Available Theses

Description:
The aim of this thesis is to develop and test a multimodal hearing testing platform to assess and characterize human individual listening responses and listening effort. This study will recruit among a general population of adults (age>18) with and without hearing impairment and will use a multimodal testing procedure able to characterize participants with varying auditory profiles. The range of measures used includes self-assessment surveys, audiometric testing, new speech in noise tests, and physiological measures (using unobtrusive sensors) in various experimental setups. Advanced signal processing techniques and data analysis methods, including AI, will be used to analyze the recorded data, supporting the development of new, multidimensional models of hearing impairment and listening effort.

Hide Description

Description:

Hide Description

Description:
Pancreatic cancer is a very aggressive disease, and surgery is often the only treatment available. Unfortunately, surgery for pancreatic cancer is extremely complex, given the location of the organ, its consistency, and the skills required. Even worse, surgery can be performed in a minority of cases when the tumor has not yet infiltrated the vessels. Therefore, the estimation of vessel infiltration is crucial from a clinical point of view. The goal of this thesis is to design a graphical user interface (GUI) for a web application to support clinicians in the decision-making process, namely to assess the infiltration of pancreatic tumor in the surrounding vessels. The GUI will leverage Cornerstone3D, a JavaScript framework to build interactive medical imaging web applications. The application will integrate state-of-the-art deep learning models for interactive medical image segmentation, and data from a large public dataset of computed tomography scans.

Hide Description

Description:
This thesis explores the use of autonomous unmanned aerial vehicles (UAVs) as passive sensing platforms for multi-user identification and localization in 5G cellular networks. Building on recent advances in UAV-based signal intelligence, the work aims to enhance the accuracy, robustness, and efficiency of user localization by combining advanced signal processing techniques with adaptive UAV mission planning. The research will investigate methods for extracting spatial information from uplink reference signals, addressing challenges such as multi-user interference, multipath propagation, and limited sensing time. In addition, the thesis will study intelligent trajectory optimization strategies that allow the UAV to dynamically adapt its flight path based on real-time signal measurements, balancing localization performance and energy consumption. The expected outcome is a scalable and infrastructure-independent framework for reliable user localization, with strong potential impact in emergency response, disaster recovery, and next-generation wireless network monitoring.

Hide Description

Description:
Pancreatic cancer reported the lowest 5-year survival rate among all cancer types in the United States in the surveilled period 2014-2020. A clinical issue is represented by small pancreatic cancers (<2 cm), which may sometimes escape the attention of experts looking at computed tomography (CT) scans. This thesis focuses on the use of artificial intelligence (AI) to support clinicians in the diagnosis of pancreatic cancer. More specifically, the goal is to develop an AI agent that can automatically detect small pancreatic cancers on CT. This work continues the previous work conducted at DEIB and published on arXiv as a pre-print (Moglia A, et al., arXiv:2412.15925). The thesis provides for the possibility of completing an internship at NVIDIA.

Hide Description

Description:
The rapid advancement of generalist models in medical image segmentation has increasingly challenged specialist models, designed for a single task and trained on one or few datasets. This thesis focused on the development of an AI agent capable of systematically comparing generalist and specialist models across different organs and datasets. The objective is to address the needs of computer scientists, engineers, and medical doctors seeking to know which approach yields superior performance, to support them in the selection of the most suitable one. This work builds upon previous research conducted at DEIB and published in Information Fusion (Moglia A. et al., Inf. Fus., 2025, https://doi.org/10.1016/j.inffus.2025.103709). The thesis provides for the possibility of completing an internship at NVIDIA.

Hide Description