Qimei Liao

Prior image based anisotropic edge guided TV minimization for few-view CT reconstruction

To improve the spatial resolution of the image reconstructed by the conventional total variation (TV) algorithm, we propose a prior image based anisotropic edge guided TV minimization (PIEGTV) algorithm for few-view CT reconstruction. In this study, an anisotropic edge of the prior image is detected using the proposed edge detector. Then the weights of the TV discretization term for the to-be-estimated image are updated by the anisotropic edge information. To solve the minimization problem of the PIEGTV reconstruction, a similar TV-based minimization implementation is developed to deal with the raw data fidelity and other constraints. The results with computer simulations for the Shepp-Logan phantom and experimental data from a physical phantom demonstrate that the proposed algorithm can yield images with noticeable gains in edge preserving and shape preserving for small structures, compared to the conventional and a few modified TV algorithms.

Decreased middle temporal gyrus connectivity in the language network in schizophrenia patients with auditory verbal hallucinations.

As the most common symptoms of schizophrenia, the long-term persistence of obstinate auditory verbal hallucinations (AVHs) brings about great mental pain to patients. Neuroimaging studies of schizophrenia have indicated that AVHs were associated with altered functional and structural connectivity within the language network. However, effective connectivity that could reflect directed information flow within this network and is of great importance to understand the neural mechanisms of the disorder remains largely unknown. In this study, we utilized stochastic dynamic causal modeling (DCM) to investigate directed connections within the language network in schizophrenia patients with and without AVHs. Thirty-six patients with schizophrenia (18 with AVHs and 18 without AVHs), and 37 healthy controls participated in the current resting-state functional magnetic resonance imaging (fMRI) study. The results showed that the connection from the left inferior frontal gyrus (LIFG) to left middle temporal gyrus (LMTG) was significantly decreased in patients with AVHs compared to those without AVHs. Meanwhile, the effective connection from the left inferior parietal lobule (LIPL) to LMTG was significantly decreased compared to the healthy controls. Our findings suggest aberrant pattern of causal interactions within the language network in patients with AVHs, indicating that the hypoconnectivity or disrupted connection from frontal to temporal speech areas might be critical for the pathological basis of AVHs.

Medical image retrieval system using multiple features from 3D ROIs

Compared to a retrieval using global image features, features extracted from regions of interest (ROIs) that reflect distribution patterns of abnormalities would benefit more for content-based medical image retrieval (CBMIR) systems. Currently, most CBMIR systems have been designed for 2D ROIs, which cannot reflect 3D anatomical features and region distribution of lesions comprehensively. To further improve the accuracy of image retrieval, we proposed a retrieval method with 3D features including both geometric features such as Shape Index (SI) and Curvedness (CV) and texture features derived from 3D Gray Level Co-occurrence Matrix, which were extracted from 3D ROIs, based on our previous 2D medical images retrieval system. The system was evaluated with 20 volume CT datasets for colon polyp detection. Preliminary experiments indicated that the integration of morphological features with texture features could improve retrieval performance greatly. The retrieval result using features extracted from 3D ROIs accorded better with the diagnosis from optical colonoscopy than that based on features from 2D ROIs. With the test database of images, the average accuracy rate for 3D retrieval method was 76.6%, indicating its potential value in clinical application.

Algorithm for x-ray beam hardening and scatter correction in low-dose cone-beam CT: phantom studies

X-ray scatter poses a significant limitation to image quality in cone-beam CT (CBCT), as well as beam hardening, resulting in image artifacts, contrast reduction, and lack of CT number accuracy. Meanwhile the x-ray radiation dose is also non-ignorable. Considerable scatter or beam hardening correction methods have been developed, independently, and rarely combined with low-dose CT reconstruction. In this paper, we combine scatter suppression with beam hardening correction for sparse-view CT reconstruction to improve CT image quality and reduce CT radiation. Firstly, scatter was measured, estimated, and removed using measurement-based methods, assuming that signal in the lead blocker shadow is only attributable to x-ray scatter. Secondly, beam hardening was modeled by estimating an equivalent attenuation coefficient at the effective energy, which was integrated into the forward projector of the algebraic reconstruction technique (ART). Finally, the compressed sensing (CS) iterative reconstruction is carried out for sparse-view CT reconstruction to reduce the CT radiation. Preliminary Monte Carlo simulated experiments indicate that with only about 25% of conventional dose, our method reduces the magnitude of cupping artifact by a factor of 6.1, increases the contrast by a factor of 1.4 and the CNR by a factor of 15. The proposed method could provide good reconstructed image from a few view projections, with effective suppression of artifacts caused by scatter and beam hardening, as well as reducing the radiation dose. With this proposed framework and modeling, it may provide a new way for low-dose CT imaging.

Scattering-compensated cone beam x-ray luminescence computed tomography

X-ray luminescence computed tomography (XLCT) opens new possibilities to perform molecular imaging with x-ray. It is a dual modality imaging technique based on the principle that some nanophosphors can emit near-infrared (NIR) light when excited by x-rays. The x-ray scattering effect is a great issue in both CT and XLCT reconstruction. It has been shown that if the scattering effect compensated, the reconstruction average relative error can be reduced from 40% to 12% in the in the pencil beam XLCT. However, the scattering effect in the cone beam XLCT has not been proved. To verify and reduce the scattering effect, we proposed scattering-compensated cone beam x-ray luminescence computed tomography using an added leading to prevent the spare x-ray outside the irradiated phantom in order to decrease the scattering effect. Phantom experiments of two tubes filled with Y2O3:Eu3+ indicated that the proposed method could reduce the scattering by a degree of 30% and can reduce the location error from 1.8mm to 1.2mm. Hence, the proposed method was feasible to the general case and actual experiments and it is easy to implement.

Modulation transfer function determination using the edge technique for cone-beam micro-CT

Evaluating spatial resolution is an essential work for cone-beam computed tomography (CBCT) manufacturers, prototype designers or equipment users. To investigate the cross-sectional spatial resolution for different transaxial slices with CBCT, the slanted edge technique with a 3D slanted edge phantom are proposed and implemented on a prototype cone-beam micro-CT. Three transaxial slices with different cone angles are under investigation. An over-sampled edge response function (ERF) is firstly generated from the intensity of the slightly tiled air to plastic edge in each row of the transaxial reconstruction image. Then the oversampled ESF is binned and smoothed. The derivative of the binned and smoothed ERF gives the line spread function (LSF). At last the presampled modulation transfer function (MTF) is calculated by taking the modulus of the Fourier transform of the LSF. The spatial resolution is quantified with the spatial frequencies at 10% MTF level and full-width-half-maximum (FWHM) value. The spatial frequencies at 10% of MTFs are 3.1±0.08mm-1, 3.0±0.05mm-1, and 3.2±0.04mm-1 for the three transaxial slices at cone angles of 3.8°, 0°, and -3.8° respectively. The corresponding FWHMs are 252.8μm, 261.7μm and 253.6μm. Results indicate that cross-sectional spatial resolution has no much differences when transaxial slices being 3.8° away from z=0 plane for the prototype conebeam micro-CT.

Beam hardening correction for sparse-view CT reconstruction

Beam hardening, which is caused by spectrum polychromatism of the X-ray beam, may result in various artifacts in the reconstructed image and degrade image quality. The artifacts would be further aggravated for the sparse-view reconstruction due to insufficient sampling data. Considering the advantages of the total-variation (TV) minimization in CT reconstruction with sparse-view data, in this paper, we propose a beam hardening correction method for sparse-view CT reconstruction based on Brabant’s modeling. In this correction model for beam hardening, the attenuation coefficient of each voxel at the effective energy is modeled and estimated linearly, and can be applied in an iterative framework, such as simultaneous algebraic reconstruction technique (SART). By integrating the correction model into the forward projector of the algebraic reconstruction technique (ART), the TV minimization can recover images when only a limited number of projections are available. The proposed method does not need prior information about the beam spectrum. Preliminary validation using Monte Carlo simulations indicates that the proposed method can provide better reconstructed images from sparse-view projection data, with effective suppression of artifacts caused by beam hardening. With appropriate modeling of other degrading effects such as photon scattering, the proposed framework may provide a new way for low-dose CT imaging.

A ROC-based feature selection method for computer-aided detection and diagnosis

Image-based computer-aided detection and diagnosis (CAD) has been a very active research topic aiming to assist physicians to detect lesions and distinguish them from benign to malignant. However, the datasets fed into a classifier usually suffer from small number of samples, as well as significantly less samples available in one class (have a disease) than the other, resulting in the classifier’s suboptimal performance. How to identifying the most characterizing features of the observed data for lesion detection is critical to improve the sensitivity and minimize false positives of a CAD system. In this study, we propose a novel feature selection method mR-FAST that combines the minimal-redundancymaximal relevance (mRMR) framework with a selection metric FAST (feature assessment by sliding thresholds) based on the area under a ROC curve (AUC) generated on optimal simple linear discriminants. With three feature datasets extracted from CAD systems for colon polyps and bladder cancer, we show that the space of candidate features selected by mR-FAST is more characterizing for lesion detection with higher AUC, enabling to find a compact subset of superior features at low cost.

MRI-Based Thickness Analysis of Bladder Cancer: A Pilot Study

To find an effective way to quantitatively analyze the thickness variation of human bladder wall under different states, in this paper, we proposed a novel pipeline for thickness measurement, analysis, and mapping of bladder wall based on T2-weighted MRI images. The pipeline includes major steps of data acquisition, automatic segmentation of bladder wall, 3D thickness calculation, thickness normalization, and standardized bladder shape mapping. Based on the proposed pipeline, 20 datasets including 10 patients and 10 volunteers were used to explore the distribution pattern of wall thickness and find the difference between cancerous tissue and normal bladder wall. The results demonstrated the potential of wall thickness as a good indicator of bladder abnormalities, indicating its possible use in lesion detection on the bladder wall.

A New Method to Detect Soft-Tissue Deformation Based on HAMMER Algorithm

The integration of preoperative high-resolution and high tissue contrast MRI 3D data with intraoperative Ultrasound (US) images would be a possible way to reflect organ deformation during abdominal surgery. In this study, a new approach is presented to detect and evaluate the deformation of soft tissue through the registration and fusion of MRI data with US scans based on HAMMER algorithm. To evaluate the proposed method quantitatively, an ultrasound simulation module was developed and then ultrasound images were simulated based on deformed MRI images that were generated from acquired abdominal MRI data using rigid and non-rigid transformations. The deformed ultrasound images and original MRI images were registered by modified HAMMER algorithm, and the deformation correction was evaluated by the average displacement and mutual information. Preliminary experimental results verifies the feasibility of proposed method on the detection and correction of soft tissue deformation.