Project Code: PN-III-P2-2.1-PED-2016-125
Contract Number: 41PED/03.01.2017
Project Title: An Experimental Machine Intelligence Framework for the Automated Differentiation of Healthy, Dysplastic and Malignant Tissues Based on Multiphoton Microscopy Datasets
Duration: 18 months (03/01/2017–02/07/2018)
Grant value: 599 200 RON (~133 155 Euro)
Abstract:
Cancer is a group of diseases characterized by the uncontrolled growth and spread of abnormal cells, and represents a leading cause of global mortality. The three main steps that are of utmost importance for the accurate and timely diagnosis of cancers are: ‘detection’, ‘characterization’ and ‘confirmation’, histopathology playing a key role with respect to the latter two. Over the past decade significant efforts have been placed on transferring multiphoton microscopy (MPM) to the realm of histopathology because it is non-invasive and holds great potential for replacing traditional approaches based on excisional biopsies and tissue staining which have a series of disadvantages such as long diagnosis time, invasiveness, sampling error or high-costs. The ‘non-invasive’ character of MPM derives from the fact that MPM techniques exploit endogenous optical signals generated by the tissues upon interaction with a laser beam to provide optical sections (virtual biopsies) that reflect the tissular architecture and biochemical composition at controlled depths. MPM can thus provide label-free information of similar pathologic relevance to the information collected for characterization/confirmation purposes with traditional histopathology approaches. To date, MPM data sets have been successfully interpreted by human experts in a large number of experiments, however it cannot be neglected that such manual analysis approaches are both time consuming and still prone to errors due to inter- and intra-observer discrepancies. MICAND will develop a set of interconnected methods that make use of machine intelligence for reaching a fast and precise classification of MPM data sets collected on human epihelial tissues in a fully automated manner. MICAND will develop a digital pathology solution consisting in an experimental software platform, based on the Bag-of-Features paradigm, for the automated differentiation of healthy, dysplastic and malignant epithelial tissues using MPM datasets
Objectives:
The main objective of MICAND is to develop a digital pathology solution based on the Bag-of-Features (BoF) paradigm, consisting in an experimental software platform for the automated differentiation of healthy, dysplastic and malignant epithelial tissues, based on multiphoton laser scanning microscopy data sets.
Expected Results:
MICAND will develop an experimental Bag-of-Features software platform for the automated classification of healthy, dysplastic (pre-cancerous) and malignant tissues based on Multiphoton Laser Scanning Microscopy (MPM) data, which can be collected with both ex-vivo and in-vivo modalities. The resulted digital pathology framework will represent a reliable and easy to use solution for automatically classifying complex MPM data sets in the purpose of non-invasively diagnosing a wide range of epithelial oncologic pathologies with high sensitivity and specificity. Besides this main expected result, the project team will contribute to an increase in the visibility of the two partner institutions by means of publications in high-impact factor journals and presentations at prestigious international conferences.
Funding Agency :
Executive Unit for Higher Education, Research, Development and Innovation Funding (UEFISCDI)