Getra (Italy)

Abstract This paper presents the development of a machine vision (MV) system powered by deep learning (DL) to automate quality assurance in high-mix, low-volume manufacturing. In such environments, where multiple part numbers with varied configurations are produced within a single cell, the risk of shipping assemblies with missing or incorrect components is high. MV systems, which are non-contact and integrate easily into existing workflows, offer an ideal solution. However, their success depends heavily on optimized imaging conditions, careful camera selection, and model tuning. To address this, extensive experiments were conducted to determine the ideal imaging setup, including camera type, lighting arrangement, and background configuration, ensuring consistent, high-quality image capture suitable for training robust DL models. The captured images were used to evaluate and fine-tune multiple DL architectures, such as ResNet, EfficientNet, and GoogLeNet, using transfer learning techniques. These models were trained to identify component mismatches and verify assembly accuracy against the bill of materials (BOM). The best-performing models achieved over 96% classification accuracy across 25 product types, enabling detection of defects before packaging. Designed for seamless integration into existing assembly lines, the system supports real-time inspection and feedback, facilitating immediate corrective action. Future work includes deploying the system directly onto production lines, optimizing, and extending it to other product categories. This research demonstrates the effectiveness of AI-driven inspection in improving quality control, minimizing human error, and supporting scalable, efficient manufacturing operations.

This paper investigates the impact of non-performing loan data on the perceived riskiness of banks, providing information to manage them. We used the quantitative methodology based on statistical analysis to represent the role of non-performing loans. We built our sample starting from European listed banks. We developed a regression analysis establishing the relationship between non-performing data and risk. Our results show achievements in the impact of the non-performing loans (NPLs) ratio on perceived risk, particularly on the beta factor. The positioning of non-performing loan data is directed to influence capital asset pricing model (CAPM) beta. In the European banks of the analysis, a higher value of NPL ratio brings to a higher perception of risk by the market, and this is central for the future management of NPLs.

The assessment of breast density is one of the main goals of radiologists because the masking effect of dense fibroglandular tissue may affect the mammographic identification of lesions. The BI-RADS 5th Edition has revised the mammographic breast density categories, focusing on a qualitative evaluation rather than a quantitative one. Our purpose is to compare the concordance of the automatic classification of breast density with the visual assessment according to the latest available classification. A sample of 1075 digital breast tomosynthesis images from women aged between 40 and 86 years (58 ± 7.1) was retrospectively analyzed by three independent readers according to the BI-RADS 5th Edition. Automated breast density assessment was performed on digital breast tomosynthesis images with the Quantra software version 2.2.3. Interobserver agreement was assessed with kappa statistics. The distributions of breast density categories were compared and correlated with age. The agreement on breast density categories was substantial to almost perfect between radiologists (κ = 0.63-0.83), moderate to substantial between radiologists and the Quantra software (κ = 0.44-0.78), and the consensus of radiologists and the Quantra software (κ = 0.60-0.77). Comparing the assessment for dense and non-dense breasts, the agreement was almost perfect in the screening age range without a statistically significant difference between concordant and discordant cases when compared by age. The categorization proposed by the Quantra software has shown a good agreement with the radiological evaluations, even though it did not completely reflect the visual assessment. Thus, clinical decisions regarding supplemental screening should be based on the radiologist's perceived masking effect rather than the data produced exclusively by the Quantra software.

The endothelium plays a fundamental role in the biological processes that ensure physiological vessel integrity, synthesizing numerous substances that are capable of modulating the tone of vessels, inflammation and the immune system, and platelet function. Endothelial dysfunction refers to an anomaly that develops at the level of the tunica that lines the internal surface of arterial and venous vessels, or, more precisely, an alteration to normal endothelial function, which involves the loss of some structural and/or functional characteristics. Studies on sex differences in endothelial function are conflicting, with some showing an earlier decline in endothelial function in men compared to women, while others show a similar age of onset between the sexes. Since increased cardiovascular risk coincides with menopause, female hormones, particularly estrogen, are generally believed to be cardioprotective. Furthermore, it is often proposed that androgens are harmful. In truth, these relationships are more complex than one might think and are not just dependent on fluctuations in circulating hormones. An increase in serum uric acid is widely regarded as a possible risk factor for cardiovascular disease; however, its role in the occurrence of endothelial dysfunction has not yet been elucidated. Several studies in the literature have evaluated sex-related differences in the association between elevated uric acid levels and cardiovascular events, with conflicting results. The association between uric acid and cardiovascular disease is still controversial, and it is not yet clear how gender differences affect the serum concentration of these substances. This review was primarily aimed at clarifying the effects of uric acid at the level of the vascular endothelium and describing how it could theoretically cause damage to endothelial integrity. The second aim was to determine if there are gender differences in uric acid metabolism and how these differences interact with the vascular endothelium.

<div>The stringent emission regulations for Internal Combustion Engines (ICEs) spawned a great amount of research in the field of innovative combustion approaches characterized by high efficiency and low emissions. Previous research demonstrate that such promising techniques, named Low-Temperature Combustion (LTC), combine the benefits of Compression Ignition (CI) engines, such as high compression ratio and unthrottled lean mixture, with low engine-out emissions using a properly premixed air-fuel mixture. Due to longer ignition delay and high volatility compared to diesel, gasoline-like fuels show good potential for the generation of a highly premixed charge, which is needed to reach LTC characteristics. In this scenario, gasoline Partially Premixed Combustion (PPC), characterized by the high-pressure direct injection of gasoline, showed good potential for the simultaneous reduction of pollutants and emissions in CI engines. However, previous research on gasoline CI highlight that a key factor for the optimization of both efficiency and pollutants is the proper management of Exhaust Gas Recirculation (EGR). This work presents the experimental investigation performed running a light-duty CI engine, operated with gasoline PPC, and varying the mass of recirculated gases trapped in the combustion chamber. To guarantee the stability of gasoline autoignition in all the tested conditions, a specific experimental layout has been developed to accurately quantify the amount of trapped residual gases due to the internal and external EGR. The obtained results clearly highlight the impact of EGR on the combustion process and emissions, demonstrating that optimization of charge dilution with EGR is fundamental to guarantee the optimal compromise between efficiency and emissions over the whole operating range.</div>

PurposeTo investigate the imaging features of emerging COVID-19 pneumonia on chest ultrasound, radiographs and computed tomography examinations performed at admission. In addition, we provide a review of the literature and compare our results with recent evidence regarding the imaging characteristics of this novel disease.Material and methodsFrom March 17, 2020 to April 25, 2020, 23 patients with real-time polymerase chain reaction (RT-PCR) assay confirmed COVID-19 were identified. All 23 patients were evaluated and admitted at San Giuseppe Moscati Hospital in Aversa, Italy. Multi-modality imaging findings were evaluated and compared. Literature research was conducted through a methodical search on PubMed.ResultsTwenty-three patients were included in the study. Chest transthoracic ultrasound (US), chest X-ray (CXR), and computed tomography (CT) were performed respectively in 11, 16 and 21 patients. Chest US findings were consistent with diffuse B lines (91%), subpleural consolidations (45%), and thickened pleural line (18%). CXR showed prevalent manifestations of consolidations (50%) and hazy increased opacities (37%). Typical CT features are bilateral and multilobar ground-glass opacities (GGO). Indeed GGO were present in 100% of our patients. Consolidations were visible in 76% of our study population. Notably both GGO and consolidations had a peripheral distribution in all our patients. Other CT imaging features included crazy-paving pattern, fibrous stripes, subpleural lines, architectural distortion, air bronchogram sign, vascular thickening and nodules. Our literature review identified thirty original studies supporting our imaging chest findings.ConclusionsAt admission, COVID-19 pneumonia can manifest in chest imaging as B-lines and consolidations on US, hazy opacities and consolidations on CXR, multiple GGO and consolidations on CT scan.

The American Society for Gastrointestinal Endoscopy (ASGE) has produced numerous algorithms for the management of gastrointestinal bleeding (GIB) in which endoscopy plays a major role. The aim of this Systematic Review was to evaluate the diagnostic accuracy of computerized tomography (CT) angiography in detecting non-variceal gastrointestinal bleeding (NVGIB). Studies were identified on PubMed, Web of Science and Scopus databases from 2009 to 2019. The search performed included the following terms: "digestive" and "bleeding" as well as "acute bleeding gastrointestinal" or "acute bleeding digestive". Studies were included if enrolled adult patients and included measures of diagnostic accuracy (sensibility and specificity) of CT angiography in detecting GIB. Studies on variceal bleeding were excluded. First, a bivariate diagnostic randomeffects meta-analysis was fitted through REML (Restricted maximum likelihood approach) method, with the estimation of pooled sensitivity, specificity and SROC curve. Then, an univariate model was fitted for the 11 studies, with the estimation of pooled Diagnostic Odds Ratio and Cochrane's Q for heterogeneity. Eleven studies were eligible for inclusion criteria and so included in the analysis. The pooled sensitivity is 85% [75%, 92%], the pooled specificity 93% [89%, 96%]. The pooled DOR is 94.35 [37.91, 234.82]. Q for heterogeneity is not significant (p=0.377). CT angiography showed a good sensibility and specificity in detecting NVGIB. Therefore, it would be useful to consider CT angiography use also in the suspicion of NVGIB, especially when endoscopy is not immediately available and there are signs and symptoms of bleeding in progress.

Background and study aims Withdrawal time (WT) monitoring and full-spectrum endoscopy (FUSE) have been suggested to increase adenoma detection rate (ADR) due to more accurate evaluation of the hidden areas of the colon. We aimed to evaluate the efficacy of WT monitoring and FUSE on ADR.Patients and methods This was a prospective observational study involving consecutive outpatients, aged 18 to 85 years, undergoing colonoscopy with unselected indications. In phase 1, endoscopists performed 660 colonoscopies either with standard forward-viewing endoscope (SFVE) (n = 330) or with FUSE (n = 330). In this phase, WTs were measured without endoscopist awareness of being monitored. In phase 2, endoscopists were informed of being monitored and performed additional 660 colonoscopies either with SFVE (n = 330) or with FUSE (n = 330).Results WT was lower in phase 1 compared to phase 2 (SFVE: 269 ± 83 vs. 386 ± 60 sec,P < 0.001; FUSE: 289 ± 97 vs. 403 ± 65 sec,P < 0.001). Use of FUSE increased ADR both in phase 1 (33.0 % vs. 27.3 %,P = 0.127) and in phase 2 (41.8 % vs. 33.6 %,P = 0.037). When endoscopists were aware of being monitored, ADR was higher in SFVE (33.6 % vs. 27.3 %;P = 0.090) and FUSE arms (41.8 % vs. 33.0 %;P = 0.024). Improvement in detection of proximal adenomas was associated with WT monitoring [OR 1.577 (95 % C. I. 1.158 – 2.148);P = 0.004], whereas detection of distal adenomas was associated with use of FUSE [OR 1.320 (95 % C. I. 1.022 – 1.705);P = 0.037].Conclusions Unmonitored endoscopists have suboptimal WT, which increases when they are monitored. WT monitoring and use of FUSE are two reliable and alternative strategies to increase ADR.

BackgroundAn unexpected increased HCC recurrence and occurrence rate among HCV patients treated with direct acting antivirals combination has been reported. Aim of the study was the evaluation of early HCC occurrence rate and its risk factors in a HCV infected population, treated with direct-acting-antivirals.MethodsAccording to the Italian ministerial guidelines for direct-acting-antivirals treatment, 1022 consecutive HCV patients treated with direct-acting-antivirals were enrolled. Patients either with active HCC at imaging or history of previous treated HCC, HBV or HIV co-infection, or liver transplant recipients were excluded. The SVR, defined as the persistent absence of detectable serum HCV-RNA 12 weeks after the end of treatment (SVR12), was assessed for all enrolled patients. Abdominal ultrasound was performed before starting antiviral therapy, and repeated every 6 months. HCC was diagnosed according to the international guidelines. Patients showing either nodular patterns suggestive of HCC or with uncertain dynamic vascular behaviour were excluded from a further follow-up.ResultsNine hundred and eighty-five patients completed the 48 weeks follow-up after the end of treatment. A Sofosbuvir-based regimen was administered in the 74.9% of patients, among whom, the 71.6% underwent a simultaneous Ribavirin administration. A sustained virological response at 12 weeks off treatment was documented in 966 patients (98.2%). During the post treatment follow-up HCC was detected in 35 patients, with a cumulative incidence rate of the 3.55%. At multivariate analysis, four variables resulted independently associated with HCC development, both in a cirrhosis based and a class B Child based model, respectively: cirrhosis/class B Child, therapeutic schedule including Sofosbuvir without Ribavirin, liver stiffness values, male gender and presence of diabetes. A multivariate analysis performed on Child A cirrhotic patients, showed that Sofosbuvir based therapeutic treatment without Ribavirin had a HCC occurrence 5.7 higher than Ribavirin-based schedules with or without Sofosbuvir (p < 0.0001, OR: 5.686, 95% CI 2.455–13.169).ConclusionsOur data suggest that early HCC occurrence appears more frequently related to Sofosbuvir-based therapy without Ribavirin which, indeed, seems to play a protective role on HCC onset. Therefore, a careful follow-up should be mandatory, especially in those regimens including Sofosbuvir without Ribavirin.

The aim of this study was to assess the role of lung ultrasound (LUS) in a diagnostic algorithm of respiratory diseases, and to establish the accuracy of LUS compared with chest radiography (CXR). Over a period of 2 years, 509 consecutive patients admitted for respiratory-related symptoms to both emergency and general medicine wards were enrolled and evaluated using LUS and CXR. LUS was conducted by expert operators who were blinded to the medical history and laboratory data. Computed tomography (CT) of the chest was performed in case of discordance between the CXR and LUS, suspected lung cancer and an inconclusive diagnosis. Diagnosis made by CT was considered the gold standard. The difference in sensitivity and specificity between LUS and CXR as demonstrated by ROC curve analyses (LUS-AUROC: 0.853; specificity: 81.6%; sensitivity: 93.9% vs CXR-AUROC: 0.763; specificity: 57.4%; sensitivity: 96.3%) was significant (P = 0.001). Final diagnosis included 240 cases (47.2%) of pneumonia, 44 patients with cancer (8.6%), 20 patients with chronic obstructive pulmonary disease (COPD, 3.9%), 24 patients with heart failure (4.7%) and others (6.1%). In 108 patients (21.2%) with any lung pathology, a CT scan was performed with a positive diagnosis in 96 cases (88.9%); we found that CXR and LUS detected no abnormality in 24 (25%) and 5 (5.2%) cases, respectively. LUS was concordant with the final diagnosis (P < 0.0001), and in healthy patients, there was a low percentage of false positives (5.9%). The results support the routine use of LUS in the clinical context.