The distribution of lymphocytes within the tissue samples and with respect to major histopathological landmarks may be used for predicting clinically-relevant outcomes. However, the application of graph-based methods to whole-slide image data is hindered by the sheer amount of data. The present thesis will explore methods for graph-based modelling of spatial distributions of lymphocytes in colorectal cancer samples. The objective is to study both the utility of said methods for extracting actionable features from spatial data as well as the scalability of these methods. The candidate will be provided with a set of pre-processed data, including the type and position of the lymphocytes in tumor sections, and histopathological annotations (morphological regions and - if present - invasion margin).

The candidate will
(i) review the relevant literature;
(ii) in accord with the supervisor, select one or two approaches for detailed investigation;
(iii) implement or adapt existing implementation to the problem at hand;
(iv) use the implemented method(s) for extracting spatial features and study their correlations with clinical outcomes (such as relapse-free survival) or other clinical parameters (e.g. tumor stage);
(v) study the scalability of the method(s) with respect to the number of cells and propose methods for improving the scalability.

Literature:
1. Z. Wu et al., Graph deep learning for the characterization of tumour microenvironments from spatial protein profiles in tissue specimens, Nat. Biomed. Eng, vol. 6, no. 12, Art. no. 12, Dec. 2022, doi: 10.1038/s41551-022-00951-w.
2. G. Palla et al., Squidpy: a scalable framework for spatial omics analysis, Nat Methods, vol. 19, no. 2, Art. no. 2, Feb. 2022, doi: 10.1038/s41592-021-01358-2.
3. J. D. Garciá-Arteaga, G. Corredor, X. Wang, V. Velcheti, A. Madabhushi, and E. Romero, A lymphocyte spatial distribution graph-based method for automated classification of recurrence risk on lung cancer images, in 13th International Conference on Medical Information Processing and Analysis, SPIE, Nov. 2017, pp. 9–15. doi: 10.1117/12.2285653.
4. H. Failmezger, S. Muralidhar, A. Rullan, C. E. de Andrea, E. Sahai, and Y. Yuan, Topological Tumor Graphs: A Graph-Based Spatial Model to Infer Stromal Recruitment for Immunosuppression in Melanoma Histology, Cancer Research, vol. 80, no. 5, pp. 1199–1209, Mar. 2020, doi: 10.1158/0008-5472.CAN-19-2268.