Classification Fusion Techniques for Automatic Diabetic Retinopathy Screening

• Grant number: 11/15349-3
• Support Opportunities: Scholarships in Brazil - Master
• Start date: March 01, 2012
• End date: February 28, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Anderson de Rezende Rocha
• Grantee: José Ramon Trindade Pires
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

In medicine, early diagnosis of diseases is paramount to improve the patients' quality of life and reduce treatment costs. For ophthalmologists, for instance, an efficient automatic screening system of diabetics' ocular-fundus images would ensure that only patients presenting some abnormality associated with Diabetic Retinopathy (DR) would be referred to an appointment. There are hundreds of abnormalities that may affect the retina of a person with diabetes, some of which have similar characteristics, while the majority has its own peculiarities. Given the large number of abnormalities related to the disease, one approach that detect only one kind of abnormality would not be enough. This research project is a natural continuation of a research project funded by MSR-FAPESP virtual institute between 2008 and 2011 conducted by our research group, intitled "Automated screening for diabetic retinopathies: IT in the fight against preventable blindness". In that project our team created the initial conditions necessary for further in-depth investigations about the problem. The research has resulted in three detectors for abnormalities related to DR. Complementing that work, this research project aims at developing detectors associated with new abnormalities, as well as developing novel classifier combination methods aiming at performing automatic DR screening tasks. We intend to pursue such objective by conducting the research in four stages: (1) verification of the quality of the ocular-fundus images to be used in the analysis; (2) improvement in image quality; (3) development of new detectors; and (4) application of classifiers fusion techniques. The validation of this work will be performed through ocular-fundus images collected weekly in the House of Diabetes, in São Paulo, one of our partners in this project. We intend to validate our approaches with images from our international partners as well.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

PIRES, RAMON; JELINEK, HERBERT F.; WAINER, JACQUES; GOLDENSTEIN, SIOME; VALLE, EDUARDO; ROCHA, ANDERSON. Assessing the Need for Referral in Automatic Diabetic Retinopathy Detection. IEEE Transactions on Biomedical Engineering, v. 60, n. 12, p. 3391-3398, DEC 2013. (11/15349-3, 10/05647-4)

PIRES, RAMON; JELINEK, HERBERT F.; WAINER, JACQUES; VALLE, EDUARDO; ROCHA, ANDERSON. Advancing Bag-of-Visual-Words Representations for Lesion Classification in Retinal Images. PLoS One, v. 9, n. 6, JUN 2 2014. (11/15349-3, 10/05647-4)