Autor: |
Cuypers E; Maastricht MultiModal Molecular Imaging Institute (M4i), Division of Imaging Mass Spectrometry, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands., Claes BSR; Maastricht MultiModal Molecular Imaging Institute (M4i), Division of Imaging Mass Spectrometry, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands., Biemans R; The M-Lab, Department of Precision Medicine, GROW─School for Oncology, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands., Lieuwes NG; The M-Lab, Department of Precision Medicine, GROW─School for Oncology, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands., Glunde K; Russell H. Morgan Department of Radiology and Radiological Science, Division of Cancer Imaging Research, The Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, United States.; The Sidney Kimmel Comprehensive Cancer Center, The Johns Hopkins University School of Medicine, Baltimore, Maryland 21205, United States., Dubois L; The M-Lab, Department of Precision Medicine, GROW─School for Oncology, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands., Heeren RMA; Maastricht MultiModal Molecular Imaging Institute (M4i), Division of Imaging Mass Spectrometry, University of Maastricht, Universiteitssingel 50, 6229 ER Maastricht, The Netherlands. |
Abstrakt: |
The molecular pathology of breast cancer is challenging due to the complex heterogeneity of cellular subtypes. The ability to directly identify and visualize cell subtype distribution at the single-cell level within a tissue section enables precise and rapid diagnosis and prognosis. Here, we applied mass spectrometry imaging (MSI) to acquire and visualize the molecular profiles at the single-cell and subcellular levels of 14 different breast cancer cell lines. We built a molecular library of genetically well-characterized cell lines. Multistep processing, including deep learning, resulted in a breast cancer subtype, the cancer's hormone status, and a genotypic recognition model based on metabolic phenotypes with cross-validation rates of up to 97%. Moreover, we applied our single-cell-based recognition models to complex tissue samples, identifying cell subtypes in tissue context within seconds during measurement. These data demonstrate "on the spot" digital pathology at the single-cell level using MSI, and they provide a framework for fast and accurate high spatial resolution diagnostics and prognostics. |