Predicting S-phase Progression from Cytomorphological Fingerprints: A Deep Learning Model for Blood-Based Clinical Cancer Risk Profiling – Sandro Dragone

This research aims to develop a deep learning-based predictive model for cancer risk assessment, leveraging cytological imaging generated by the application of the Charactex method to peripheral blood samples. The experimental workflow integrates non-hematological cell enrichment and short-term (14-day) culture, followed by whole-slide imaging (WSI/SVS) digitization. By integrating clinical metadata, we extract high-dimensional morphological and cytometric biomarkers to distinguish healthy cohorts from cancer patients. A comprehensive dataset, including phenotypic patterns, cytopathological variables, and proliferation profiles, was used to train a neural network. The primary objective is to establish and validate an automated predictive model capable of determining the proliferation profile solely from morphological and cytoarchitectural characteristics compared to gold-standard analytical measurements.

About Sandro Dragone