Vahid Taimouri

Web-based interactive 2D/3D medical image processing and visualization software

There are many medical image processing software tools available for research and diagnosis purposes. However, most of these tools are available only as local applications. This limits the accessibility of the software to a specific machine, and thus the data and processing power of that application are not available to other workstations. Further, there are operating system and processing power limitations which prevent such applications from running on every type of workstation. By developing web-based tools, it is possible for users to access the medical image processing functionalities wherever the internet is available. In this paper, we introduce a pure web-based, interactive, extendable, 2D and 3D medical image processing and visualization application that requires no client installation. Our software uses a four-layered design consisting of an algorithm layer, web-user-interface layer, server communication layer, and wrapper layer. To compete with extendibility of the current local medical image processing software, each layer is highly independent of other layers. A wide range of medical image preprocessing, registration, and segmentation methods are implemented using open source libraries. Desktop-like user interaction is provided by using AJAX technology in the web-user-interface. For the visualization functionality of the software, the VRML standard is used to provide 3D features over the web. Integration of these technologies has allowed implementation of our purely web-based software with high functionality without requiring powerful computational resources in the client side. The user-interface is designed such that the users can select appropriate parameters for practical research and clinical studies.

Intra-patient supine-prone colon registration in CT colonography using shape spectrum

CT colonography (CTC) is a minimally invasive screening technique for colorectal polyps and colon cancer. Since electronic colon cleansing (ECC) cannot completely remove the presence of pseudo-polyps, most CTC protocols acquire both prone and supine images to improve the visualization of the lumen wall and to reduce false positives. Comparisons between the prone and supine images can be facilitated by computerized registration between the scans. In this paper, we develop a fully automatic method for registering colon surfaces extracted from prone and supine images. The algorithm uses shape spectrum to extract the shape characteristics which are employed as the surface signature to find the correspondent regions between the prone and supine lumen surfaces. Our experimental results demonstrate an accuracy of 12.6 +/- 4.20 mm over 20 datasets. It also shows excellent potential in reducing the false positive when it is used to determine polyps through correspondences between prone and supine images.

Passive fMRI mapping of language function for pediatric epilepsy surgical planning: Validation using Wada, ECS, and FMAER

In this study we validate passive language fMRI protocols designed for clinical application in pediatric epilepsy surgical planning as they do not require overt participation from patients. We introduced a set of quality checks that assess reliability of noninvasive fMRI mappings utilized for clinical purposes. We initially compared two fMRI language mapping paradigms, one active in nature (requiring participation from the patient) and the other passive in nature (requiring no participation from the patient). Group-level analysis in a healthy control cohort demonstrated similar activation of the putative language centers of the brain in the inferior frontal (IFG) and temporoparietal (TPG) regions. Additionally, we showed that passive language fMRI produced more left-lateralized activation in TPG (LI=+0.45) compared to the active task; with similarly robust left-lateralized IFG (LI=+0.24) activations using the passive task. We validated our recommended fMRI mapping protocols in a cohort of 15 pediatric epilepsy patients by direct comparison against the invasive clinical gold-standards. We found that language-specific TPG activation by fMRI agreed to within 9.2mm to subdural localizations by invasive functional mapping in the same patients, and language dominance by fMRI agreed with Wada test results at 80% congruency in TPG and 73% congruency in IFG. Lastly, we tested the recommended passive language fMRI protocols in a cohort of very young patients and confirmed reliable language-specific activation patterns in that challenging cohort. We concluded that language activation maps can be reliably achieved using the passive language fMRI protocols we proposed even in very young (average 7.5 years old) or sedated pediatric epilepsy patients.

Colon Segmentation for Prepless Virtual Colonoscopy

A novel segmentation framework for a prepless virtual colonoscopy (VC) is presented, which reduces the necessity for colon cleansing before the CT scan. The patient is injected rectally with a water-soluble iodinated contrast medium using manual insufflators and a small rectal catheter. Compared to the air-based contrast medium, this technique can better preserve the color lumen and reduce the partial volume effect. However, the contrast medium, together with the fecal materials and air, makes colon wall segmentation challenging. Our solution makes no assumptions about the shape, size, and location of the fecal material in the colon. This generality allows us to label the fecal material accurately and extract the colon wall reliably. The accuracy of our technique has been verified on 60 human subjects. Compared with current VC technologies, our method is shown to be better in terms of both sensitivity and specificity. Further, in our experiments, the accuracy of the technique was comparable to that of optical colonoscopy results.

Segmentation of cell nuclei in heterogeneous microscopy images: A reshapable templates approach

Histological tissue images typically exhibit very sophisticated spatial color patterns. It is of great clinical importance to extract qualitative and quantitative information from these images. As an ad hoc solution, various unsupervised approaches address the object detection and segmentation problem which are suitable for limited classes of histology images. In this paper, we propose a general purpose localization and segmentation method which utilizes reshapable templates. The method combines both pixel- and object-level features for detecting regions of interest. Segmentation is carried out in two levels including both the coarse and fine ones. A set of simple-shaped templates is used for coarse segmentation. A content based template reshaping algorithm is proposed for fine segmentation of target objects. Experimentation was done using a publicly available image data set which contains 7931 manually labeled cells of heterogeneous histology images. The experiments have demonstrated acceptable level of detection and segmentation results for the proposed approach (precision=0.904, recall=0.870 and Zijdenbos similarity index=73%). Thus, the prototype software developed based on proposed method can be considered as a potential tool for pathologists in clinical process.

Deformation similarity measurement in quasi-conformal shape space

This paper presents a novel approach based on the shape space concept to classify deformations of 3D models. A new quasi-conformal metric is introduced which measures the curvature changes at each vertex of each pose during the deformation. The shapes with similar deformation patterns follow a similar deformation curve in shape space. Energy functional of the deformation curve is minimized to calculate the geodesic curve connecting two shapes on the shape space manifold. The geodesic distance illustrates the similarity between two shapes, which is used to compute the similarity between the deformations. We applied our method to classify the left ventricle deformations of myopathic and control subjects, and the sensitivity and specificity of our method were 88.8% and 85.7%, which are higher than other methods based on the left ventricle cavity, which shows our method can quantify the similarity and disparity of the left ventricle motion well.

A template-to-slice block matching approach for automatic localization of brain in fetal MRI

To enhance post-acquisition processing of fetal brain MRI we have developed a template-to-slice block matching technique that matches a spatiotemporal (4D) atlas of the fetal brain to the corresponding section of the brain in each 2D fetal MRI scan. As compared to the recent studies which used feature based approaches for fetal brain localization, we propose a template matching approach that registers the template to each slice of the brain in 3D. In addition to brain localization, this novel technique registers the template to each fetal MRI scan thus is useful for slice-level motion correction, brain segmentation, and reconstruction. In this paper we describe the details of our algorithm and report statistical analysis of the accuracy of localization and registration. We examined the algorithm on 366 scans obtained from fetal MRI of 30 subjects and used a 4D MRI atlas of the fetal brain as template. Our results show 94% success rate in brain localization, and 73% successful matching, with a median localization error of 6.45 mm for all scans, and median target registration error of 4.2 mm for successfully localized scans. This algorithm can be used for automatic localization of fetal brain as the first step of fetal MRI analysis pipeline.

Automated template-based brain localization and extraction for fetal brain MRI reconstruction

ABSTRACT Most fetal brain MRI reconstruction algorithms rely only on brain tissue‐relevant voxels of low‐resolution (LR) images to enhance the quality of inter‐slice motion correction and image reconstruction. Consequently the fetal brain needs to be localized and extracted as a first step, which is usually a laborious and time consuming manual or semi‐automatic task. We have proposed in this work to use age‐matched template images as prior knowledge to automatize brain localization and extraction. This has been achieved through a novel automatic brain localization and extraction method based on robust template‐to‐slice block matching and deformable slice‐to‐template registration. Our template‐based approach has also enabled the reconstruction of fetal brain images in standard radiological anatomical planes in a common coordinate space. We have integrated this approach into our new reconstruction pipeline that involves intensity normalization, inter‐slice motion correction, and super‐resolution (SR) reconstruction. To this end we have adopted a novel approach based on projection of every slice of the LR brain masks into the template space using a fusion strategy. This has enabled the refinement of brain masks in the LR images at each motion correction iteration. The overall brain localization and extraction algorithm has shown to produce brain masks that are very close to manually drawn brain masks, showing an average Dice overlap measure of 94.5%. We have also demonstrated that adopting a slice‐to‐template registration and propagation of the brain mask slice‐by‐slice leads to a significant improvement in brain extraction performance compared to global rigid brain extraction and consequently in the quality of the final reconstructed images. Ratings performed by two expert observers show that the proposed pipeline can achieve similar reconstruction quality to reference reconstruction based on manual slice‐by‐slice brain extraction. The proposed brain mask refinement and reconstruction method has shown to provide promising results in automatic fetal brain MRI segmentation and volumetry in 26 fetuses with gestational age range of 23 to 38 weeks. HIGHLIGHTSWe offer a template‐based fetal brain localization, extraction and segmentation.We reconstruct fetal brain MRI in a standard common coordinate space.We achieve brain extraction in addition to localization success rate of 93%.Brain segmentation accuracy (Dice overlap) compared to manual delineation is 94.5%.We report fetal brain tissue volume growth maps using atlas‐based segmentation.

Visualization of Shape Motions in Shape Space

Analysis of dynamic object deformations such as cardiac motion is of great importance, especially when there is a necessity to visualize and compare the deformation behavior across subjects. However, there is a lack of effective techniques for comparative visualization and assessment of a collection of motion data due to its 4-dimensional nature, i.e., timely varying three-dimensional shapes. From the geometric point of view, the motion change can be considered as a function defined on the 2D manifold of the surface. This paper presents a novel classification and visualization method based on a medial surface shape space, in which two novel shape descriptors are defined, for discriminating normal and abnormal human heart deformations as well as localizing the abnormal motion regions. In our medial surface shape space, the geodesic distance connecting two points in the space measures the similarity between their corresponding medial surfaces, which can quantify the similarity and disparity of the 3D heart motions. Furthermore, the novel descriptors can effectively localize the inconsistently deforming myopathic regions on the left ventricle. An easy visualization of heart motion sequences on the projected space allows users to distinguish the deformation differences. Our experimental results on both synthetic and real imaging data show that this method can automatically classify the healthy and myopathic subjects and accurately detect myopathic regions on the left ventricle, which outperforms other conventional cardiac diagnostic methods.

Spatially Constrained Incoherent Motion SCIM Model Improves Quantitative Diffusion-Weighted MRI Analysis of Crohn's Disease Patients

Quantitative analysis of fast and slow diffusion from abdominal Diffusion-weighted MRI has the potential to provide important new insights into physiological and microstructural properties of the body. However, the commonly used, independent voxel-wise fitting of the signal decay model leads to imprecise parameter estimates, which has hampered their practical usage. In this work we evaluated the improvement in the precision of the fast and slow diffusion parameter estimates achieved by using a spatially-constrained Incoherent Motion SCIM model of DW-MRI signal decay in 5 healthy subjects and 24 Crohns disease patients. We found that the improvement in Coefficient of Variation CV of the parameter estimates achieved using the SCIM model was significantly larger compared to thus achieved by repeated acquisition and signal averaging n=5, paired Students t-test, p ≤ 0.05. We also found that the SCIM model reduced the coefficient of variation of the parameter estimates of the D * and f parameter estimates in the ileum by 30% compared to the independent voxel-wise fitting of the signal decay model in the Crohns patients data n=24, paired Students t-test, p ≤ 0.05. The SCIM model is more precise for quantitative analysis of abdominal DW-MRI signal decay.