Determining the histological structure of tumors is important for understanding spatial tumor biology and to identify the pathological mechanisms underlying cancer. By analyzing tissue architecture, cellular organization, and interactions within the tumor microenvironment, researchers can gain insight into mechanisms behind, for example cancer recurrence, response to treatment and other clinically important features where prediction models are warranted.
The project proposal
The last couple of years have focused on deep learning models for analyzing images derived from biological tissue. There have been models trained to detect different cell type representations in H&E images. One is Hover-Net which both segments and classifies cells into normal epithelial, malignant/dysplastic epithelial, fibroblast, muscle, inflammatory, endothelial or miscellaneous (necrotic, mitotic and cells)1. Another software ConvPath uses a model to define lymphocytes, tumor cells, stroma cells and their regional border2. Detecting and identifying the different cells and regions enables downstream tasks such as investigation of tumour infiltrating lymphocytes (TILs). For example, TILs have proven to predict cancer recurrence3.
The proposed project will explore the different interactions/spatial metrics related to defined cell types, including cell to cell distances, cellular niches and correlate these with clinical parameters such as outcome or previously defined high-risk patient parameters. The goal is to profile and understand the TME better and to find patterns or structures that can be used to stratify patient’s tumor in more detail. The steps would be to identify cells in the tissue by segmentation, then further classify the cells to known cell types. The downstream task involves calculating the distance between cell types based on known metrics such as using the nearest neighbor distance, (Ripley’s) K-function (cross), pair correlation function, neighborhood analysis etc.
Data-set available for the project
The project will have access to HTX staining of duplicate tissues from 650 patients diagnosed with diffuse large B-cell lymphoma (DLBCL). This is a unique clinical dataset with high potential for translational publications, as well as method development.
Methods
The project will evaluate different workflows for segmenting and classifying cells in DLBCL using python mainly because of the model implementations and Pythons image handling. To tailor the task to lymphoma, different models will be evaluated including HoLy-Net4, Hover-Net with the aim to extract spatial metrics from the full image dataset and to perform data integration with clinical data. Evaluation of the classification model will be performed based on previously available multiplex immunofluorescence data. Downstream statistical tasks can be done in R or Python.
Requirements
We seek a bioinformatic student that is proficient in R, and with basic knowledge of Python. You will have the possibility to deepen your experience in Python and gain hands-on experience on high-throughput image and down-stream data handling as well as data integration.
Principle investigator/supervisor: Sara Ek
Practical supervisor: Daniel Nilsson
Department of Immunotechnology, Lund University
Starting date: flexible, reach out at sara.ek@immun.lth.se to discuss your interest
Length/credits: (30-60 hp) the project can be adapted to fit as either course project or longer combined master thesis projects
References: previous students from the bioinformatic program include Teodor Alling, Mattis Knulst, Daniel Nilsson and Markus Heidrich. Two of previous students are today employed within the group.
Literature
- 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. https://doi.org/10.1016/j.media.2019.101563
- Wang, S., Wang, T., Yang, L., Yang, D. M., Fujimoto, J., Yi, F., Luo, X., Yang, Y., Yao, B., Lin, S., Moran, C., Kalhor, N., Weissferdt, A., Minna, J., Xie, Y., Wistuba, I. I., Mao, Y., & Xiao, G. (2019). ConvPath: A software tool for lung adenocarcinoma digital pathological image analysis aided by a convolutional neural network. EBioMedicine, 50, 103–110. https://doi.org/10.1016/j.ebiom.2019.10.033
- Corredor, G., Wang, X., Zhou, Y., Lu, C., Fu, P., Syrigos, K., Rimm, D. L., Yang, M., Romero, E., Schalper, K. A., Velcheti, V., & Madabhushi, A. (2019). Spatial Architecture and Arrangement of Tumor-Infiltrating Lymphocytes for Predicting Likelihood of Recurrence in Early-Stage Non-Small Cell Lung Cancer. Clinical cancer research : an official journal of the American Association for Cancer Research, 25(5), 1526–1534. https://doi.org/10.1158/1078-0432.CCR-18-2013
- Naji, H., Sancere, L., Simon, A., Büttner, R., Eich, M. L., Lohneis, P., & Bożek, K. (2024). HoLy-Net: Segmentation of histological images of diffuse large B-cell lymphoma. Computers in biology and medicine, 170, 107978. https://doi.org/10.1016/j.compbiomed.2024.107978