The paper’s main contributions are twofold: to demonstrate how to apply the general European Union’s High-Level Expert Group’s (EU HLEG) guidelines for trustworthy AI in practice for the domain of healthcare; and to investigate the research question of what does “trustworthy AI” mean at the time of the COVID-19 pandemic. To this end, we present the results of a post-hoc self-assessment to evaluate the trustworthiness of an AI system for predicting a multi-regional score conveying the degree of lung compromise in COVID-19 patients, developed and verified by an interdisciplinary team with members from academia, public hospitals, and industry in time of pandemic. The AI system aims to help radiologists to estimate and communicate the severity of damage in a patient’s lung from Chest X-rays. It has been experimentally deployed in the radiology department of the ASST Spedali Civili clinic in Brescia (Italy) since December 2020 during pandemic time. The methodology we have applied for our post-hoc assessment, called Z-Inspection, uses socio-technical scenarios to identify ethical, technical and domain-specific issues in the use of the AI system in the context of the pandemic.

Modeling and representing 3D shapes of the human body and face is a prominent field due to its applications in the healthcare, clothes, and movie industry. In our work, we tackled the problem of 3D face and body synthesis by reducing 3D meshes to 2D image representations. We show that the face can naturally be modeled on a 2D grid. At the same time, for more challenging 3D body geometries, we proposed a novel non-bijective 3D–2D conversion method representing the 3D body mesh as a plurality of rendered projections on the 2D grid. Then, we trained a state-of-the-art vector-quantized variational autoencoder (VQ-VAE-2) to learn a latent representation of 2D images and fit a PixelSNAIL autoregressive model to sample novel synthetic meshes. We evaluated our method versus a classical one based on principal component analysis (PCA) by sampling from the empirical cumulative distribution of the PCA scores. We used the empirical distributions of two commonly used metrics, specificity and diversity, to quantitatively demonstrate that the synthetic faces generated with our method are statistically closer to real faces when compared with the PCA ones. Our experiment on the 3D body geometry requires further research to match the test set statistics but shows promising results.

Background: Oral fluids may provide an alternative to traditional invasive sampling methods for screening and monitoring chronic conditions, such as Type 2 diabetes mellitus (T2DM), which is commonly assessed using blood HbA1c levels. We aimed to identify putative proteomic signatures in oral rinse samples to differentiate individuals with low versus high HbA1c levels using a machine learning (ML) approach. Methods: Participants were selected from an existing population cohort (The Maastricht Study), and oral rinse samples were derived from the associated Biobank. A two-threshold ML approach used the population median threshold (37 mmol/mol), followed by the clinical cutoff for elevated HbA1c (48 mmol/mol), the diagnostic threshold for diabetes, to stratify participants into two HbA1c-based groups. Proteomic analysis of the oral rinse samples was performed using the Proximity Extension Assay (Olink ^® ) targeting 92 cardiometabolic proteins. The ML approach employed repeated nested cross-validation with Boruta feature selection and evaluated Logistic Regression, Random Forest, and XGBoost. SHapley Additive exPlanations (SHAP) were used to quantify feature importance and interpret the contribution and directionality of individual proteins in the model predictions. Post hoc analyses evaluated the levels of identified proteins across HbA1c and T2DM categories and assessed potential confounding factors, including periodontitis, an oral inflammatory disease. Results: Among the 176 participants, no discriminatory proteins were identified at the population median threshold. Using the clinical cutoff, Logistic Regression achieved the highest performance (ROC AUC: 0.78 ± 0.09; 95% CI: [0.74, 0.82]). REG1A (Regenerating Islet-Derived Protein 1 Alpha), also known as Pancreatic Stone Protein, emerged as the sole oral biomarker for elevated blood HbA1c. In the post hoc multivariable analysis, after adjusting for potential confounders, REG1A remained significantly associated with elevated HbA1c, and REG1A levels showed significant positive trends across the clinically relevant HbA1c and T2DM categories (normal, prediabetes, and diabetes). Conclusions: REG1A, a protein linked to pancreatic β-cell regeneration, emerged as an oral biomarker for elevated HbA1c levels. This finding suggests that oral rinse proteomics may offer a simple, non-invasive tool for screening and monitoring elevated HbA1c levels, with potential applicability to point-of-care testing and large epidemiological studies. However, validation in larger cohorts is warranted.