Yachna Sharma
PhD candidate
School of Electrical and Computer Engineering
Georgia Institute of Technology
Atlanta, GA 30332
E-mail: ysharma3@gatech.edu, yachna3@gmail.com

I am a Ph.D. student at Georgia Tech working with Prof. Irfan Essa at Computational Perception Laboratory (CPL). My dissertation research is focused on automated competence scoring of human activitites using video data.

My previous research experience at Georgia Tech involved Computer Aided Diagnosis (CAD) of cancer, medical image registration and 3D volume rendering.

Before coming to Georgia Tech, I did my Masters in electrical and Computer engineering from University of Tennessee, Knoxville. I worked in the wireless and communication research group at UT Knoxville with Dr. Mostofa Howlader. During my Masters, I also worked in collaboration with Dr. Paul D. Ewing and Dr. Stephen F. Smith at RF and microwave systems group, Oak Ridge National Laboratory. For my Masters thesis, I designed and implemented a simulation framework to study the performance of direct sequence, slow and fast frequency hopping, and hybrid spread spectrum systems under multiuser interference and Rayleigh fading.
Research interests: Video analysis, activity recognition, medical image and signal processing, feature selection, machine learning

Resume: PDF    TXT

Research Projects
Human activity recognition and competency scoring
The goal of this project is to measure how well an action, activity, or a task is accomplished. Our work is motivated by visual activity and behavior recognition, but we seek to extend this work with a framework to support automatic skill assessment by analyzing time series motion data, captured as humans perform various tasks. Traditionally, visual activity recognition aims to identify what is happening and when it happened in a scene. On the other hand, our goal is to extract skill relevant information, which requires a holistic analysis of time series motion data, within a particular domain. Automated competency scoring is a challenging task due to diferent skill defining criteria in various domains. It is also difficult to isolate skill-relevant information from video data. We address these issues by designing a general framework for competency scoring using time series analysis. Preliminaty results show the applicability of our approach in different domains.

Significance: Competency asssessment is a time and labor intensive task in several domains. Automated assessment using computer vision and machine learning techniques can speed up the process and also reduce the human bias involved.

Image registration and 3D rendering of high resolution tissue images
Registration of high-resolution tissue images is a critical step in the 3D analysis of protein expression. Because the distance between images (~4-5μm thickness of a tissue section) is nearly the size of the objects of interest (~10-20μm cancer cell nucleus), a given object is often not present in both of two adjacent images. Without consistent correspondence of objects between images, registration becomes a difficult task. This work assesses the feasibility of current registration techniques for such images. We used simulation to help select appropriate features and methods for image registration by estimating best-case-scenario errors for given data constraints in histological images. The results of this study suggest that much of the difficulty of stained tissue registration can be reduced to the problem of accurately identifying feature points, such as the center of nuclei.

Significance: Simulation framework can assist in selecting the tissue preparation (e.g. slice thickness) and staining (e.g. nuclei) techniques beforehand for different protein expressions and tissue types. This can speed up the data preparation process for high resolution 3D rendeing of histological tissues.

Computer aided diagnosis of head and neck cancer
Squamous cell carcinoma of Head and Neck (SCCHN) represents more than 90% of all head and neck cancers. An important area of cancer research is to identify biomarkers for cancer prognosis such as folate receptors (FR) in cancer. The automated system designed in this project quantitatively evaluates biomarker expression in addition to cancerous and textural features from biopsy images. The proposed methodology encompasses image based morphological techniques to provide new quantitative measures for typical cancer attributes (nuclear atypia, pleomorphism and necrosis) and allows usage of these attributes along with biomarker expression characteristics. Grading accuracies up to 94% were achieved for a heterogeneous image dataset and the results demonstrated that image based quantified biomarker features result in considerable improvement in computer assisted grading of SCCHN as compared to cancerous and textural features.

Significance:The system is intended to assist pathologists in a clinical setting by correlating the biomarker expression to cancer grade.

Biopsy image preprocessing and cancer region segmentation
Manual analysis of cancer biopsies can be time consuming and challenging due to variations in tissue morphology, inconsistencies in preparation of tissue specimen and errors in the image acquisition process. The tool is designed to automatically standardize the variations in different images due to changing illumination and experimental conditions. Segmentation of cancerous regions from non-cancerous areas is a mandatory step before extracting relevant information from cancer images such as the number and size of nuclei and subsequently using it for classification and quantitative analysis. The tool was tested for two completely different cancers: Head and Neck Cancer (HNC) and Renal Cell Carcinoma (RCC). The tool enables the user to successfully segment the cancerous areas for both types of cancers and the results matched with the manual validation by an expert pathologist.

Significance: This tool can potentially help the pathologists by highlighting the cancerouis regions in large biopsy slides. It can also provide pre-processing for computer-aided diagnostic (CAD) systems to compute cancer-relevant features from the segmented regions.

Performance analysis of spread-spectrum wireless communication system
In this project, we designed and implemented a simulation framework for the performance analysis of hybrid direct sequence/slow frequency hopping (DS/SFH) and hybrid direct sequence/fast frequency hopping (DS/FFH) systems under multi-user interference and Rayleigh fading. The performance of direct sequence spread spectrum (DSSS), slow frequency hopping (SFH) and fast frequency hopping (FFH) systems was studied for varying processing gains under interference environment assuming equal bandwidth constraint with Binary Phase Shift Keying (BPSK) modulation and synchronous system. The results showed that the hybrid DS/FFH systems outperform both SFH and hybrid DS/SFH systems under Rayleigh fading and multi-user interference. Also, both hybrid DS/SFH and hybrid DS/FFH showed performance improvement with increasing spreading factor and decreasing number of hopping frequencies.

Significance: The framework provides relative performance analysis and merits of different signal spreading mechanisms in a wireless communication channel shared by several users with Rayleigh fading conditions.


[1] Y. Sharma, Thomas Ploetz, Nils Hammerla, Sebastian Mellor, Roisin McNaney, Patrick Olivier, Sandeep Deshmukh, Andrew McCaskie, and Irfan Essa,“Automated Surgical OSATS Prediction from Videos”, accepted for International Symposium on Biomedical Imaging (ISBI) 2014. PDF

[2] Y. Sharma, R. A. Moffitt, T. H. Stokes, Q. Chaudry, M. D. Wang, “Feasibility Analysis of High Resolution Tissue Image Registration Using 3-D Synthetic Data”, Journal of Pathology Informatics (JPI), 2012.

[3] A. J. Pierrot, E. E. Pujol, Y. Sharma, M. D. Wang, “Autonomous Point-Based Registration of Prostate Gland Tissue Images”, 4th International Congress on Image and Signal Processing (CISP), 15-17 October, Shangai, China, 2011.

[4] Q. Chaudry, Y. Sharma, S. H. Raza, M. D. Wang, “Development of a Novel 2D Color Map for Interactive Segmentation of Histological Images” International Symposium on Biomedical Imaging (ISBI), 2012.

[5] S. H. Raza, R.M. Parry,Y. Sharma, Q. Chaudry, R. A. Moffitt, A. N. Young, M. D. Wang, “Automated classification of renal cell carcinoma subtypes using bag-of-features”, 32nd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2010 Page(s): 6749-6752.

[6] Y. Sharma, S. H. Raza, Q. Chaudry, K. Y. Kong, S. Muller, Z. (Georgia) Chen and M. D. Wang, “Computer Assisted Grading of Squamous Cell Carcinoma of Head and Neck (SCCHN) Using Folate Receptor (FR) Expression in IHC Images”, Microscopic Image Analysis with Applications in Biology (MIAAB) 2009, NIH, Bethesda (available at http://www.miaab.org/miaab-2009-proceedings-02.html#cancer-imaging).

[7] S. H. Raza, Y. Sharma, Q. Chaudry, A. N. Young, M. D. Wang, “Automated classification of renal cell carcinoma subtypes using scale invariant feature transform”, Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2009, Page(s): 6687-6690.

[8] Y. Sharma, S. H. Raza, K. Y. Kong, Q. Chaudry, S. Muller, A. N. Young, Z. (Georgia) Chen, M. D. Wang, “PASuite: A preprocessing algorithm suite for cellular and molecular image classification in cancer diagnosis and treatment”, Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2008, Page(s): 3114 – 3117.

[9] K. Y. Kong, Y. Sharma, S. H. Raza, Zhuo Chen, S. Muller, M. D. Wang, “Using spiral intensity profile to quantify head and neck cancer”, IEEE International Conference on Bioinformatics and Bioengineering, (BIBE), 2008, Page(s): 1 – 6.

[10] Q. Chaudry, S. H. Raza, Y. Sharma, A. N. Young, M. D. Wang, Improving renal cell carcinoma classification by automatic region of interest selection, IEEE International Conference on Bioinformatics and Bioengineering, (BIBE), 2008, Page(s): 1 -6.

[11] M. L. Caldwell, R. A. Moffitt, J. Liu, R. M. Parry, Y. Sharma, M. D. Wang, "Simple quantification of multiplexed quantum dot staining in clinical tissue samples," International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), 2008, Page(s): 1907-1910.