
Jens Rittscher
Dr. Jens Rittscher has been appointed as a University Research Lecturer in 2013 and he is the first joint academic appointment between the Institute of Biomedical Engineering and the Nuffield Department of Medicine. In particular his work supports the Target Discovery Institute and Ludwig Institute of Cancer Research. In addition to his research in the field of biomedical imaging, Jens Rittscher has worked extensively in the area of video surveillance, the automatic annotation of video, and understanding of volumetric seismic data.
Before coming to Oxford in 2013 Jens Rittscher led the Computer Vision Laboratory at GE Global Research in Niskayuna, NY, USA. He joined GE in 2001 after completing his PhD at the Department of Engineering Science at University of Oxford. During this time he was part of the Visual Dynamics Group led by Andrew Blake. He received his Diploma in Mathematics and Computer Science from the University of Bonn, Germany. Jens Rittscher held a position as an adjunct assistant professor at the Rensselaer Polytechnic Institute. He is a member of IEEE and acts as an elected member of the IEEE SPS Technical Committee on Bio Image and Signal Processing.
Computational Tissue Analysis
Lecture 1: Background & imaging methods
Learning objective: Become familiar with the workflow in cellular pathology workflow in the clinic. Learn what imaging methods are being applied (H&E, IHC). Become familiar with more advanced imaging methods such as multiplexing and see examples of imaging tissues in 3D.
Outline:
- Highlight the clinical importance of cellular pathology
- Work through and example: tissue sampling, fixation, staining
- How can we generate contrast H&C, IHC
- What do pathology reports look like
- Summarise digitisation of histology slides
- What is the potential impact of digital pathology?
- Provide a few examples
- The digital pathology workflow
- Example study: automatic assessment of tissue quality [L1-quality]
- Tools for annotating histology slides
- The automated annotator - what went wrong?
References:
[L1-quality] Haghighat, M., Browning, L., Sirinukunwattana, K., Malacrino, S., Khalid Alham, N., Colling, R., ... & Rittscher, J. (2022). Automated quality assessment of large digitised histology cohorts by artificial intelligence. Scientific Reports, 12(1), 1-16. DOI: 10.1038/s41598-022-08351-5
Lecture 2: Segmenting tissue architecture components
Learning objectives: Become familiar with established tools for segmenting tissue components such as nuclei, vessels, and glands. Understand what the limitations of accurate segmentation on 2D sections are.
Outline:
- Segmenting tissue components on thin 2D histology sections. Outline the tasks and highlight challenges
- A brief review of classical segmentation methods (e.g. superpixels, etc.)
- UNet for segmenting nuclei
- Cellpose with an example
- StarDist with a 3D example
- Exemplar: segmentation of glands in prostate cancer
- Hover-net as an example of learning segmentation and classification simultaneously
- Example: A quantitative analysis that supports the assessment of donor kidneys for transplantation
References:
[L2-cellpose] Stringer, C., Wang, T., Michaelos, M., & Pachitariu, M. (2021). Cellpose: a generalist algorithm for cellular segmentation. Nature methods, 18(1), 100-106. DOI: 10.1038/s41592-020-01018-x
[L2-stardist] Schmidt, U., Weigert, M., Broaddus, C., & Myers, G. (2018, September). Cell detection with star-convex polygons. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 265-273). Springer, Cham DOI: 10.1007/978-3-030-00934-2_30
[L2-hovernet] Graham, S., Vu, Q. D., Raza, S. E. A., Azam, A., Tsang, Y. W., Kwak, J. T., & Rajpoot, N. (2019). Hover-net: Simultaneous segmentation and classification of nuclei in multi-tissue histology images. Medical Image Analysis, 58, 101563. DOI: 10.1016/j.media.2019.101563
[L2-kidney] Tam, K. H., Soares, M. F., Kers, J., Sharples, E., Ploeg, R., Kaisar, M., & Rittscher, J. (2022). Predicting Clinical Endpoints and Visual Changes with Quality-Weighted Tissue-based Histological Features. medRxiv. DOI: 10.1101/2022.03.30.22269826
Lecture 3: Modelling spatial relationships using graphs
Learning objectives: Become familiar with techniques of modelling spatial relationships. Study the potential of applying graph-based deep learning to model biologically relevant features (e.g. fibrosis)
Outline:
- Motivate the need for modelling the spatial relationships in tissues
- Graph convolutional networks
- Modelling spatial relationships in tissues [L4-hact-net]
- Liver-fibrosis: a new way for scoring liver fibrosis
- Identifying cell communities in the bone marrow
References:
[L4-hact-net] Pati, P., et al.: HACT-Net: a hierarchical cell-to-tissue graph neural network for histopathological image classification. In: Sudre, C.H., et al. (eds.) UNSURE/GRAIL -2020. LNCS, vol. 12443, pp. 208–219. Springer, Cham (2020). DOI 10.1007/978-3-030-60365-6_20
[L3-fibrosis] Wojciechowska, M., Malacrino, S., Garcia Martin, N., Fehri, H., & Rittscher, J. (2021, September). Early Detection of Liver Fibrosis Using Graph Convolutional Networks. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 217-226). Springer, Cham. DOI 10.1007/978-3-030-87237-3_21
[L3-communities] unpublished work, perhaps we can use a different paper
Lecture 4: Correlating morphology with molecular information
Learning objectives: Become familiar with the relationship between pathology and genetic and molecular information. Reflect on the opportunities these approaches can provide. Discuss examples of how morpho-molecular features need to be linked with tissue segmentation.
Outline:
- Going beyond established clinical grading systems
- Molecular subtyping in colorectal cancer
- Introduce the concept of morph-molecular correlates
- Review the MSI example [L4-msi]
- Review the imCMS example [L4-imCMS]
- Outline what opportunities related methods open up
- Include attention here?
References:
[L4-msi] Kather, J. N., Pearson, A. T., Halama, N., Jäger, D., Krause, J., Loosen, S. H., ... & Luedde, T. (2019). Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer. Nature medicine, 25(7), 1054-1056. DOI: https://doi.org/10.1038/s41591-019-0462-y
[L4-imCMS] Sirinukunwattana, K., Domingo, E., Richman, S. D., Redmond, K. L., Blake, A., Verrill, C., ... & Koelzer, V. H. (2021). Image-based consensus molecular subtype (imCMS) classification of colorectal cancer using deep learning. Gut, 70(3), 544-554. DOI: 10.1136/gutjnl-2019-319866