Institut du Thorax

The Interest Public Group ARRONAX's C70XP cyclotron, used for radioisotope production for medical and research applications, relies on complex and costly systems that are prone to failures, leading to operational disruptions. In this context, this study aims to develop a machine learning-based method for early anomaly detection, from sensor measurements over a temporal window, to enhance system performance. One of the most widely recognized methods for anomaly detection is Isolation Forest (IF), known for its effectiveness and scalability. However, its reliance on axis-parallel splits limits its ability to detect subtle anomalies, especially those occurring near the mean of normal data. This study proposes a hybrid approach that combines a fully connected Autoencoder (AE) with IF to enhance the detection of subtle anomalies. In particular, the Mean Cubic Error (MCE) of the sensor data reconstructed by the AE is used as input to the IF model. Validated on proton beam intensity time series data, the proposed method demonstrates a clear improvement in detection performance, as confirmed by the experimental results.

Arterial dilation can lead to aneurysm formation, most commonly affecting the aorta and intracranial arteries. Platelets are now recognized as key mediators of vascular inflammation and remodeling, contributing to the initiation and progression of several vascular diseases. While the role of platelets in abdominal aortic aneurysm formation has gained considerable attention and is under active investigation, their contribution to intracranial aneurysm pathophysiology remains poorly understood. In this review, we summarize current mechanistic and preclinical evidence on the development and rupture of intracranial aneurysms. We discuss how platelet interactions with leukocytes, vascular cells, and their procoagulant role may influence inflammatory and tissue damage within the intracranial aneurysm, highlighting gaps in knowledge that could reveal new mechanisms.

There is a significant need for trials that evaluate the treatment of University of Texas (UT) grade 2 and 3 diabetic foot ulcers (bone, joint, or tendon exposed wounds). We undertook a trial looking at the effect of intact fish skin graft (IFSG) on these deep and difficult-to-heal ulcers. 262 patients Intent to Treat (ITT) patients with UT grade 2 and 3 DFUs were randomised to receive intact fish skin graft (IFSG) or a standardised treatment (SOC) that adhered to the International Working Group on the Diabetic Foot (IWGDF) guidelines. The secondary endpoints that were measured included wound area reduction (WAR), healing rates at 20 and 24 weeks; closure rates by UT grade, perfusion, quality of life, pain reduction and IFSG safety. We report ITT (all randomised) (mITT previosly reported) The (WAR) at 12 weeks was 65.53% for IFSG versus 30.82% for SOC (p = 0.007). UT 2 wounds (60% of total) exhibited a closure rate of 47% versus 23% at 16 weeks for IFSG versus SOC (p = 0.0033). Target wound infections were comparable (39 vs. 37) and major outcomes were comparable during the 24 week period (target-limb amputations 8% vs. 7%). Time-to-heal favoured IFSG (restricted mean to 24 weeks 17.31 vs. 19.37 weeks; KM/log-rank significant; Cox HR 1.59). The in the treatment of deep complex diabetic foot wounds the addition of IFSG significantly improved the number of patients with total wound closure as well as the time to wound closure without increased risk of complications. This improvement in total wound closure and time to wound closure was noted across prior amputation status, quality of perfusion, and UT grade.

Intracranial aneurysms (ICA) commonly occur in specific segments of the Circle of Willis (CoW), primarily, onto thirteen major arterial bifurcations. An accurate detection of these critical landmarks is necessary for a prompt and efficient diagnosis. We introduce a fully automated landmark detection approach for CoW bifurcations using a two-step neural networks process. Initially, an object detection network identifies regions of interest (ROIs) proximal to the landmark locations. Subsequently, a modified U-Net with deep supervision is exploited to accurately locate the bifurcations. This two-step method reduces various problems, such as the missed detections caused by two landmarks being close to each other and having similar visual characteristics, especially when processing the complete MRA Time-of-Flight (TOF). Additionally, it accounts for the anatomical variability of the CoW, which affects the number of detectable landmarks per scan. We assessed the effectiveness of our approach using two cerebral MRA datasets: our In-House dataset which had varying numbers of landmarks, and a public dataset with standardized landmark configuration. Our experimental results demonstrate that our method achieves the highest level of performance on a bifurcation detection task.

BackgroundTo develop a 10-year risk score to predict lower-limb amputation (LLA) in individuals with type 1 diabetes, and to assess whether this score also predicts kidney failure, and cardiovascular disease (CVD).MethodsThe LLA risk score was derived from GENEDIAB and GENESIS, two prospective French and Belgian cohorts of 828 individuals with type 1 diabetes. External validation was assessed in SURGENE, an independent cohort of 247 individuals with type 1 diabetes. LLA was defined as a non-traumatic amputation above the metatarsophalangeal joint, kidney failure as the need for renal replacement therapy, transplantation, or an estimated glomerular filtration rate (eGFR) < 15 ml/min/1.73 m2, CVD as the occurrence of myocardial infarction, heart failure, or stroke, and major adverse cardiac event (MACE) as a composite of myocardial infarction, stroke, heart failure, or all-cause death.ResultsDuring 12 years of follow-up, LLA, kidney failure, CVD and MACE occurred in 71 (9%), 84 (11%), 138 (17%) and 265 (32.0%) of participants, respectively. Sex, diabetes duration, prior LLA, eGFR, and albuminuria were identified as independent determinants of incident LLA and were used to construct the risk score. Compared to participants in the lowest score, those in the highest score had a significantly increased risks of LLA (multivariable-adjusted HR 6.02 [95% CI, 1.92–18.89]), kidney failure (8.97 [3.95–20.37]), CVD (2.50 [1.34–4.64]) and MACE (1.91 [1.24–2.96]). The score demonstrated good discriminative performance for LLA (C-index 0.82 [0.77–0.87]), kidney failure (0.86 [0.82–0.91]), CVD (0.73 [0.68–0.78]), and MACE (0.71 [0.68–0.75]). Similar predictive performance was observed in the validation cohort.ConclusionThis newly developed 10-year LLA risk score achieved excellent ability to identify individuals with type 1 diabetes at high risk for LLA, kidney failure, CVD, and MACE.Graphical abstractSupplementary InformationThe online version contains supplementary material available at 10.1186/s12933-025-03056-1.

Many patients with severe aortic stenosis present with concomitant multivalvular heart disease. The management of this condition remains challenging and requires a multidisciplinary approach that integrates clinical, hemodynamic, and multimodality imaging data to define the most effective and durable treatment strategy. Although randomized evidence to guide treatment decisions in this setting is limited, percutaneous interventions to address additional valvular lesions are being increasingly adopted. This review provides an overview of the pathophysiology of valvular diseases commonly associated with severe aortic stenosis, highlighting their prognostic implications after surgical or transcatheter treatment and their impact on risk stratification and therapeutic management.

Abstract The global rise in obesity poses a major public health challenge. While chronic energy surplus is a well-established driver of weight gain and obesity, the mechanisms linking adipose tissue (AT) expansion to cardiometabolic complications remain incompletely understood. In obese individuals, dysfunctional AT loses its capacity to store excess lipids, leading to ectopic fat accumulation and contributing to cardiometabolic complications such as type 2 diabetes. However, the molecular events that drive the transition from healthy to dysfunctional adipocytes are poorly defined. At the opposite end of the adiposity spectrum, lipodystrophies represent a heterogeneous group of disorders characterized by selective loss of AT, often accompanied by severe metabolic disturbances. Despite these contrasting adipose phenotypes, both obesity and lipodystrophy result in similar metabolic complications. In this study, we investigated whether AT in these two contrasting conditions shares a common molecular signature. We performed an unbiased comparative transcriptomic analysis of AT from lipodystrophic BSCL2-deficient and obese mice, identifying a shared signature of 129 genes. Using publicly available datasets, we replicated this signature and refined it to 102 genes whose expression is consistently altered in both obese and lipodystrophic adipose tissue. Correlation network analysis, gene ontology, and literature-based refinement revealed that these genes fall into nine functional categories: lipogenesis, adipocyte differentiation, carbohydrate metabolism, mitochondrial function, amino acid metabolism, reactive oxygen species, metabolic processes, immune response, and a group with no clear functional association. Most of these genes’ expression levels correlated strongly with insulin sensitivity across lipodystrophic and obese mice, as well as human samples. Finally, 52 genetic loci containing these genes harbor variants associated with type 2 diabetes, including 11 loci where genetic associations directly influence candidate gene expression levels. In conclusion, our findings demonstrate that a shared “energetic collapse” of adipocytes, characterized by profound metabolic inflexibility in pathways spanning glucose utilization, lipogenesis, and amino acid catabolism, represents a common pathogenic mechanism underlying adipose tissue dysfunction in both obesity and lipodystrophy. This convergent molecular signature underscores the critical role of intrinsic adipocyte metabolic health in systemic energy homeostasis and insulin sensitivity.

Abstract Cyclic stretch, generated by pulsatile blood pressure, has been identified as a pivotal regulator of vascular smooth muscle cell (VSMC) behavior and arterial wall homeostasis. Research has demonstrated that RhoA activation is a fundamental component of VSMC responses to mechanical stress through its action on cytoskeletal remodeling. However, the upstream mechanosensitive regulators of RhoA activation remain insufficiently defined. In this study, we identify the guanine nucleotide exchange factor (RhoGEF) NET1/ARHGEF8 as a stretch-responsive protein. NET1 appears to be highly expressed in arteries and VSMCs. Under physiological levels of cyclic stretch, NET1 localizes to the cytosol and interacts with RhoA. In contractile cells, loss of NET1 blunted stretch-induced MYPT1 phosphorylation and impaired cell adhesion and spreading, without exerting any effect on cell proliferation. As we demonstrated, NET1’s contribution to the stretch-induced response of VSMC was due to its localization in the cytosol. The expression of a cytosolic NET1 mutant promoted contractile gene expression and increased cell contractile capacity. Collectively, these findings identify NET1 as a stretch-sensitive RhoGEF in VSMCs that contributes to their adhesion and contractile behavior.

Abstract Background Tricuspid transcatheter edge-to-edge repair (T-TEER) is an emerging therapeutic intervention for patients with significant tricuspid regurgitation (TR). However, patient selection for T-TEER remains challenging. Prior studies have identified several hemodynamic markers, including the TAPSE/systolic PAP ratio (TAPSE/sPAP), cardiac output (CO), and pulmonary artery pulsatility index (PAPi), as potentially predictive of relevant outcome. The prognostic utility of these parameters in guiding patient selection for T-TEER remains to be fully elucidated. Objective To generate hypothesis as regards the prognostic value of these three invasive hemodynamic markers obtained from right heart catheterization (RHC) in patients eligible to T-TEER in terms of clinical response at one year ("improvement" versus "worse or no change" – the main outcome of TRI-FR trial, based on clinical events (unscheduled hospitalization for heart failure, death, tricuspid surgery) as well as PGA, NYHA and KCCQ). Methods All randomized patients in TRI-FR (N = 300) were eligible and analyzed, except those who withdrew their consent (N = 5), those allocated to T-TEER who experience a technical failure (TR grade 4+ or more in post-intervention) (N = 25), and those who had missing values for at least one of three RHC markers (N = 66). In each treatment arm, a linear relationship between RHC markers and outcome was first assessed using a generalized additive model, transformations were then applied accordingly. We fitted univariate and multivariable logistic regression models. The performances of the final model were investigated through discrimination, expressed as the area under the receiver-operating curve (AUC) and calibration (graphically). Results Out of 148 patients allocated to guideline directed medical therapy (GDMT) alone, and 152 patients allocated to T-TEER on top of GDMT, 109 and 100 were analyzed, and an improvement was observed in 39 out of 106 (36.8%) and 77 (77.8%) out of 99 patients assessed, respectively. In the GDMT alone arm, PAPi (log-transformed) and CO (3-degree polynomial) moderately predicted improvement (AUC = 0.696 and calibration far from good). In the T-TEER on top of GDMT, CO (3-degree polynomial) and TAPSE/sPAP (2-degree polynomial) moderately predicted improvement (AUC = 0.730 and moderate calibration). There were no significant relationship between hemodynamic measures and change in KCCQ or in 6-minute walk distance. Conclusion Key hemodynamic measures might be taken in account for patient selection for tricuspid valve interventions. This hypothesis has to be confirmed and probably added to other key clinical, biological and echocardiographic parameters.