A statistical modeling approach for tumor-type identification in surgical neuropathology using tissue mass spectrometry imaging.

Current clinical practice involves classification of biopsied or resected tumor tissue based on a histopathological evaluation by a neuropathologist. In this paper, we propose a method for computer-aided histopathological evaluation using mass spectrometry imaging. Specifically, mass spectrometry imaging can be used to acquire the chemical composition of a tissue section and, hence, provides a framework to study the molecular composition of the sample while preserving the morphological features in the tissue. The proposed classification framework uses statistical modeling to identify the tumor type associated with a given sample. In addition, if the tumor type for a given tissue sample is unknown or there is a great degree of uncertainty associated with assigning the tumor type to one of the known tumor models, then the algorithm rejects the given sample without classification. Due to the modular nature of the proposed framework, new tumor models can be added without the need to retrain the algorithm on all existing tumor models.