Project Description
Project Identification
Project name: Intelligent System for Automatic Assistance of
Cervical Cancer Diagnosis
Contract No. 7/2014
Program name within PN II: Joint Applied Research
Projects
("Parteneriate in domenii prioritare")
Funding Agency: Executive Unit for Higher Education,
Research, Development and Innovation Funding
(UEFISCDI)
Project code: PN-II-PT-PCCA-2013-4-0202
Abstract
Cervical cancer is the second most widespread type of cancer in the world among women, with an estimated number of 530,000 cases (and 270,000 deaths) recorded in 2008. In Europe, around 60,000 women are diagnosed with this disease every year and the associated mortality rate is over 50%. Romania ranks first in Europe in terms of cervical cancer mortality rates. There are more than 3,400 new cases and 2,000 deaths each year, which is 16 times more than the European average. Unlike other types of cancer which are incurable, cervical cancer can be treated if detected in an early stage.
Cervical cancer can only be detected by means of periodic medical exams due to its asymptomatic nature. Current screening tests require the involvement of a cythopathologist. Due to the extensive volume of samples that need to be tested within a classic screening test, the minimum timeframe for analyzing a single Pap smear sample is around 30 minutes, so more than 61,5% of the tests remain unclassified. Moreover, the repetitive nature of the screening activity and also the input data's natural complexity (un-even layering, high density, cell overlaying or artifacts) invariably lead to human errors in the interpretation process.
This project aims to develop an intelligent computer-aided diagnosis system for physicians when diagnosing cervical cancer. One of the objectives is to significantly improve the diagnosis efficiency by cutting down the analysis timeframe by more than 75% and by reducing by at least 10% the false positive and false negative rates. Rigorous classification and feature extraction techniques will be applied in order to achieve this main objective, so that the diagnosis will be as accurate as possible.
Promising feature extraction techniques will be applied, such as extending the usecase domain for fractals using color information, flexible cell separation methods by partitioning and describing contour curves in order to identify the structural components as well as new classifying techniques such as Deep Neural Networks or adaptive algorithms like Boosting and Random Forrests.
Thus, an intelligent automated system for Pap smears samples would have a significant impact from several points of view: i) firstly, developing an automated diagnosis system leads to an increase in the overall quality of life due to the early detection and treatment of cervical cancer within its earliest stages of development and by drastically reducing the analysis timeframe, by diminishing the human errors associated with traditional analysis techniques, reducing the costs for the population associated with extra medical exams required by the traditional diagnosis procedures and by reducing the population's anxiety related to known human error diagnosis issues, etc., ii) increasing competitivity in Romania's economy by directly transferring our research results in the industry through our actively involved partner Genetic Lab, which is directly interested in obtaining a practical automated system, iii) stimulating the growth of the private sector's expense figures by designing a system by means of both own current activities and the collaboration with UPB.
The consortium, formed by the Politehnica University of Bucharest, with highly experienced experts in Image Processing and S.C. GeneticLab S.R.L. - the first and fore-most private diagnosis laboratory in Romania, checks all the necessary requirements to form a very able team to accomplish all the project's objectives. The project time span is 24 months, with an allocated budget of 1,437,500 lei for a team of 15 researchers, 2 PostDoc students.
Using the various complementary abilities of the two partner institutions involved is a necessary requirement to obtain a good synergy which will facilitate obtaining the expected results and is, in the same time, a solid guarantee factor achieving the proposed objectives