MSACL 2024 Abstract

Cancer Surgery remains the first pillar of therapy in oncology. Surgery quality is of utmost importance because of the huge impact it has on patient relapse and survival. This is also true for the choice of the subsequent therapy. Moving forward personalized surgery and therapy remain thus the main goal of most of the research in oncology. However, tailoring the surgery and the therapy is tightly linked to the ability to collect solid molecular data already by the time point of the surgery offering accurate information which could be used for diagnosis and prognosis as well as surgical margin delineation. Measuring the composition of the tumor microenvironment (TME) is one way to this objective.

Among other technologies, Ambient Ionization Mass Spectrometry (AIMS) has shown to be a very sensitive and specific approach to get the molecular fingerprint of different cell types. Recent developments have made it possible to operate MS intraoperatively and in vivo. SpiderMass technology is one of the AIMS that has demonstrated its potential for in-man intraoperative analysis.

Our objective is to develop an autonomous robotized system that will be able to scan a suspicious area and propose an interpretation of the molecular information for the surgeon to be able to make the right decision and tailor the surgery. We have been working therefore on the development of a robotized system, testing both stiff and looking forward soft robot systems, as well as developing a robust processing pipeline to generate MS-based digital twins. More specifically, we show the ability of the SpiderMass to perform topographic MS imaging suited for the real environment of the surgery room. On the other hand, we are developing a machine-learning-based pipeline to predict the ratio of the different cell populations in the TME and its surroundings. Indeed, based on the MS profiles of the different cells present in the tissues analyzed individually, we can build a LightGBM model to predict the ratio of these different cell populations and sub-populations for each pixel of the image. This can be applied to predict the infiltration of the different immune cell sub-populations or highlight the presence of specific bacterial populations according to the cancer sub-type of the different areas of the tissues for diagnosis and prognosis purposes. Additionally, this approach opens an interesting way to depict the infiltration of cancer cells in the normal tissue and propose an accurate delineation of the tumor margins. Interestingly these predictions were cross-validated by immunohistochemistry (IHC) based on specific markers known for these different cells including MALDI IHC. We are now investigating generating these scores in real-time and unrolling them on the topographic maps recorded during the MS Imaging to reconstruct digital twins based on the MS Imaging scoring which can then be used by the physicians to personalize their surgery.