#Dermatologic Diagnosis through Computer Vision and Pattern Recognition

This is the code, data and monograph of my final project, a requirement to complete my Bachelor’s in Information Systems at the Federal University of the State of Rio de Janeiro. The project is entitled Dermatologic Diagnosis through Computer Vision and Pattern recognition, developed under the mentorship of Sean Wolfgand Matsui Siqueira Ph.D.

The objective is to measure the efficacy of a selection of image analysis algorithms that aim to detect malignant melanomas in images of skin lesions.

The project entails the re-implementation of Staunton and Ma’s border irregularity analysis algorithm and the implementation of 3 additional algorithms using different methods to quantify the potential malignancy.

Staunton and Ma’s algorithm was originally written in Matlab; the software was recently lost and needed to be rebuilt to include their algorithm in my study. The code was written in Octave due to budget limitations, and is located in the directory StauntonMa.

The code for the 3 additional algorithms is in the src folder and is written in Java. The algorithm makes use of Yuri Pourre’s Java image manipulation library e-motion (see https://github.com/yuripourre/e-motion) and some of Tom Gibara’s computer vision code (see http://www.tomgibara.com).

The folders Melanoma and Nevi contain the images used in the study. These images were acquired from the Dermaltas library and from the records of dermatologists who contributed to my project. These images were curated by Dr. Jefferson Braga Louback and Dr. Ricardo Barbosa Lima.

Mono v5 is the 5th version of my monograph, presented for defense on June 4th, 2014 and awarded the highest grade by the evaluation committee. The evaluation committee was composed of Sean Wolfgand Matsui Siqueira Ph.D., Asterio Kiyoshi Tanaka Ph.D. and Fernanda Araujo Baião Ph.D.

Obs. The monograph is to undergo a final review before submission to the university’s records.