Texture analysis based on randomized neural networks with application in disease diagnosis

• Grant number: 22/15840-3
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): February 01, 2023
• Effective date (End): January 31, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Lucas Correia Ribas
• Grantee: Ana Catarina Marques Vicentim
• Host Institution: Instituto de Biociências, Letras e Ciências Exatas (IBILCE). Universidade Estadual Paulista (UNESP). Campus de São José do Rio Preto. São José do Rio Preto , SP, Brazil
• Associated research grant: 18/22214-6 - Towards a convergence of technologies: from sensing and biosensing to information visualization and machine learning for data analysis in clinical diagnosis, AP.TEM

Abstract

Texture analysis is a classic yet still challenging task in computer vision, for which artificial neural networks are actively being applied. Most approaches take advantage of large and complex architectures of convolutional neural networks trained with huge databases and that demand high processing power. In contrast to these approaches based on complex architectures, Randomized Neural Networks (RNNs) can provide texture descriptors based on a simple representation learning paradigm that has low computational cost (thanks to training based on a closed-form solution), which uses few data and is easy to interpret. In this sense, this project aims to study, implement and develop improvements in techniques for learning texture representations using Randomized Neural Networks, in order to extract rich representations with low computational cost and using few data. In addition, the texture analysis methods studied and developed will be applied to microscopy images (such as sensory units and tissues) with the purpose of diagnosing diseases. Thus, it is expected that this research results in contributions in the computational scope with improved techniques for texture analysis and also related to the areas of applications with better performance in disease diagnoses. It should be noted that these contributions are congruent with the objectives of the thematic project (process, 2018/22214-6) of which this project is part, whose goal is the convergence of technologies for diagnosis.