Dr. Tanishq Mathew Abraham

Dr. Tanishq Mathew Abraham

20 year old BME PhD from UC Davis. Founding CEO of MedARC. Part-time @ Stability AI. Former member of the Levenson Lab. Applying generative AI to medicine.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

portfolio

publications

FIBI (Fluorescence Imitating Brightfield Imaging) for rapid, slide-free dermatopathology

Published in Journal of Cutaneous Pathology, 2022

Images acquired with FIBI are comparable to traditional H&E-stained slides, suggesting that this rapid, inexpensive, and non-destructive microscopy technique is a conceivable alternative to standard histopathology processes especially for time-sensitive procedures and in settings with limited histopathology resources.

Download here

talks

Slide-free MUSE Microscopy to H&E Histology Modality Conversionvia Unpaired Image-to-Image Translation GAN Models

Published:

Link to paper

Link to slides

Abstract: MUSE is a novel slide-free imaging technique for histological examination of tissues that can serve as an alternative to traditional histology. In order to bridge the gap between MUSE and traditional histology, we aim to convert MUSE images to resemble authentic hematoxylin- and eosin-stained (H&E) images. We evaluated four models: a non-machine-learning-based color-mapping unmixing-based tool, CycleGAN, DualGAN, and GANILLA. CycleGAN and GANILLA provided visually compelling results that appropriately transferred H&E style and preserved MUSE content. Based on training an automated critic on real and generated H&E images, we determined that CycleGAN demonstrated the best performance. We have also found that MUSE color inversion may be a necessary step for accurate modality conversion to H&E. We believe that our MUSE-to-H&E model can help improve adoption of novel slide-free methods by bridging a perceptual gap between MUSE imaging and traditional histology.

What are CycleGANs? (a novel deep learning tool in pathology)

Published:

Link to talk

Link to code

Abstract: A CycleGAN is a variant of the generative adversarial network (GAN) architecture designed to handle unpaired image conversion problems. In this talk, Tanishq will describe what are CycleGANs, what applications they are best suited for, and their pitfalls. Additionally, he will walk through code for model training and inference as part of a code demo. Finally, he will present his own research applying CycleGANs (and related models) to pathology and microscopy.

Bio: Tanishq Abraham is considered a child genius and a prodigy. He graduated high school at 10 years old with a 4.0 GPA and at 11 he obtained 3 college Associate Degrees also with a perfect 4.0 GPA. At 14, he graduated from University of California, Davis as a biomedical engineer with summa cum laude. As an undergrad, he has presented papers at several conferences. At 14, he was the first author on a review paper about smart bio-inspired vesicles and it’s biomedical engineering applications published in IOP Physical Biology. At 16, he is a published book chapter author in the book “Artificial Intelligence and Deep Learning in Pathology”. Now at 17, he is a 2nd year PhD student in biomedical engineering at UC Davis. He has worked for a year in the Levenson lab at UC Davis on applying deep learning to novel microscopy techniques in order to enable digital pathology applications. His recent conference was presenting his PhD research at the ICML2020 Computational Biology workshop. Tanishq has also reviewed few books that includes two science fiction books and recently a machine learning book.

Follow him on Twitter: https://twitter.com/iScienceLuvr

#MachineLearning #DeepLearning #GAN

teaching

BIM 1 Fall 2019

Introduction to Biomedical Engineering, UC Davis, Department of Biomedical Engineering, 2019

Led two 3 hr long weekly discussion sections that involved:

  • Mentoring teams for design challenge 1 & 2
  • Design Challenge #1 (Rube Goldberg machine with various constraints and requirements)
  • Design Challenge #2 (Assistive shoe-tying machine for hemiplegic children)
  • Solidworks tutorial
  • lecturing about the engineering design process
  • reviewing the main lecture content
  • grading quizzes and Solidworks assignment for about 50 undergrad freshman and sophomore ‏‏‎ ‎‏‏‎ ‎‏‏engineering students

Mentored 11 teams for the design challenges. Out of the total 35 teams in the course, one of the teams that I mentored won the challenge.

BIM 1 Fall 2020

Introduction to Biomedical Engineering, UC Davis, Department of Biomedical Engineering, 2020

Led two 3 hr long weekly discussion sections that involved:

  • Mentoring teams for the class design challenge (assistive shoe-tying machine for hemiplegic children)
  • Solidworks tutorial
  • lecturing about the engineering design process
  • reviewing the main lecture content
  • grading quizzes and Solidworks assignment for about 60 undergrad freshman and sophomore ‏‏‎ ‎‏‏‎ ‎‏‏engineering students
  • organization and coordination of the design challenge for the entire class Due to the virtual nature of the class that year, significant work was put in to ensure that the class format (which usually involves lots of hands-on activities as part of the design challenge) translated well to the new format.

Fast.ai Part 1, 2022

Practical Deep Learning for Coders, https://course.fast.ai, 2022

As a teaching assistant for the class, I was answering student questions during and after lessons and managing the course forum.

Fast.ai Part 2, 2022

Practical Deep Learning for Coders, Online, 2022

As an instructor for this ongoing course, I am preparing Python notebooks and code for the class, recording lectures with other instructors and course contributors (led by Jeremy Howard), answering student questions, managing the course forum, and conducting research along the way.

The first set of lectures were released here.