I like science, cooking, and mountain stuff.
My Curriculum Vitae
My Science:
Project Euler (417130_6d81d59107417228c89893205d252259)
My Cooking:
![]() |
![]() |
![]() |
![]() |
(made with my brother Jonathan Buckley) |
My Mountain Stuff:
![]() |
![]() |
![]() |
![]() |
(I also happen to like Crystie and beer 🙂 )
Love you!!!!
thanks for the citation, bro!
You’re one in a million. Keep up the great work and keep having fun!!!
Dear Andrew
I want invite you to participate in seminar Warsaw, Poland about deep learning in histopathology but can’t find your address.
Replay me ; September 2018
I can be contacted at: andrew.janowczyk[AT]case.edu
Great job you are doing. So proud of your work.
Hey, thanks for the information about .svs. I am at my Bioengineering Grade Project and it was really useful. Also I like your mountaim stuff. Reminds me the book ‘Mount analogue’ by René Dauman.
glad to hear it and thanks!
Andrew, you were my coolest podcast guest in 2020! Thanks so much for the interview and I hope to meet you again soon when we can travel to conferences again!
My pleasure! Thanks for having me!
Dear Andrew
As we are aware, in the realm of digital pathology, especially in the context of The Cancer Genome Atlas (TCGA) project, Whole Slide Images (WSIs) at 40X magnification typically correspond to a resolution of 0.25 microns per pixel (mpp), and those at 20X magnification correspond to 0.5 mpp. However, I’ve encountered an intriguing anomaly in some of the TCGA WSIs where the images are at 0.25 mpp resolution, yet the stated magnification is 20X.
This observation leads me to wonder if such images have undergone a downsampling process or if there’s another explanation behind this discrepancy. Additionally, I am curious if there is any documentation or resources within the TCGA project or elsewhere that address this specific scenario
I don’t think its a downsampling issue – i think the slides themselves were scanned at 20x due to cost/time/resource decisions. 20x is not actually that “uncommon” for many situations as its already provides enough context + detail to make accurate diagnoses. To my knowledge, there is no TCGA documentation which systematically discusses these characteristics of the data set. That said, i would point out that the TCGA was initially/primarily created as a resource for -omics modality, not digital pathology. the DP images were originally included as a value add to the omics to help the researchers better characterize their findings. This can help explain the heterogeneity we see — they were not originally collected for the purpose of computational analysis!
Andrew, it is known that TCGA’s WSI, 0.25 mpp corresponds to a 40x magnification and 0.5 mpp corresponds to a 20x magnification. But when I downloaded TCGA-COAD and READ Diagnostic Whole Slide Imaging (WSI) data and observed that the micron-per-pixel (mpp) is approximately 0.25 across all the WSIs no matter the objective magnifications is 20 or 40. Can you tell me why?
This is a great question, i unforunately don’t know the reason why. Sometimes this is an error in the scanner – for example i know one scanner which scans at 40x because it uses a 20x mag with a 2x doubler, but the metadata is only aware of the 20x mag, and thus this gets recorded in the metadata (incorrectly). I think the only real options here are to look at e.g., (a) file size, which is going to be much larger for 40x specimens of the same tissue area, or (b) try to find lymphocytes (there are always a few!) and measure them. They are constant in size/presentation everywhere in the body, and thus are a good landmark for estimating magnification