15th Annual Symposium
Physics of Cancer
Leipzig, Germany
Sept. 30 - Oct. 2, 2024
Contributed Talk
Automated quantitative histology of colorectal liver metastases to assess tumour response to chemotherapy in correlation with magnetic resonance elastography.
Kristin Brüggemann1, Lisa-Marie Skrip2, Yasmine Safraou4, Simon Moosburner2,3, Peter Tang2, Jing Guo4, Steffen Görner4, Heiko Tzschätzsch1, Kilian Alexander Walter2, Clarissa Hosse4, Uli Fehrenbach4, Alexander Arnold5, Dominik Paul Modest6, Felix Krenzien2,3, Wenzel Schöning2, Thomas Malinka2, Johann Pratschke2, Björn Papke5, Josef Alfons Käs7, Ingolf Sack4, Igor Maximilian Sauer2, Karl Herbert Hillebrandt2,3
1Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt- Universität zu Berlin, Institute of Medical Informatics, Charitéplatz 1, 10117 Berlin, Germany
2Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt- Universität zu Berlin, Department of Surgery, Augustenburger Platz 1, 13353 Berlin, Germany
3Berlin Institute of Health at Charité – Universitätsmedizin Berlin, BIH Academy, Clinician Scientist Program, Anna-Louisa-Karsch-Straße 2, 10178 Berlin, Germany
4Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt- Universität zu Berlin, Department of Radiology, Charitéplatz 1, 10117 Berlin, Germany
5Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt- Universität zu Berlin, Institute of Pathology, Charitéplatz 1, 10117 Berlin, Germany
6Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt- Universität zu Berlin, Department of Hematology, Oncology, and Cancer Immunology (CVK/CCM), Charitéplatz 1, 10117 Berlin, Germany
7Soft Matter Physics Division, Peter Debye Institute for Soft Matter Physics, University of Leipzig, Linnéstraße 5, 04103 Leipzig, Germany
Contact: 
Colorectal cancer is the third most common cancer and the fourth leading cause of cancer-related mortality worldwide [1]. There is a need in oncology for new diagnostic methods and tumour markers which are specific to the efficacy of therapeutic approaches. Classical histology-based methods can be supported by quantitative histology and correlated with biophysical imaging modalities such as magnetic resonance elastography (MRE) [2].
Colorectal liver metastases (CRLM) samples were resected from 34 patients. 28 from the 34 CRLM patients (82.4 %) were treated with chemotherapy before the tumour resection. The samples were manually analysed according to their histology and automatically using quantitative histology (qHisto). The regression grades were used to evaluate the response of the tumour to chemotherapy. Therefore, the H&E stained samples were evaluated by a pathologist. For the automated analysis of the histological slides by qHisto a pre-trained segmentation model StarDist was used to segment the tumour cell nuclei [3, 4]. To quantify the shape of the nuclei, three different parameters were calculated: Nucleus size (in µm2), aspect ratio and density (in 1/µm2) [5]. The nucleus size was calculated by measuring the area of each nucleus and taking the mean over all nuclei. The aspect ratio for elliptical shapes was calculated by dividing the major axis length by the minor axis length. Finally, the nucleus density was calculated by taking the number of nuclei per tissue area. MRE was performed in a compact benchtop MRI scanner with MRE vibration unit operated in a frequency range of 500 to 5300 Hz as further detailed in [6].
According to the histopathological analysis 14.7% of the samples showed a major response and 17.6% showed a partial response to chemotherapy while the rest of the samples (67.6%) showed no response to chemotherapy. qHisto reflected these results by showing smaller nucleus area and nucleus density in samples with major response (19.8 µm2 ± 0.98, p = 0.0002) and partial response (20.9 µm2 ± 1.00, p = 0.0053) than no response (23.4 µm2 ± 1.4). In addition, aspect ratio showed higher values for both response types compared to the non-responder and showed a significant difference between partial and no response (1.54 ± 0.046 vs. 1.48 ± 0.027, p = 0.0003). The area under the curve (AUC) for major vs. non-major was 0.95 (sensitivity = 1, specificity = 0.86) for the nucleus area, 0.67 (sensitivity = 0.5, specificity = 0.93) for the nucleus aspect ratio and 0.83 (sensitivity = 0.75, specificity = 0.93) for the nucleus density. From MRE the shear wave speed c (in m/s) and the penetration rate a (in m/s) were measured. In addition, the frequency-independent parameters shear modulus (in Pa) µ and the dimensionless powerlaw exponent α according to the springpot model were calculated to evaluate stiffness and fluidity of the tissues. c showed a correlation with the tumour regression grade for 1300 Hz and 2100 – 5300 Hz (all p < 0.05). In comparison to the histopathological response a significant correlation for c was observed for 4900 Hz (p = 0.03) and 5300 Hz (p = 0.027) overall making CRLM samples with major response stiffer than partial or non-responders. The frequency-independent α-parameter showed lower values in major response than non-response samples (0.43 ± 0.08 vs. 0.52 ± 0.08, p = 0.03) indicating a more elastic-solid tumour behaviour due to chemotherapy. AUC for comparing major vs. non-major was 0.82 (sensitivity = 0.86, specificity = 0.8) for α.
The automated qHisto analysis was able to reproduce the gold standard of manual histological analysis in correlation with tissue fluidity as quantified by ex-vivo MRE. qHisto can be used to evaluate the tumour response to therapy in CRLM and, in combination with ex-vivo MRE, could be of value for improving the reliability and objectivity of standard histology.
[1]Mármol, I., Sánchez-de-Diego, C., Pradilla Dieste, A., Cerrada, E., & Rodriguez Yoldi, M. J.: Colorectal carcinoma: a general overview and future perspectives in colorectal cancer, International journal of molecular sciences, 18(1), 197. (2017)
[2]Venkatesh, S. K., Yin, M., & Ehman, R. L.: Magnetic resonance elastography of liver: technique, analysis, and clinical applications, Journal of magnetic resonance imaging, 37(3), 544-555. (2013)
[3]Schmidt, U., Weigert, M., Broaddus, C., & Myers, G.: Cell detection with star-convex polygons., In Medical Image Computing and Computer Assisted Intervention–MICCAI 2018: 21st International Conference, Granada, Spain, September 16-20, 2018, Proceedings, Part II 11 (pp. 265-273). Springer International Publishing. (2018)
[4]Weigert, M., Schmidt, U., Haase, R., Sugawara, K., & Myers, G.: Star-convex polyhedra for 3D object detection and segmentation in microscopy, In Proceedings of the IEEE/CVF winter conference on applications of computer vision (pp. 3666-3673). (2020)
[5]Gottheil, P., Lippoldt, J., Grosser, S., Renner, F., Saibah, M., Tschodu, D., Poßögel, A., Wegscheider, A., Ulm, B., Friedrichs, K., Lindner, C., Engel, C., Löffler, M., Wolf, B., Höckel, M., Aktas, B., Kubitschke, H., Niendorf, A., Käs, J.: State of cell unjamming correlates with distant metastasis in cancer patients, Physical Review X, 13(3), 031003. (2023)
[6]Skrip, L. M., Moosburner, S., Tang, P., Guo, J., Görner, S., Tzschätzsch, H., Brüggemann, K., Walter, K. A., Hosse, C., Fehrenbach, U., Arnold, A., Modest, D., Krenzien, F., Schöning, W., Malinka, T., Pratschke, J., Papke, B., Käs, J. A., Sack, I., Sauer, I. M. & Hillebrandt, K.: Viscoelastic properties of colorectal liver metastases reflect tumour cell viability., Journal of Translational Medicine, 22(1), 1-14. (2024)
University of Leipzig  |  Faculty of Physics and Earth Sciences  |  Peter Debye Institute  |  Soft Matter Physics Division
© Soft Matter Physics Division, University of Leipzig. Designed and created by sp design. Imprint & Disclaimer