Vaishnavi Subramanian

PhD Candidate
Electrical and Computer Engineering
University of Illinois at Urbana-Champaign

My research advisor is Prof Minh Do of the Electrical and Computer Engineering department.

I am interested in the broad areas of signal processing, computer vision, image processing and machine learning.

:)

Current & Past Affiliations

University of Illinois at Urbana-Champaign | PhD Candidate, Electrical and Computer Engineering
Advisor: Prof Minh Do | Aug 2018-Present
IBM Research, Almaden | Research Internship
Advisor: Dr Tanveer Syeda-Mahmood | May-Aug 2018
University of Illinois at Urbana-Champaign | Master of Science, Electrical and Computer Engineering
Advisor: Prof Minh Do | Aug 2016-May 2018
Indian Institute of Technology Bombay | Bachelor of Technology, Electrical Engineering
Advisors: Prof V Rajbabu & Prof Sibiraj Pillai | 2012-2016
Adobe Systems, Bangalore | Research Internship
Advisor: Dr Natwar Modani | May-July 2015

My Work

For a PDF version, please see here: [short], [long].

Radiogenomics for Lung Cancer | Research Project (Aug '18 - Present)
University of Illinois at Urbana-Champaign | Guides: Dr. Tanveer Syeda-Mahmood, IBM Research & Prof Minh Do, UIUC

Working on multimodal data fusion in the context of radiology and genomics to better predict survival of lung cancer patients. Aiming to integrate different data sources by designing a comprehensive multimodal neural network for the task with attention modules. Targeting the incorporation of existing knowledge in terms of gene interaction networks as a prior for the task. Developing the network with additional features of explainability and visualization for effective use in biological settings.

Identifying Central Vascular Lines in Chest X-Rays | Research Internship (Jun'18 - Aug '18)
IBM Research, Almaden, CA | Guide: Dr. Tanveer Syeda-Mahmood, IBM Research

Designed a multi-stage hybrid learning network to detect and classify central vascular lines, which are used to monitor heart function or to deliver medications. Effectively incorporated a top-down U-Net segmentation module, followed by a bottom-up feature extraction to identify the presence and type of the vascular lines.  Achieved 82% accuracy with 100% precision in line presence identification, and 80% accuracy with 95% precision in line classification, outperforming current techniques 

Correlating Features of Histopathology Images and Genomics | Master's Thesis Project (Feb '17 - May '18)
University of Illinois at Urbana-Champaign | Advisor: Prof Minh Do

Worked on nuclei segmentation tasks to separate out nuclei in histopathology images of breast cancer patients. Developed correlation techniques based on canonical correlation analysis to effectively combine nuclei features with gene expression data of the same patient. Identified the gene pathways represented by the specific correlations. Studied the effect of explicitly incorporating spatial distribution using Ripley's K-function of the different cells on the correlation. Published work at ISBI 2018 [link] and MICCAI 2018 [link]

Fat - Water Separation in Liver MRI Images | Software Development Project (Jan '17 - Sept '17 )
University of Illinois at Urbana-Champaign with PAREXEL, MA | Advisors: Prof Minh Do, David Bennett & Elif Sikoglu

Developed a user-friendly GUI tool in MATLAB to accurately determine the Proton Density Fat Fraction (PDFF) from magnitude MRI images of liver, for the diagnosis of Non-Alcoholic Fatty Liver Disease (NAFLD). Incorporated sparsity based optimization techniques to reduce the computation time 10-fold. Tested various models to identify the best model for the PDFF determination.

Matrix Completion in Atmospheric Radar | Undergraduate Thesis (July '15 - July '16)
Indian Institute of Technology Bombay | Advisors: Prof. Sibi Raj Pillai & Prof. V. Rajbabu 

Thesis PDF : here

Studied and modeled the mechanism of the radar system in order to recognize possibilities of improving its accuracy in Doppler frequency estimation. Identified methods to utilize the sparsity of radar returns in time and frequency domains to develop a more efficient signal processing block. Explored possibilities to fill in missing values from a set of incomplete measurements of radar data using matrix completion techniques in the range-frequency domain.

Multimedia Content Summarization | Research Internship (May - July '15)
Adobe Systems, Bangalore | Advisor: Dr Natwar Modani

Designed a novel summarization method to produce a more informative and coherent summary by accounting for information content of both images and sentences. Developed two algorithms to generate the summaries - an iterative greedy graph-based algorithm and a submodular-based optimization approach. Integrated deep learning methods such as convolutional neural nets for images and recursive autoencoders for text to represent the text and images.

Publications & Patents

  1. Subramanian, Vaishnavi, Hongzhi Wang, Joy T. Wu, Hui Tang, Arjun Sharma, Tanveer Syeda-Mahmood, "Identification of Central Vascular Lines in Chest X-rays." [Under Review]
     
  2. Subramanian, Vaishnavi, Weizhao Tang, Benjamin Chidester, Jian Ma, Minh N. Do, " Integration of Spatial Distribution in Imaging-Genetics " In International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2018 [link ]
     
  3. Subramanian, Vaishnavi, Benjamin Chidester, Jian Ma, Minh N. Do, "Correlating Cellular Features with Gene Expression using CCA " In IEEE International Symposium on Biomedical Imaging (ISBI) 2018, pp. 805-808 [link]
     
  4. Modani, Natwar, Pranav Maneriker, Gaurush Hiranandani, Atanu R. Sinha, Vaishnavi Subramanian, and Shivani Gupta. "Summarizing Multimedia Content." In International Conference on Web Information Systems Engineering (WISE), pp. 340-348. Springer International Publishing, 2016. [link]
     
  5. Modani, Natwar, Vaishnavi Subramanian, Shivani Gupta, Pranav Ravindra Maneriker, Gaurush Hiranandani, and Atanu R. Sinha. "Determining quality of a summary of multimedia content." U.S. Patent 9,454,524, issued September 27, 2016. [link]
     
  6. Modani, Natwar, Vaishnavi Subramanian, Shivani Gupta, Pranav R. Maneriker, Gaurush Hiranandani, and Atanu R. Sinha. "Multimedia Document Summarization." U.S. Patent Application 14/947,964, filed November 20, 2015. [link]

Drop By

... for a cup of tea, a smile or anything else!

111, Coordinated Science Lab
University of Illinois at Urbana-Champaign

... or email at

vs5[AT]illinois[DOT]edu