Pedro Sousa

António Cardoso, Pedro Sousa, Tania Pereira et al.

Currently, a central challenge and bottleneck in the deployment and validation of computer-aided diagnosis (CAD) models within the field of medical imaging is data scarcity. For lung cancer, one of the most prevalent types worldwide, limited datasets can delay diagnosis and have an impact on patient outcome. Generative AI offers a promising solution for this issue, but dealing with the complex distribution of full Hounsfield Unit (HU) range lung CT scans is challenging and remains as a highly computationally demanding task. This paper introduces a novel decomposition strategy that synthesizes CT images one HU interval at a time, rather than modelling the entire HU domain at once. This framework focuses on training generative architectures on individual tissue-focused HU windows, then merges their output into a full-range scan via a learned reconstruction network that effectively reverses the HU-windowing process. We further propose multi-head and multi-decoder models to better capture textures while preserving anatomical consistency, with a multi-head VQVAE achieving the best performance for the generative task. Quantitative evaluation shows this approach significantly outperforms conventional 2D full-range baselines, achieving a 6.2% improvement in FID and superior MMD, Precision, and Recall across all HU intervals. The best performance is achieved by a multi-head VQVAE variant, demonstrating that it is possible to enhance visual fidelity and variability while also reducing model complexity and computational cost. This work establishes a new paradigm for structure-aware medical image synthesis, aligning generative modelling with clinical interpretation.

Sérgio Guerreiro, Pedro Sousa

BPMN is a specification language widely used by industry and researchers for business process modeling and execution. It defines clearly how to articulate its concepts, but do not provide mechanism to represent the semantics of the produced models. This paper addresses the problem of how to improve the expressiveness of BPMN models, proposing a definition for the semantics of a business process within a BPMN model, and improving the completeness of the models in a systematic manner, so that models can describe far more situations with few extra managed complexity. We conceive a framework based on the business transaction patterns available in the enterprise ontology body of knowledge to prescribe the foundations of semantic BPMN models. A tool has been developed to automate the framework. Then, two industrial proof-of-concepts are used to measure its coverage, both positive and negative, and to argue about our proposal's usefulness. After that, the proposal is compared with others using a systematic literature review. A full BPMN pattern is proposed encompassing the happy flow, the declinations, the rejections and the revocations, without adding any new element to the BPMN specification. A software tool has been developed, and made publicly available, to support the automatic generation of the BPMN models from the proposed patterns. Our semantified BPMN pattern allowed the identification of a large amount of implicit, and other not implemented, situations in both proof-of-concepts. It is concluded that the usage of a semantic solution, grounded on a sound pattern, allows the systematic enrichment of the BPMN models with a bounded effort. Moreover, to simplify the BPMN executable models' implementation, its elements could be classified as implicit, explicit, or not implemented yet. Finally, related work indicates that this work is demanded, but no full solutions are available.