Saurabh Vyas

MRBrainS challenge: online evaluation framework for brain image segmentation in 3T MRI scans

Many methods have been proposed for tissue segmentation in brain MRI scans. The multitude of methods proposed complicates the choice of one method above others. We have therefore established the MRBrainS online evaluation framework for evaluating (semi)automatic algorithms that segment gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) on 3T brain MRI scans of elderly subjects (65–80 y). Participants apply their algorithms to the provided data, after which their results are evaluated and ranked. Full manual segmentations of GM, WM, and CSF are available for all scans and used as the reference standard. Five datasets are provided for training and fifteen for testing. The evaluated methods are ranked based on their overall performance to segment GM, WM, and CSF and evaluated using three evaluation metrics (Dice, H95, and AVD) and the results are published on the MRBrainS13 website. We present the results of eleven segmentation algorithms that participated in the MRBrainS13 challenge workshop at MICCAI, where the framework was launched, and three commonly used freeware packages: FreeSurfer, FSL, and SPM. The MRBrainS evaluation framework provides an objective and direct comparison of all evaluated algorithms and can aid in selecting the best performing method for the segmentation goal at hand.

Estimating physiological skin parameters from hyperspectral signatures

Abstract. We describe an approach for estimating human skin parameters, such as melanosome concentration, collagen concentration, oxygen saturation, and blood volume, using hyperspectral radiometric measurements (signatures) obtained from in vivo skin. We use a computational model based on Kubelka-Munk theory and the Fresnel equations. This model forward maps the skin parameters to a corresponding multiband reflectance spectra. Machine-learning-based regression is used to generate the inverse map, and hence estimate skin parameters from hyperspectral signatures. We test our methods using synthetic and in vivo skin signatures obtained in the visible through the short wave infrared domains from 24 patients of both genders and Caucasian, Asian, and African American ethnicities. Performance validation shows promising results: good agreement with the ground truth and well-established physiological precepts. These methods have potential use in the characterization of skin abnormalities and in minimally-invasive prescreening of malignant skin cancers.

Hyperspectral signature analysis of skin parameters

The temporal analysis of changes in biological skin parameters, including melanosome concentration, collagen concentration and blood oxygenation, may serve as a valuable tool in diagnosing the progression of malignant skin cancers and in understanding the pathophysiology of cancerous tumors. Quantitative knowledge of these parameters can also be useful in applications such as wound assessment, and point-of-care diagnostics, amongst others. We propose an approach to estimate in vivo skin parameters using a forward computational model based on Kubelka-Munk theory and the Fresnel Equations. We use this model to map the skin parameters to their corresponding hyperspectral signature. We then use machine learning based regression to develop an inverse map from hyperspectral signatures to skin parameters. In particular, we employ support vector machine based regression to estimate the in vivo skin parameters given their corresponding hyperspectral signature. We build on our work from SPIE 2012, and validate our methodology on an in vivo dataset. This dataset consists of 241 signatures collected from in vivo hyperspectral imaging of patients of both genders and Caucasian, Asian and African American ethnicities. In addition, we also extend our methodology past the visible region and through the short-wave infrared region of the electromagnetic spectrum. We find promising results when comparing the estimated skin parameters to the ground truth, demonstrating good agreement with well-established physiological precepts. This methodology can have potential use in non-invasive skin anomaly detection and for developing minimally invasive pre-screening tools.

Computational modeling of skin reflectance spectra for biological parameter estimation through machine learning

A computational skin re ectance model is used here to provide the re ectance, absorption, scattering, and transmittance based on the constitutive biological components that make up the layers of the skin. The changes in re ectance are mapped back to deviations in model parameters, which include melanosome level, collagen level and blood oxygenation. The computational model implemented in this work is based on the Kubelka- Munk multi-layer re ectance model and the Fresnel Equations that describe a generic N-layer model structure. This assumes the skin as a multi-layered material, with each layer consisting of specic absorption, scattering coecients, re ectance spectra and transmittance based on the model parameters. These model parameters include melanosome level, collagen level, blood oxygenation, blood level, dermal depth, and subcutaneous tissue re ectance. We use this model, coupled with support vector machine based regression (SVR), to predict the biological parameters that make up the layers of the skin. In the proposed approach, the physics-based forward mapping is used to generate a large set of training exemplars. The samples in this dataset are then used as training inputs for the SVR algorithm to learn the inverse mapping. This approach was tested on VIS-range hyperspectral data. Performance validation of the proposed approach was performed by measuring the prediction error on the skin constitutive parameters and exhibited very promising results.

Non-invasive estimation of skin thickness from hyperspectral imaging and validation using echography

BACKGROUND The skin is the largest organ and is subject to the greatest exposure to outside elements throughout one׳s lifetime. Current data by the American Academy of Dermatology suggests that approximately ten people die each hour worldwide due to skin related conditions. Cancers such as melanoma are growths that originate in the epidermis. Therefore, an accurate and non-invasive method to estimate skin constitutive elements can play an important clinical role in detecting the early onset of skin tumors. It can also serve as a valuable tool for research and development in cosmetics and pharmaceuticals in general. METHODS In our prior work, we developed a method that combined a physics-based model of human skin with machine learning and Hyperspectral imaging to non-invasively estimate physiological skin parameters, including melanosomes, collagen, oxygen saturation, and blood volume. In this work, we extend that model to also estimate skin thickness. Moreover, for the first time, we develop a protocol to test our methodology for skin thickness estimation using Ultrasound to acquire a gold standard dataset. RESULTS We tested our methodology for skin thickness estimation on a dataset of 48 Hyperspectral signatures obtained in vivo from six patients under IRB at Johns Hopkins Hospital. We found mean absolute errors on the order of the Ultrasound resolution (i.e., our gold standard). DISCUSSION This is the first study of its kind to validate skin thickness estimates using a gold standard. Our preliminary results constitute a proof-of-concept that hyperspectral-based methods can accurately and non-invasively estimate skin thickness in clinical settings.

Machine learning methods for in vivo skin parameter estimation

The WHO estimates three million new cases of skin cancer each year. Therefore, there exists a need for prescreening tools that can estimate the biological parameters of human skin, as they can help detect cancers before metastasis. In this paper, we present a novel inverse modeling technique based on Kubelka-Munk theory and machine learning to estimate biological skin parameters from in vivo hyperspec-tral imaging. We use the k-nearest neighbors (k-NN) algorithm in order to estimate skin parameters from their hy-perspectral signatures. We test our methods on 241 hyper-spectral signatures obtained from both genders and three ethnicities, and find encouraging results.

Computing Cardiac Strain from Variational Optical Flow in Four-Dimensional Echocardiography

Myocardial strain is important to assess cardiac function and diagnose cardiovascular disease. Despite the adoption of 4D (volume + time) echocardiography for diagnostic and therapeutic purposes, current clinical practice often relies exclusively on 2D measurements of strain or flow information resulting from Doppler echography. However, strain is a 3D measure of deformation in the radial, circumferential and longitudinal directions and therefore full 3D strain, and in particular out-of- sagittal plane strain components, include important information for diagnostic purposes since they provide additional information on the manner in which the heart lengthens and contracts during diastole and systole. In our prior work, we have developed robust variational optical flow methods to estimate dense myocardial motion. In this study, we extend this methodology to track ventricular outlines, which are subsequently used to compute displacement and deformation fields. This in turn is used to compute volumetric estimates of strain. We test our methods on a dataset of 4D ultrasound acquired in vivo from seven patients, and find good agreement with physiological precepts.

Endocardium segmentation in 3D Transesophageal Echocardiography

Segmentation of endocardial walls from 3D Transesophageal Echocardiography (3D TEE) is a critical step in helping characterize the pathophysiology of heart disease. The goal of this paper is (a) to present a new level set method based on the use of a novel inhibition function and (b) to asses its performance and that of other best of breed segmentation algorithms such as graph cuts and random walker as applied to 3D TEE segmentation. The method is tested on 14 sequences collected from 8 patients; it shows good agreement with the ground truth developed by physicians, and demonstrates similar (if not better) performance than current segmentation methods.

Endocardial surface delineation in 3-D transesophageal echocardiography.

We describe and compare several methods for recovering endocardial walls from 3-D transesophageal echocardiography (3-D TEE), which can help with diagnostics or providing input into biomechanical models. We employ a segmentation method based on 3-D level sets that maximizes enclosed volume while minimizing surface area and uses a growth inhibition function that includes 3-D gradient magnitude (to locate the endocardial walls) and a thin tissue detector (for the mitral valve leaflets). We also study delineation using a graph cut method that performs automated seeding by leveraging a fast radial symmetry transform to determine a central axis along which the 3-D volume is warped into a cylindrical coordinate space. Finally, a random walker approach is also used for automated delineation. The methods are used to estimate clinically relevant cardiovascular volumetric parameters such as stroke volume and left ventricular ejection fraction. Experiments are performed on clinical data collected from patients undergoing cardiothoracic surgery. Performance evaluation includes comparisons of the automated delineations against expert-defined ground truth using a number of error metrics, as well as errors between automatically computed and expert-derived physiologic parameters.