Project description
Cellular metabolism is essential for maintaining normal tissue function. Abnormal metabolism, in many cases, is a hallmark of several pathologies, such as cancer and neurodegenerative or immunological disorders. Metabolic imaging is a cutting-edge tool in precision medicine, as it offers a unique perspective on the biochemical and functional aspects that characterize both health and disease. Non-invasive imaging, combined with high sensitivity and resolution, improves our understanding of the disease state without generating phototoxicity.
Our group has developed METAPHOR (Metabolic Evaluation through Phasor-based Hyperspectral Imaging and Object Recognition for Mammalian Blastocysts and Oocytes). This innovative imaging method combines hyperspectral illumination with artificial intelligence to extract metabolic footprints and assess mitochondrial distribution within biological samples, using exclusively autofluorescence. Our success in developing an imaging device and associated classification software to assess the quality of embryos and oocytes has led us to expand the application of our technology to address new needs in research and clinical practice.
The METAPHOR system is based on hyperspectral (HS) imaging, a technique that allows capturing the full autofluorescence spectrum in each pixel of an image, thus encoding a large amount of metabolic information from the living sample. In our system, multidimensional HS images are transformed into a normalized spectral histogram and a phasor plot. Both representations allow for easy handling of complex multidimensional data due to their dimensionality reduction. These representations are then fed into a supervised machine learning classifier, trained to discriminate cells in different metabolic states. METAPHOR is able to identify more than six metabolites in vivo that exhibit natural autofluorescence, including those essential in key metabolic pathways (such as glycolysis and oxidative phosphorylation) and markers of oxidative stress.
The main objective of the project is to develop a reliable software based on artificial intelligence for obtaining non-invasive metabolic images in multiple cell types, with a special focus on immune cells such as PBMCs (Peripheral Blood Mononuclear Cells) and all cell subsets present in menstrual effluent for the diagnosis of endometriosis. Here we propose to take advantage of our experience in the use of HS images of living tissues to develop a diagnostic tool for non-invasive early detection of endometriosis, based on the analysis of cell populations in menstrual effluent.
Menstrual effluent contains a representative sample of various endometrial cell populations, which are involved in homeostatic and reproductive functions, such as endometrial mesenchymal stem cells (enMSCs) and uterine natural killer (uNK) cells. enMSCs have the capacity for self-renewal and play a physiological role in endometrial remodeling, although they can also be found in endometrial lesions. On the other hand, uNK cells are key regulators in the decidualization process and during pregnancy; however, in endometriosis their number is significantly reduced.
Since endometriosis is an inflammatory disorder characterized by exacerbated levels of reactive oxygen species (ROS), which affects cell types with homeostatic functions, altered metabolism could be a biomarker of interest. We propose to use this methodology to characterize the endometrial metabolic context through cells isolated from menstrual effluent, developing a personalized AI algorithm capable of distinguishing between samples from healthy and endometriosis patients. Furthermore, we plan to apply this technology to the study of PBMCs, where redox balance is fundamental for activation and metabolic phenotype. This characterization would be innovative and could lead to a new drug screening platform for T cells, macrophages, among others.
