Curricular Internship

Segmentation, the problem of locating and outlining objects of interest in images, is a central problem in biomedical image analysis. In brain research in particular, segmentation of MRI and other imaging modalities is crucial for studying the effects of behavior, disease, and treatment on brain anatomy and function. We use Atlas-based segmentation in order to insight in the image registration techniques applied in medical image processing.

The developed program provides a classification of fruits and vegetables. The data is obtained from Horea Muresan's "Fruit recognition from images using deep learning" and it contains 75,937 images of 120 fruits. It is realized in collaboration with Robovision, we use Keras and Google Colab to get familiar with different theoretical and practical aspects of Neural Networks in a way that the software will be able to distinguish different pictures by itself.

The purpose of this project is to obtain insight into feature extraction techniques applied on medical image processing of lungs as well as processing of the extracted features.

A qualitative knowledge about the medical diagnosis of several respiratory diseases is considered. The foreseen developed expert system is meant to infer the preliminary diagnosis and probabilities of the diseases only based on the observed symptoms and the accumulated expert knowledge (statistical data), without having access to any further tests. The system can operate online posing the relevant questions to the user and making inference in real time based on the answers. Based on the results of the Bayesian inference, further tests (including some standard tests the considered diseases) should be recommended.

This report focuses on two advantages of iterative reconstruction algorithms. We will look at the influence of resolution modeling in reconstruction making a projection that mimics a PET system with realistic resolution model. Then the influence of noise will be investigated. For both parts, the most used iterative reconstruction algorithm, will be used: Maximum Likelihood Expectation Maximization (MLEM)

Model development using Matlab and Simulink to describe the glycemic and insulinemic response to a 75 g oral glucose tolerance test (OGTT). A target point of this model is to represent separately the incretin effect of the two hormones GIP and GLP-1, considering their different influence on the insulin secretion based on physiological considerations. Estimated model parameters reflect the respective values found in literature with an acceptable coefficient of variance.

Ho tenuto lezioni di teoria ritmica e percezione musicale nell'ambito di corsi amatoriali e preaccademici