Vanathi Gopalakrishnan

YNIMG Journal 2018 Journal Article

A computational framework for the detection of subcortical brain dysmaturation in neonatal MRI using 3D Convolutional Neural Networks

• Rafael Ceschin
• Alexandria Zahner
• William Reynolds
• Jenna Gaesser
• Giulio Zuccoli
• Cecilia W. Lo
• Vanathi Gopalakrishnan
• Ashok Panigrahy

Deep neural networks are increasingly being used in both supervised learning for classification tasks and unsupervised learning to derive complex patterns from the input data. However, the successful implementation of deep neural networks using neuroimaging datasets requires adequate sample size for training and well-defined signal intensity based structural differentiation. There is a lack of effective automated diagnostic tools for the reliable detection of brain dysmaturation in the neonatal period, related to small sample size and complex undifferentiated brain structures, despite both translational research and clinical importance. Volumetric information alone is insufficient for diagnosis. In this study, we developed a computational framework for the automated classification of brain dysmaturation from neonatal MRI, by combining a specific deep neural network implementation with neonatal structural brain segmentation as a method for both clinical pattern recognition and data-driven inference into the underlying structural morphology. We implemented three-dimensional convolution neural networks (3D-CNNs) to specifically classify dysplastic cerebelli, a subset of surface-based subcortical brain dysmaturation, in term infants born with congenital heart disease. We obtained a 0. 985 ± 0. 0241-classification accuracy of subtle cerebellar dysplasia in CHD using 10-fold cross-validation. Furthermore, the hidden layer activations and class activation maps depicted regional vulnerability of the superior surface of the cerebellum, (composed of mostly the posterior lobe and the midline vermis), in regards to differentiating the dysplastic process from normal tissue. The posterior lobe and the midline vermis provide regional differentiation that is relevant to not only to the clinical diagnosis of cerebellar dysplasia, but also genetic mechanisms and neurodevelopmental outcome correlates. These findings not only contribute to the detection and classification of a subset of neonatal brain dysmaturation, but also provide insight to the pathogenesis of cerebellar dysplasia in CHD. In addition, this is one of the first examples of the application of deep learning to a neuroimaging dataset, in which the hidden layer activation revealed diagnostically and biologically relevant features about the clinical pathogenesis. The code developed for this project is open source, published under the BSD License, and designed to be generalizable to applications both within and beyond neonatal brain imaging.

IJCAI Conference 2007 Conference Paper

• Yan Liu
• Jaime Carbonell
• Vanathi Gopalakrishnan
• Peter Weigele

Protein fold recognition is a crucial step in inferring biological structure and function. This paper focuses on machine learning methods for predicting quaternary structural folds, which consist of multiple protein chains that form chemical bonds among side chains to reach a structurally stable domain. The complexity associated with modeling the quaternary fold poses major theoretical and computational challenges to current machine learning methods. We propose methods to address these challenges and show how (1) domain knowledge is encoded and utilized to characterize structural properties using segmentation conditional graphical models; and (2) model complexity is handled through efficient inference algorithms. Our model follows a discriminative approach so that any informative features, such as those representative of overlapping or long-range interactions, can be used conveniently. The model is applied to predict two important quaternary folds, the triple beta-spirals and double-barrel trimers. Cross-family validation shows that our method outperforms other state-of-the art algorithms.

TIME Conference 1998 Conference Paper

Representing and Learning Temporal Relationships among Experimental Variables

• Vanathi Gopalakrishnan
• Bruce G. Buchanan

The authors describe the necessity to capture temporal information in scientific experiment design for analysis by machine learning algorithms that can learn useful temporal patterns among experimental variables. They have identified three types of temporal information, namely duration, rate of change, and sequence of application of laboratory operators that are useful to learn from experimental data. Their motivation stems from study of experimental design in the domain of macromolecular crystallography. They identify the challenges posed both by the domain as well as the temporal information on machine learning programs, and describe work in progress. They outline the method of temporal specialization for inducing temporal relations between experimental variables, and illustrate with an example from the domain.

AAAI Conference 1996 Short Paper

Inducing Design Biases that Characterize Successful Experimentation in Weak-Theory Domains: TIPS

• Vanathi Gopalakrishnan

Experiment design in domains with weak theories is largely a trial-and-error process. In such domains, the effects of actions are unpredictable due to insufficient knowledge about the causal relationships among entities involved in an experiment. Thus, experiments are designed based on heuristics obtained from prior experience. Assuming that past experiment designs leading to success or failure can be recorded electronically, this thesis research proposes one method for analyzing these designs to yield hints regarding effective operator application sequences. This work assumes that the order in which operators are applied matters to the overall success of experiments. Experiment design can also be thought of as a form of planning, since it involves generation of a sequence of steps comprising of one or more operations that can change the environment by changing values of some of the parameters that describe the environment. Experiment design operators can therefore be thought of as plan operators at higher levels of abstraction. This thesis proposes a method for learning contexts within which applying certain sequences of operators has favored successful experimentation in the past.

AAAI Conference 1994 Short Paper

The Crystallographer’s Assistant

• Vanathi Gopalakrishnan
• Bruce Buchanan

The only routinely used technique available today for obtaining the 3-D structure of a protein or DNA molecule is by X-ray diffracting a crystal of the macromolecule. The rate limiting step in structure determination is the process of growing a crystal of the macromolecule. This process is not very well understood, and can take a few weeks to several years. Crystallographers, therefore, are in great need of tools to aid them in the process of designing and performing experiments. There is a great deal of experiential data in this domain, in the form of scientific notebooks with graphical and textual representations of previous experiments.