I have ran across this paper and had a great time reading it. Namely Stress Testing Pathology Models with Generated Artifacts. This research was carried out by a team of people from Department of Computational Medicine and Bioinformatics, University of Michigan and Department of Pathology, University of Michigan Medical School. In this research paper they studied 7 types of artifacts that could appear on tissue images:
\nAlong side the said paper they also made their custom toolkit which they used for this research publicly accessible. You can find the repository link in the Conclusions section of their work or you can get to it from here: Histopath Failure Modes from Systems Imaging Bioinformatics Lab at the University of Michigan.
\nAs I have said in the beginning I had a great time reading their work. And I wanted to share it. So if you are interested in how the said artifacts are generated you can find the notebook I have put together using multiple sources here.
\nNow then I want to take this chance to again put an emphasis on the fact that THE NOTEBOOK I HAVE PROVIDED HERE DOES NOT SHOWCASE MY WORK. In the notebook I have PUT TOGETHER:
\nSuch that this notebook only showcases works and researches carried out by those included above. I have only put some additional reading materials in the Conclusions section and defined some utility functions for ease of use throughout the notebook.
\nThen again, I have greatly enjoyed reading their work so please do check them out:
\nHave fun!
", "messages": [ { "id": 1839186, "postDate": "2022-07-01T07:20:22.053Z", "content": "I have ran across this paper and had a great time reading it. Namely Stress Testing Pathology Models with Generated Artifacts. This research was carried out by a team of people from Department of Computational Medicine and Bioinformatics, University of Michigan and Department of Pathology, University of Michigan Medical School. In this research paper they studied 7 types of artifacts that could appear on tissue images:
\nAlong side the said paper they also made their custom toolkit which they used for this research publicly accessible. You can find the repository link in the Conclusions section of their work or you can get to it from here: Histopath Failure Modes from Systems Imaging Bioinformatics Lab at the University of Michigan.
\nAs I have said in the beginning I had a great time reading their work. And I wanted to share it. So if you are interested in how the said artifacts are generated you can find the notebook I have put together using multiple sources here.
\nNow then I want to take this chance to again put an emphasis on the fact that THE NOTEBOOK I HAVE PROVIDED HERE DOES NOT SHOWCASE MY WORK. In the notebook I have PUT TOGETHER:
\nSuch that this notebook only showcases works and researches carried out by those included above. I have only put some additional reading materials in the Conclusions section and defined some utility functions for ease of use throughout the notebook.
\nThen again, I have greatly enjoyed reading their work so please do check them out:
\nHave fun!
", "rawMarkdown": "I have ran across this paper and had a great time reading it. Namely [Stress Testing Pathology Models with Generated Artifacts](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8721870/pdf/JPI-12-54.pdf). This research was carried out by a team of people from *Department of Computational Medicine and Bioinformatics, University of Michigan* and *Department of Pathology, University of Michigan Medical School*. In this research paper they studied 7 types of artifacts that could appear on tissue images:\n\n- Bubbles\n- Tissue fold\n- Illumination\n- Marker line\n- Sectioning\n- Stain alteration\n- Tissue tears\n\nAlong side the said paper they also made their **custom toolkit** which they used for this research publicly accessible. You can find the repository link in the Conclusions section of their work or you can get to it from here: [Histopath Failure Modes](https://github.com/Systems-Imaging-Bioinformatics-Lab/histopath_failure_modes.git) from [Systems Imaging Bioinformatics Lab at the University of Michigan](https://sibl.lab.medicine.umich.edu). \n\nAs I have said in the beginning I had a great time reading their work. And I wanted to share it. So if you are interested in how the said artifacts are generated you can find the notebook I have put together using multiple sources [here](https://www.kaggle.com/code/temuujinerdene/augmentation-with-generated-artifacts/settings). \n\nNow then I want to take this chance to **again** put an emphasis on the fact that **THE NOTEBOOK I HAVE PROVIDED HERE DOES NOT SHOWCASE MY WORK**. In the notebook I have **PUT TOGETHER**:\n\n1. Splitting the original training image into tiles and overlaying the corresponding annotation masks of that image are entirely from [FTUs⚕️Segm: decompose 🖽 large images to tiles](https://www.kaggle.com/code/jirkaborovec/ftus-segm-decompose-large-images-to-tiles) notebook by [Jirka Borovec](https://www.kaggle.com/jirkaborovec)\n2. Toolkit and methodology is from *Department of Computational Medicine and Bioinformatics, University of Michigan* and *Department of Pathology, University of Michigan Medical School*.\n\nSuch that this notebook only showcases works and researches carried out by those included above. I have only put some additional reading materials in the Conclusions section and defined some utility functions for ease of use throughout the notebook.\n\nThen again, I have greatly enjoyed reading their work so please do check them out:\n- [Stress Testing Pathology Models with Generated Artifacts](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8721870/pdf/JPI-12-54.pdf)\n- [Histopath Failure Modes](https://github.com/Systems-Imaging-Bioinformatics-Lab/histopath_failure_modes.git)\n- [FTUs⚕️Segm: decompose 🖽 large images to tiles](https://www.kaggle.com/code/jirkaborovec/ftus-segm-decompose-large-images-to-tiles)\n\nHave fun!", "votes": 17 }, { "id": 1841151, "postDate": "2022-07-02T21:07:19.303Z", "content": "Really appreciate you taking the time to both read and summarize the research in this post. very helpful!
", "rawMarkdown": "Really appreciate you taking the time to both read and summarize the research in this post. very helpful!\n", "votes": 1 }, { "id": 1847879, "postDate": "2022-07-08T08:16:47Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ], "comments": [ { "id": 1841151, "author_name": "The Devastator", "author_url": "", "post_date": "2022-07-02T21:07:19.303000", "content": "Really appreciate you taking the time to both read and summarize the research in this post. very helpful!
", "votes": 1, "replies": [] }, { "id": 1847879, "author_name": "", "author_url": "", "post_date": "2022-07-08T08:16:47", "content": "", "votes": 1, "replies": [] } ], "raw_markdown_by_id": { "1839186": "I have ran across this paper and had a great time reading it. Namely [Stress Testing Pathology Models with Generated Artifacts](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8721870/pdf/JPI-12-54.pdf). This research was carried out by a team of people from *Department of Computational Medicine and Bioinformatics, University of Michigan* and *Department of Pathology, University of Michigan Medical School*. In this research paper they studied 7 types of artifacts that could appear on tissue images:\n\n- Bubbles\n- Tissue fold\n- Illumination\n- Marker line\n- Sectioning\n- Stain alteration\n- Tissue tears\n\nAlong side the said paper they also made their **custom toolkit** which they used for this research publicly accessible. You can find the repository link in the Conclusions section of their work or you can get to it from here: [Histopath Failure Modes](https://github.com/Systems-Imaging-Bioinformatics-Lab/histopath_failure_modes.git) from [Systems Imaging Bioinformatics Lab at the University of Michigan](https://sibl.lab.medicine.umich.edu). \n\nAs I have said in the beginning I had a great time reading their work. And I wanted to share it. So if you are interested in how the said artifacts are generated you can find the notebook I have put together using multiple sources [here](https://www.kaggle.com/code/temuujinerdene/augmentation-with-generated-artifacts/settings). \n\nNow then I want to take this chance to **again** put an emphasis on the fact that **THE NOTEBOOK I HAVE PROVIDED HERE DOES NOT SHOWCASE MY WORK**. In the notebook I have **PUT TOGETHER**:\n\n1. Splitting the original training image into tiles and overlaying the corresponding annotation masks of that image are entirely from [FTUs⚕️Segm: decompose 🖽 large images to tiles](https://www.kaggle.com/code/jirkaborovec/ftus-segm-decompose-large-images-to-tiles) notebook by [Jirka Borovec](https://www.kaggle.com/jirkaborovec)\n2. Toolkit and methodology is from *Department of Computational Medicine and Bioinformatics, University of Michigan* and *Department of Pathology, University of Michigan Medical School*.\n\nSuch that this notebook only showcases works and researches carried out by those included above. I have only put some additional reading materials in the Conclusions section and defined some utility functions for ease of use throughout the notebook.\n\nThen again, I have greatly enjoyed reading their work so please do check them out:\n- [Stress Testing Pathology Models with Generated Artifacts](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8721870/pdf/JPI-12-54.pdf)\n- [Histopath Failure Modes](https://github.com/Systems-Imaging-Bioinformatics-Lab/histopath_failure_modes.git)\n- [FTUs⚕️Segm: decompose 🖽 large images to tiles](https://www.kaggle.com/code/jirkaborovec/ftus-segm-decompose-large-images-to-tiles)\n\nHave fun!", "1841151": "Really appreciate you taking the time to both read and summarize the research in this post. very helpful!\n", "1847879": "" } }