reconstruction due to motion artefacts. Traditional CT scanners minimise such effects by taking extra projections. In this work, we focus on an innovative CT capture method that captures projections simultaneously, thus significantly minimises the scan time. Through ensuring rapid scan, the proposed method eliminates the extra projections that are needed to compensate for motion artefacts. By requiring a less number of projections, the simultaneous CT capture model ultimately minimises the radiation dose requirement. While the simultaneous CT capture model has already been proposed in our earlier works [Saha S, Tahtali M, Lambert A, Pickering M. 2012. Multi-axial CT reconstruction from few view projections. In: SPIE optical engineering+applicatio...