Navneeth Subramanian

Automatic view classification of echocardiograms using Histogram of Oriented Gradients

When imaging the heart, using a 2D ultrasound probe, different views can manifest depending on the location and angulations of the probe. Some of these views have been labeled as standard views, due to the presentation and ease of assessment of key cardiac structures in them. We present an approach for automatic recognition and classification of these standard views - namely the Parasternal Long Axis (PLAX) and the Short Axis (SAX) B-mode echocardiograms. The Histogram of Oriented Gradients (HOG) used as the discriminating feature encodes the spatial arrangement of edges/gradients in the images. The HOG feature is computed on the pre-scan converted image data in the ultrasound beam space. On a fairly large database of 703 images, with a Support Vector Machine classifier we obtained an accuracy of about 98%.

Automated interventricular septum thickness measurement from B-mode echocardiograms

In this work, we address the problem of automated measurement of the interventricular septum thickness, one of the key parameters in cardiology, from B-mode echocardiograms. The problem is challenging due to high levels of noise, multi modal intensity, weak contrast due to near field haze, and non rigid motion of the septum across frames. We introduce a complete system for automated measurement of septum thickness from B-mode echocardiograms incorporating three main components: a 1D curve evolution algorithm using region statistics for segmenting the septum, a motion clustering method to locate the mitral valve, and a robust method to calculate the septum width from these inputs in accordance with medical standards. Our method effectively handles the challenges of such measurements and runs in near real time. Results on 57 patient recordings showed excellent agreement of the automated measurements with expert manual measurements.

Motion correction for augmented fluoroscopy - application to liver embolization

Hepatic embolization is a procedure designed to cut off blood supply to liver tumors, either hepatocellular carcinomas (HCC) or metastases from other parts of the body. While it often serves as a palliative treatment, it can also be indicated as a precursor to liver resection and liver transplants. The procedure itself is conducted under fluoroscopic X-ray guidance. Contrast agent is administered to opacify the vasculature and to indicate the arterial branches that feed the treatment target. These supply routes are then blocked by embolic agents, cutting off the tumors blood supply. While methods exist to enhance fluoroscopic images and reduce the dependency on contrast agent, they are typically confounded by patient respiratory motion and are hence not effective for abdominal interventions. This paper presents an appearance based tracking algorithm that quickly and accurately compensates for the livers bulk motion due to respiration, thereby enabling the application of fluoroscopic augmentations (i.e. image overlays) for hepatic embolization procedures. To quantify the accuracy of our algorithm, we manually identified vascular and artificial landmarks in fluoroscopy sequences acquired from three patients during free breathing. The average postmotion compensation landmark misalignment was 1.9 mm, with the maximum landmark misalignment not exceeding 5.5 mm.

Quality metric for parasternal long axis b-mode echocardiograms

This paper presents a method for automatically estimating the quality of Parasternal Long AXis (PLAX) B-mode echocardiograms. The purpose of the algorithm is to provide live feedback to the user on the quality of the acquired image. The proposed approach uses Generalized Hough Transform to compare the structures derived from the incoming image to a representative atlas, thereby providing a quality metric (PQM). On 133 PLAX images from 35 patients, we show: 1) PQM has high correlation with manual ratings from an expert echocardiographer 2) PQM has high correlation with contrast-to-noise ratio, a traditional indicator of image quality 3) on images with high PQM, error in automatic septal wall thickness measurement is low, and vice versa.

Operator guidance in 2D echocardiography via 3D model to image registration

Ubiquitous use of 2D ultrasound (US) is limited by the difficulty in interpretation of images for an untrained operator. We present a solution for operator guidance through visual cues via registration of US to a 3D model. The method is demonstrated on 2D echocardiography data, where we are able to localize the scan plane in relation to the standard planes on the 3D model. Our algorithm operates by pre-processing both the US and CT images to the most basic information- muscle, blood pool - using classification. Subsequently, these labels are registered using the match cardinality metric for binary labeled images. We evaluated our method on four parasternal long-axis and three parasternal short-axis images from different patients. Results show that our system is able to correctly distinguish between the different US standard views and is able to localize the scan on the 3D model, correctly on five out of seven cases.

Volume rendering segmented data using 3D textures: a practical approach for intra-operative visualization

Volume rendering has high utility in visualization of segmented datasets. However, volume rendering of the segmented labels along with the original data causes undesirable intermixing/bleeding artifacts arising from interpolation at the sharp boundaries. This issue is further amplified in 3D textures based volume rendering due to the inaccessibility of the interpolation stage. We present an approach which helps minimize intermixing artifacts while maintaining the high performance of 3D texture based volume rendering - both of which are critical for intra-operative visualization. Our approach uses a 2D transfer function based classification scheme where label distinction is achieved through an encoding that generates unique gradient values for labels. This helps ensure that labelled voxels always map to distinct regions in the 2D transfer function, irrespective of interpolation. In contrast to previously reported algorithms, our algorithm does not require multiple passes for rendering and supports greater than 4 masks. It also allows for real-time modification of the colors/opacities of the segmented structures along with the original data. Additionally, these capabilities are available with minimal texture memory requirements amongst comparable algorithms. Results are presented on clinical and phantom data.

Automated posteriorwall thickness measurement from B-mode ultrasound

In this paper, we present a robust algorithm to segment the posterior wall region and estimate wall thickness from parasternal long axis(PLAX) view cardiac US B-mode images. Posterior wall thickness (PWd), Septal wall thickness (SWTd) and Left ventricular Internal diameter(LVId) are used to detect and measure the extent of Left Ventricular Hypertrophy (LVH). Manual measurements of PWd suffers from large inter and intra observer variability due to weak endocardial boundary intertangled with speckle and poor contrast, movement of the fibrous structures like the chordae,papillary muscles and posterior mitral leaflet. The proposed algorithm seeks to address some of these issues by automating the measurement algorithm. The algorithm initially detects epicardial boundary by pericardium detection and later segments the endocardial boundary by a 1D active contour evolution. We have designed the algorithm on a pilot data set of 42 images and validated on 88 patient data sets.The measurement values are in excellent agreement with expert measurements with error = 2.06mm ± 1.5mm.

Volumetric peeling: feature centric visualization using membership functions

Surgeons often desire a volume visualization where a region of interest (vessels, tumor) is isolated and viewed with complete clarity in relation to neighboring landmarks. A method to fade the landmarks/ surrounding structures in relation to the region of interest is desired as it provides context and an appreciation of the surgical approach. Current methods for generating these visualizations are either cumbersome (clip--planes) or require significant preprocessing (segmentation). Transfer functions, which can provide this visualization, are limited to datasets where a relation between voxel intensities and structures exists. We present a powerful, easy-to-use method for exploring volumetric data by controlling visibility according to membership volumes. These membership volumes facilitate a plethora of effects that cannot be achieved using conventional classification methods such as transfer functions. The membership volumes allow assignment of optical properties based on a combination of spatial and intensity criteria. Various membership functions, ranging from analytical to distance map based are shown. Through the use of Boolean operations in pixel shaders, we demonstrate several examples of interactive real-time visualization using our method.

GPNLPerf: Robust 4d Non-rigid Motion Correction for Myocardial Perfusion Analysis

Since the introduction of wide cone detector systems, CT myocardial perfusion has been an area of increased interest, for which non-rigid registration [NRR] is a key step to further analysis. We propose a novel motion management pipeline for perfusion data, GPNLPerf (Group-wise, non-local, NRR for perfusion analysis) centering on group-wise NRR using non-local spatio-temporal constraints. The proposed pipeline deals with the NRR challenges for 4D perfusion data and results in generating clinically relevant perfusion parameters. We demonstrate results on 9 dynamic perfusion exams comparing results quantitatively with ANTs NRR and also show qualitative results on perfusion maps.

Echocardiogram view classification with appearance and spatial distributions

When imaging the heart, using a 2D ultrasound probe, different views can manifest depending on the location and angulations of the probe. Some of these views have been labeled as standard views, due to the presentation and ease of assessment of key cardiac structures in them. We present an approach for automatic recognition and classification of these standard views, as a potential enabler for automated measurements or detection of noise - all without a human in the loop. We present an approach for view classification, Spatial Pyramid Histogram of Words which successfully models the appearance and shape distributions of object class. We demonstrate the effectiveness of this technique for the task of discrimination between the B-mode Parasternal Long Axis (PLAX) and the Short Axis (SAX) echocardiograms. For this task, our method shows a classification accuracy of 98.3% on an exhaustive database of 703 ultrasound images.