Antonio Scardace
|
“Whatever it takes.”
SCROLL DOWN
SWIPE UP
About me
In July 2025 I graduated with honours from the University of Catania with a Master’s degree in Computer Science with a final grade of 110/110 cum laude. I completed my Bachelor’s degree in March 2023, also with top marks and honours. On both occasions I was nominated for a prestigious departmental merit prize.
I have conducted research activities for over three years focusing on Theoretical Computer Science and Quantum Computing under Professor Simone Faro.
Since September 2024 I have been working on Medical Imaging research with Professor Daniele Ravì, and in October 2025 I started a PhD in Computer Science under his supervision. My research equally combines applied multimodal AI for Alzheimer’s Disease-progression modelling and theoretical studies on detecting and quantifying memorization in generative models trained on medical datasets.
In September 2020 I was selected as a mentee for LeadTheFuture, a Forbes-recognised mentorship programme for outstanding Italian students with an acceptance rate <13%. In May 2023 I also became an Ambassador for the organisation.
LeadTheFuture taught me the value of the right network for personal growth. I am eager to connect with visionary mentors and peers. With a strong devotion to Computer Science and the Public Interest, I aim to merge these fields to achieve meaningful social impact. In particular, I want to apply AI in healthcare to improve lives and make a real social impact.
Selected Publications
Auditing Patient Privacy in Medical Generative Models:
Scalable Memorization Detection with DeepSSIM++
[Submitted to] Medical Image Analysis (MedIA) • arXiv:2609.03615
In this paper, we extend DeepSSIM with DeepSSIM++, a self-supervised metric that leverages multi-scale feature aggregation, anatomy-preserving augmentations, and spatial explainability to detect and quantify memorization efficiently and robustly across large sets of generated medical images.
A Quantitative Analysis of Multimodal Biomarkers in Alzheimer's Disease
2026 IEEE Symposium on Computers and Communications (ISCC) • arXiv:2606.17867
In this paper, we systematically analyze multimodal Alzheimer’s disease biomarkers (i.e. sMRI, tau-PET, CDR, APOE4) to characterize cross-modal dependencies, identify redundant information, and uncover neurodegenerative trajectories linking tau pathology, brain atrophy, and cognitive decline.
A Novel Metric for Detecting Memorization in Generative Models for Brain MRI Synthesis
2026 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) • openaccess.thecvf.com
In this paper, we address the risk of patient data leakage in deep generative models for medical imaging, which may memorize sensitive training samples. We propose DeepSSIM, a self-supervised metric designed to detect and quantify memorization across large sets of generated images, providing a scalable solution to safeguard privacy.