Pica Code IR-2026-099 Segmentation and multimodal analysis of 3D medical images through efficient deep learning architectures

Research activities: The appointment concerns the development of deep learning methods for the segmentation and analysis of three-dimensional medical volumes, with particular reference to Cone-Beam CT (CBCT) images of the maxillofacial region and brain magnetic resonance imaging (MRI). Activities will include: - design, training and evaluation of architectures for multi-structure 3D segmentation, including convolutional, Transformer and State Space Model (e.g., Mamba) architectures, with attention to computational efficiency and scalability on HPC/GPU infrastructures; - development of multimodal fusion strategies and methods to handle missing modalities (e.g., incomplete MRI sequences), as well as image synthesis methods;(omissis)