Gravina, Michela (2022) Artificial Intelligence and Medical Imaging: Handling and Mining Multiple Sources. [Tesi di dottorato]
Anteprima |
Testo
Gravina_Michela_35.pdf Download (6MB) | Anteprima |
| Tipologia del documento: | Tesi di dottorato |
|---|---|
| Lingua: | English |
| Titolo: | Artificial Intelligence and Medical Imaging: Handling and Mining Multiple Sources |
| Autori: | Autore Email Gravina, Michela michela.gravina@unina.it |
| Data: | 13 Dicembre 2022 |
| Numero di pagine: | 204 |
| Istituzione: | Università degli Studi di Napoli Federico II |
| Dipartimento: | Ingegneria Elettrica e delle Tecnologie dell'Informazione |
| Dottorato: | Information and Communication Technology for Health |
| Ciclo di dottorato: | 35 |
| Coordinatore del Corso di dottorato: | nome email Riccio, Daniele daniele.riccio@unina.it |
| Tutor: | nome email Sansone, Carlo [non definito] |
| Data: | 13 Dicembre 2022 |
| Numero di pagine: | 204 |
| Parole chiave: | Artificial Intelligence, Medical Image Computing, Deep Learning, Multimodal Learning, Image Synthesis |
| Settori scientifico-disciplinari del MIUR: | Area 09 - Ingegneria industriale e dell'informazione > ING-INF/05 - Sistemi di elaborazione delle informazioni |
| Depositato il: | 25 Gen 2023 01:16 |
| Ultima modifica: | 09 Apr 2025 14:18 |
| URI: | http://www.fedoa.unina.it/id/eprint/14644 |
Abstract
Medical image computing refers to the process of extracting relevant information from medical images for the characterization of the area under analysis. The large amount of information to consider, and the high complexity of medical images, which make the manual inspection a very tedious and hard task, have prompted research into proposing solutions for the automatic analysis of radiological acquisitions. More recently, Artificial Intelligence (AI), and in particular Machine Learning (ML) and Deep Learning (DL), had a radical spread in medical image computing with surprising results. Moreover, the use of deep neural networks has also enabled the development of DL-based solutions in medical applications characterized by the need of leveraging information coming from multimodal data sources, raising the Multimodal Deep Learning (MDL). However, in healthcare, it is very difficult to obtain high-quality, balanced datasets with labels due to privacy and sharing policy issues. Several applications have leveraged DL approaches in data augmentation techniques, proposing models that are able to create new realistic and synthetic samples. As a consequence, it is possible to identify a new source of data, that is denoted as synthetic data source. The aim of this thesis is to investigate the DL approaches in medical image computing, considering the presence of multiple data sources. In the case of multimodal data sources, a systematic analysis of multimodal data fusion techniques is performed introducing an innovative transfer module that allows the different modalities to influence each other, while in the analysis of synthetic data source, a DL-based data augmentation method is proposed that exploits the biological characteristics of the images implementing a physiologically-aware synthetic image generation process.
Downloads
Downloads per month over past year
Actions (login required)
 |
Modifica documento |
