Project Code: PN-III-P2-2.1-PED-2019-1666
Contract Number: 434PED/2.11.2020
Project Title: Method for fast and precise diagnostic of gastric cancers based on non-linear optical microscopy and Deep Learning
Duration: 24 months (02/11/2020-01/11/2022)
Grant value: 600 000 RON (~122 700 Euro)
Abstract:
According to the World Health Organization, the proportion of world’s population over 60 years will approximately double by 2050. This progressive increase of the elderly population will lead to a dramatic growth of age-related diseases, including cancer, resulting in tremendous pressure on the sustainability of healthcare systems, globally. In this context, finding more efficient ways to address Gastric Cancer (GC), the third leading cause of cancer-related deaths worldwide, is of utmost importance. Prevention of gastric cancer to decrease morbidity relies on precise diagnosis of dysplastic tissues or of gastric tumours of a very early stage. In this project, we will address this by combining two emerging technologies, namely non-linear optical microscopies (NLO) and Deep Learning, to result in a novel method enabling automated fast and precise characterization of unstained gastric tissues. The applicability of this method will be three-fold: (i) it will be efficient in characterizing NLO data sets collected with bench-top systems on fixed, excised, samples (acting as a complementary tool to traditional histopathology protocols), (ii) it will be efficient in characterizing freshly excised unstained minute specimens (imaged with bench-top NLO systems or existing NLO tomographs) to provide a GC diagnostic very fast, next to the patient, avoiding wait times associated to tissue fixation, staining and expert interpretation, (iii)it will enable intra-vital in-vivo characterization approaches for GC diagnostics in association with forthcoming systems for NLO endomicroscopy/gastroscopy. Such capabilities will rely on novel ways to augment NLO trainig data and to transfer the learning of deep convolutional neural networks across distinct imaging modalities.
Objectives:
The main objective of GastroDeep is to combine two emerging technologies, Nonlinear Optical Microscopies and Deep Learning, to result in a novel method enabling automated fast and precise characterization of unstained gastric tissues. The applicability of this method, coined GastroDeep, will be three-fold: (i)it will be efficient in characterizing NLO data sets collected with bench-top systems on fixed, excised, samples (acting as a complementary tool to traditional histopathology protocols).(ii)it will be efficient in characterizing freshly excised unstained minute specimens(imaged with bench-top NLO systems or existing NLO tomographs)to provide a GC diagnostic very fast, next to the patient,avoiding wait times associated to tissue fixation, staining and expert interpretation, (iii)it will enable intra-vital in-vivo characterization approaches for GC diagnostics in association with forthcoming systems for NLO endomicroscopy/gastroscopy.
Expected Results:
-Method for the characterization and diagnostic of gastro-intestinal tissues, using nonlinear optical microscopies (NLO) and Deep Learning artificial intelligence techniques.
-Training sets for developing a new generation of Deep Learning methods for the automated analysis of NLO images.
-Mechanisms for hybrid training of convolutional neural networks using complementary optical datasets.
-Method for augmenting NLO datasets that compensate their limited availability.
-Consolidation of the existing collaboration between Politehnica University of Bucharest and the Carol Davila University of Medicine and Pharmacy, with perspectives for adapting the methdology developed in this project to address other high-interest pathologies.
Funding Agency :
Executive Unit for Higher Education, Research, Development and Innovation Funding (UEFISCDI)