Maxim Fradkin

Efficient model-based quantification of left ventricular function in 3-D echocardiography

Quantitative functional analysis of the left ventricle plays a very important role in the diagnosis of heart diseases. While in standard two-dimensional echocardiography this quantification is limited to rather crude volume estimation, three-dimensional (3-D) echocardiography not only significantly improves its accuracy but also makes it possible to derive valuable additional information, like various wall-motion measurements. In this paper, we present a new efficient method for the functional evaluation of the left ventricle from 3-D echographic sequences. It comprises a segmentation step that is based on the integration of 3-D deformable surfaces and a four-dimensional statistical heart motion model. The segmentation results in an accurate 3-D + time left ventricle discrete representation. Functional descriptors like local wall-motion indexes are automatically derived from this representation. The method has been successfully tested both on electrocardiography-gated and real-time 3-D data. It has proven to be fast, accurate, and robust.

Segmentation of thrombus in abdominal aortic aneurysms from CTA with nonparametric statistical grey level appearance modeling

This paper presents a new method for deformable model-based segmentation of lumen and thrombus in abdominal aortic aneurysms from computed tomography (CT) angiography (CTA) scans. First the lumen is segmented based on two positions indicated by the user, and subsequently the resulting surface is used to initialize the automated thrombus segmentation method. For the lumen, the image-derived deformation term is based on a simple grey level model (two thresholds). For the more complex problem of thrombus segmentation, a grey level modeling approach with a nonparametric pattern classification technique is used, namely k-nearest neighbors. The intensity profile sampled along the surface normal is used as classification feature. Manual segmentations are used for training the classifier: samples are collected inside, outside, and at the given boundary positions. The deformation is steered by the most likely class corresponding to the intensity profile at each vertex on the surface. A parameter optimization study is conducted, followed by experiments to assess the overall segmentation quality and the robustness of results against variation in user input. Results obtained in a study of 17 patients show that the agreement with respect to manual segmentations is comparable to previous values reported in the literature, with considerable less user interaction.

Automatic myocardium segmentation in late-enhancement MRI

We propose a novel automatic method to segment the myocardium on late-enhancement cardiac MR (LE CMR) images with a multi-step approach. First, in each slice of the LE CMR volume, a geometrical template is deformed so that its borders fit the myocardial contours. The second step consists in introducing a shape prior of the left ventricle. To do so, we use the cine MR sequence that is acquired along with the LE CMR volume. As the myocardial contours can be more easily automatically obtained on this data, they are used to build a 3D mesh representing the left ventricle geometry and the underlying myocardium thickness. This mesh is registered towards the contours obtained with the geometrical template, then locally adjusted to guarantee that scars are included inside the final segmentation. The quantitative evaluation on 27 volumes (272 slices) shows robust and accurate results.

Comprehensive Segmentation of Cine Cardiac MR Images

A typical Cardiac Magnetic Resonance (CMR) examination includes acquisition of a sequence of short-axis (SA) and long-axis (LA) images covering the cardiac cycle. Quantitative analysis of the heart function requires segmentation of the left ventricle (LV) SA images, while segmented LA views allow more accurate estimation of the basal slice and can be used for slice registration. Since manual segmentation of CMR images is very tedious and time-consuming, its automation is highly required. In this paper, we propose a fully automatic 2D method for segmenting LV consecutively in LA and SA images. The approach was validated on 35 patients giving mean segmentation error smaller than one pixel, both for LA and SA, and accurate LV volume measurements.

Segmentation of Pathologic Hearts in Long-Axis Late-Enhancement MRI

We propose a new method to segment long-axis cardiac MR images acquired with a late-enhancement protocol. Detecting the myocardium boundaries is difficult in these images because healthy myocardium appears dark while the intensity of enhanced areas ranges from gray to white, depending on the myocardial damage. In this context, geometrical template deformation, alternated with the update of a damaged tissue map, allows us to include abnormal myocardium parts in the final segmentation. The template and map are initialized using short-axis images and the deformation parameters are adapted according to the type of enhancement pattern. Good segmentation results are obtained on a database of real pathologic heart images presenting various types of abnormal myocardium tissues.

Diagnosing breast cancer using independent diffuse optical tomography and x-ray mammography scans

We have previously demonstrated the utilization of spatially co-registered diffuse optical tomography (DOT) and digital breast tomosynthesis (DBT) for joint breast cancer diagnosis. However, clinical implementation of such a multi-modality approach may require development of integrated DOT/DBT imaging scanners, which can be costly and time-consuming. Exploring effective image registration methods that combine the diagnostic information from a standalone DOT measurement and a separate mammogram can be a cost-effective solution, which may eventually enable adding functional optical assessment to all previously installed digital mammography systems. In this study, we investigate a contour-based image registration method to convert independent optical and x-ray scans into co-registered datasets that can benefit from a joint image analysis. The breast surface used in 3D optical DOT reconstruction is registered with the breast contour line extracted from an x-ray mammogram acquired separately. This allows us to map the 2D mammogram to the optical measurement space and build structural constraints for optical image reconstruction. A non-linear reconstruction utilizing structure-priors is then performed to produce hemoglobin maps with improved resolution. To validate this approach, we used a set of tumor patient measurements with simultaneous DOT/DBT and separate 2D mammographic scans. The images recovered from the registration procedure derived from DOT and 2D mammogram present similar image quality compared to those recovered from co-registered DOT/DBT measurements.

Automatic Analysis of the Left Ventricle in the Time Sequences of 3D Echo-Cardiographic Images

The 3D ultrasound imagery becomesmore and more attractive for cardiac studies due to its simplicity, its improved reproducibility, and better precision, as compared to standard 2D echographic exams. However, automatic tools are needed to fully and efficiently analyze the large amounts of data obtained. In this paper, we present an automatic tool, aimed at the quantitative analysis of heart motion, based on the segmentation of the endo-cardium of the left ventricle. Using our method, quantitative measurements on volumes such as stroke and ejection-fraction (EF) are readily available (without needing any geometrical assumption), as well as regional wall motion parameters.

Characterizing breast lesions through robust multimodal data fusion using independent diffuse optical and x-ray breast imaging.

To enable tissue function-based tumor diagnosis over the large number of existing digital mammography systems worldwide, we propose a cost-effective and robust approach to incorporate tomographic optical tissue characterization with separately acquired digital mammograms. Using a flexible contour-based registration algorithm, we were able to incorporate an independently measured two-dimensional x-ray mammogram as structural priors in a joint optical/x-ray image reconstruction, resulting in improved spatial details in the optical images and robust optical property estimation. We validated this approach with a retrospective clinical study of 67 patients, including 30 malignant and 37 benign cases, and demonstrated that the proposed approach can help to distinguish malignant from solid benign lesions and fibroglandular tissues, with a performance comparable to the approach using spatially coregistered optical/x-ray measurements.

Fully automatic segmentation of short and long axis cine cardiac MR

Introduction Quantitative analysis of cardiac function requires d lineation of the left ventricle (LV) in cine cardiac MR (CMR). Typically, this is done using sho rt-axis (SA) images, however, acquisition of several long-axis (LA) views has bec ome quite common. The latter can be used for the accurate and reproducible determination of the basal SA slice, known as one of the major inter-observer variability factors in SA LV m easurements [1,2]. Since manual LV delineation is very tedious and time-consuming, aut omatic segmentation methods, enabling to obtain reproducible LV measurements, are highly des irable. Though many publications proposed such methods, mainly for SA, only very few of them reported acceptable accuracy.

Configurable real-time motion estimation for medical imaging: application to X-ray and ultrasound

AbstractMotion estimation is a key building block of image processing pipelines in many different contexts, ranging from efficient coding of video sequences in the consumer electronics domain (TV, DVD, BD) to professional medical applications. Many block-matching approaches have been proposed in the literature for motion detection and compensation in general, including both lossless and lossy algorithms. However, in real-time medical imaging applications, characterized by high frame rates, the needs for low latency and jitter, accuracy and robustness against noise are quite difficult to achieve with standard block-matching methods. We introduce a new hybrid image processing approach to block-matching that takes advantage of both types of algorithms (lossless and lossy), adapts to the image content and noise, and provides high flexibility for the speed/accuracy tradeoff. The presented approach has been successfully tested on interventional X-ray fluoroscopy and cardiac ultrasound images sequences.