Amarjot Singh

Automatic Segmentation of Trophectoderm in Microscopic Images of Human Blastocysts

Accurate assessment of embryos viability is an extremely important task in the optimization of in vitro fertilization treatment outcome. One of the common ways of assessing the quality of a human embryo is grading it on its fifth day of development based on morphological quality of its three main components (Trophectoderm, Inner Cell Mass, and the level of expansion or the thickness of its Zona Pellucida). In this study, we propose a fully automatic method for segmentation and measurement of TE region of blastocysts (day-5 human embryos). Here, we eliminate the inhomogeneities of the blastocysts surface using the Retinex theory and further apply a level-set algorithm to segment the TE regions. We have tested our method on a dataset of 85 images and have been able to achieve a segmentation accuracy of 84.6% for grade A, 89.0% for grade B, and 91.7% for grade C embryos.

Aerial Scene Understanding Using Deep Wavelet Scattering Network and Conditional Random Field

This paper presents a fast and robust architecture for scene understanding for aerial images recorded from an Unmanned Aerial Vehicle. The architecture uses Deep Wavelet Scattering Network to extract Translation and Rotation Invariant features that are then used by a Conditional Random Field to perform scene segmentation. Experiments are conducted using the proposed framework on two annotated datasets of 1277 images and 300 aerial images, introduced in the paper. An overall pixel accuracy of 81 % and 78 % is achieved for the datasets. A comparison with another similar framework is also presented.

Multi-Resolution Dual-Tree Wavelet Scattering Network for Signal Classification

This paper introduces a Deep Scattering network that utilizes Dual-Tree complex wavelets to extract multi-scale translation invariant representations from an input signal. The computationally efficient Dual-Tree wavelets decompose the input signal into equally spaced representations over scales. Translation invariance is introduced in the representations by applying a non-linearity over a region followed by averaging. The discriminatory information from the equally spaced locally smooth signal representations aids the learning of the classi- fier. The proposed network is shown to outperform Mallat’s ScatterNet [1] on four datasets with different modalities, both for classification accuracy and computational efficiency.

Automatic blastomere detection in day 1 to day 2 human embryo images using partitioned graphs and ellipsoids

Fertility specialists have linked the size, shape and position of blastomeres in humans embryos with the viability of such embryos. We propose an automatic blastomere identification and modeling approach in an attempt to aid physicians in determining embryos viability. The proposed method applies isoperimetric graph partitioning, succeeded by a novel region merging algorithm to Hoffman Modulation Contrast (HMC) embryo images, to approximate blastomeres positions. Ellipsoidal models are then used to approximate the shape and the size of each blastomere. We discuss experimental results on a dataset of 40 embryo images, and expand on the advantages and drawbacks of our method while comparing our method to other approaches.

Dual-Tree wavelet scattering network with parametric log transformation for object classification

We introduce a ScatterNet that uses a parametric log transformation with Dual-Tree complex wavelets to extract translation invariant representations from a multi-resolution image. The parametric transformation aids the OLS pruning algorithm by converting the skewed distributions into relatively mean-symmetric distributions while the Dual-Tree wavelets improve the computational efficiency of the network. The proposed network is shown to outperform Mallats ScatterNet [1] on two image datasets, both for classification accuracy and computational efficiency. The advantages of the proposed network over other supervised and some unsupervised methods are also presented using experiments performed on different training dataset sizes.

Annotation Supported Contour Based Object Tracking With Frame Based Error Analysis

Motion tracking is a vital component of study for a video sequence having wide applications in object tracking, coding and editing the videos and mosaicking (2). Getting the actual motion for the videos existing in the real world though is a difficult task but plays a pivotal role in designing a model for any algorithms evaluation. We used an interactive computer vision system (1) which provides annotation tool for labeling and tracking the contours. Through the mutual work of user interaction and the computer vision system, the input effort is greatly reduced, simultaneously increasing the dependability of the whole system as compared to the solely computer based system. The ability of humans to easily segment and detect difference between different frames has been utilized using the human in loop methodology (1) by making use of a simple camera. The paper experiments with the capabilities of the system applied to indoor video sequence. This is the first paper which evaluates the capabilities of image annotation supported contour based object tracking with error analysis and correction, explaining the significance of human in the error incurred by the methodology. The paper studies the error incurred by the system with movement from one frame to another, supported by detailed simulations. The paper also focuses on the reasons responsible for the error incurred by the system mainly involving human intervention. Finally the paper presents the correction of the error followed by the in depth simulation indicating the in capabilities of the system on deforming objects. This system can be effectively used to analyze the error in motion tracking and further correcting the error leading to flawless tracking.

Level Set Segmentation of Brain Matter Using a Trans-Roto-Scale Invariant High Dimensional Feature

Brain matter extraction from MR images is an essential, but tedious process performed manually by skillful medical professionals. Automation can be a potential solution to this complicated task. However, it is an ambitious task due to the irregular boundaries between the grey and white matter regions. The intensity inhomogeneity in the MR images further adds to the complexity of the problem. In this paper, we propose a high dimensional translation, rotation, and scale-invariant feature, further used by a variational framework to perform the desired segmentation. The proposed model is able to accurately segment out the brain matter. The above argument is supported by extensive experimentation and comparison with the state-of-the-art methods performed on several MRI scans taken from the McGill Brain Web.