Abstract:

Federated Learning (FL) holds the potential to advance equality in health by enabling diverse institutions to collaboratively train deep learning (DL) models, even with limited data. However, the significant resource requirements of FL often exclude centres with limited computational infrastructure, further widening existing healthcare disparities. To address this issue, we propose a Green AI-oriented adaptive layer-freezing strategy designed to reduce energy consumption and computational load while maintaining model performance. We tested our approach using different federated architectures for Magnetic Resonance Imaging (MRI)-to-Computed Tomography (CT) conversion. The proposed adaptive strategy optimises the federated training by selectively freezing the encoder weights based on the monitored relative difference of the encoder weights from round to round. A patience-based mechanism ensures that freezing only occurs when updates remain consistently minimal. The energy consumption and CO2eq emissions of the federation were tracked using the CodeCarbon library. Compared to equivalent non-frozen counterparts, our approach reduced training time, total energy consumption and CO2eq emissions by up to 23%. At the same time, the MRI-to-CT conversion performance was maintained, with only small variations in the Mean Absolute Error (MAE). Notably, for three out of the five evaluated architectures, no statistically significant differences were observed, while two architectures exhibited statistically significant improvements. Our work aligns with a research paradigm that promotes DL-based frameworks meeting clinical requirements while ensuring climatic, social, and economic sustainability. It lays the groundwork for novel FL evaluation frameworks, advancing privacy, equity and, more broadly, justice in AI-driven healthcare.

Abstract:

Cone-beam computed tomography (CBCT) has become a widely adopted modality for image-guided radiotherapy (IGRT). However, CBCT is characterized by increased noise, limited soft-tissue contrast, and artifacts. These issues result in unreliable Hounsfield unit (HU) values, which limits direct dose calculation. These issues were addressed by generating synthetic CT (sCT) from CBCT, particularly by adopting deep learning (DL) methods. However, existing DL approaches were hindered by institutional heterogeneity, scanner-dependent variations, and data privacy regulations that prevented multi-center data sharing.To overcome these challenges, we proposed a cross-silo federated learning approach for CBCT-to-sCT synthesis in the head and neck region. This approach extended our original FedSynthCT framework to a different image modality and anatomical region. A conditional generative adversarial network (cGAN) was trained using data from three European medical centers within the SynthRAD2025 public challenge dataset while maintaining data privacy at each institution. A combination of the FedAvg and FedProx aggregation strategies, alongside a standardized preprocessing pipeline, was adopted to federate the DL model.The federated model effectively generalized across participating centers, as evidenced by the mean absolute error (MAE) ranging from $64.38\pm13.63$ to $85.90\pm7.10$ HU, the structural similarity index (SSIM) ranging from $0.882\pm0.022$ to $0.922\pm0.039$, and the peak signal-to-noise ratio (PSNR) ranging from $32.86\pm0.94$ to $34.91\pm1.04$ dB. Notably, performance on an external validation dataset of 60 patients yielded comparable metrics: a MAE of $75.22\pm11.81$ HU, an SSIM of $0.904\pm0.034$ and a PSNR of $33.52\pm2.06$, confirming robust cross-center generalization despite differences in imaging protocols and scanner types, without additional training. Furthermore, a visual analysis of the results revealed that the obtained metrics were influenced by registration errors.Our findings demonstrated the technical feasibility of FL for CBCT-to-sCT synthesis task while preserving data privacy, offering a collaborative solution for developing generalizable models across institutions without requiring data sharing or center-specific models.

Abstract:

The generation of Synthetic Computed Tomography (sCT) images has become a pivotal methodology in modern clinical practice, particularly in the context of Radiotherapy (RT) treatment planning. The use of sCT enables the calculation of doses, pushing towards Magnetic Resonance Imaging (MRI) guided radiotherapy treatments. Moreover, with the introduction of MRI-Positron Emission Tomography (PET) hybrid scanners, the derivation of sCT from MRI can improve the attenuation correction of PET images.Deep learning methods for MRI-to-sCT have shown promising results, but their reliance on single-centre training dataset limits generalisation capabilities to diverse clinical settings. Moreover, creating centralised multi-centre datasets may pose privacy concerns. To address the aforementioned issues, we introduced FedSynthCT-Brain, an approach based on the Federated Learning (FL) paradigm for MRI-to-sCT in brain imaging. This is among the first applications of FL for MRI-to-sCT, employing a cross-silo horizontal FL approach that allows multiple centres to collaboratively train a U-Net-based deep learning model. We validated our method using real multicentre data from four European and American centres, simulating heterogeneous scanner types and acquisition modalities, and tested its performance on an independent dataset from a centre outside the federation.In the case of the unseen centre, the federated model achieved a median Mean Absolute Error (MAE) of 102.0 HU across 23 patients, with an interquartile range of 96.7–110.5 HU. The median (interquartile range) for the Structural Similarity Index (SSIM) and the Peak Signal to Noise Ratio (PNSR) were 0.89 (0.86–0.89) and 26.58 (25.52–27.42), respectively.The analysis of the results showed acceptable performances of the federated approach, thus highlighting the potential of FL to enhance MRI-to-sCT to improve generalisability and advancing safe and equitable clinical applications while fostering collaboration and preserving data privacy.

Abstract:

PurposeIdentifying and quantifying coronary artery calcification (CAC) is crucial for preoperative planning, as it helps to estimate both the complexity of the 2D coronary angiography (2DCA) procedure and the risk of developing intraoperative complications. Despite the relevance, the actual practice relies upon visual inspection of the 2DCA image frames by clinicians. This procedure is prone to inaccuracies due to the poor contrast and small size of the CAC; moreover, it is dependent on the physician’s experience. To address this issue, we developed a workflow to assist clinicians in identifying CAC within 2DCA using data from 44 image acquisitions across 14 patients.MethodsOur workflow consists of three stages. In the first stage, a classification backbone based on ResNet-18 is applied to guide the CAC identification by extracting relevant features from 2DCA frames. In the second stage, a U-Net decoder architecture, mirroring the encoding structure of the ResNet-18, is employed to identify the regions of interest (ROI) of the CAC. Eventually, a post-processing step refines the results to obtain the final ROI. The workflow was evaluated using a leave-out cross-validation.ResultsThe proposed method outperformed the comparative methods by achieving an F1-score for the classification step of 0.87 (0.77-0.94) (median ± quartiles), while for the CAC identification step the intersection over minimum (IoM) was 0.64 (0.46-0.86) (median ± quartiles).ConclusionThis is the first attempt to propose a clinical decision support system to assist the identification of CAC within 2DCA. The proposed workflow holds the potential to improve both the accuracy and efficiency of CAC quantification, with promising clinical applications. As future work, the concurrent use of multiple auxiliary tasks could be explored to further improve the segmentation performance.

Abstract:

BACKGROUND AND OBJECTIVE: Radiomics extracts quantitative features from magnetic resonance images (MRI) and is especially useful in the presence of subtle pathological changes within human soft tissues. This scenario, however, may not cover the effects of intrinsic, e.g., aging-related, (physiological) neurodegeneration of normal brain tissue. The aim of the work was to study the repeatability of radiomic features extracted from an apparently normal area in longitudinally acquired T1-weighted MR brain images using three different intensity normalization approaches typically used in radiomics: Z-score, WhiteStripe and Nyul. METHODS: Fifty-nine images of hearing impaired, yet cognitively intact, patients were repeatedly acquired in two different time points within six months. Ninety-one radiomic features were obtained from an area within the pons region, considered to be a healthy brain tissue according to previous analyses and reports. The Intraclass Correlation Coefficient (ICC) and the Concordance Correlation Coefficient (CCC) in the repeatability study were used as metrics. RESULTS: Features extracted from the MRI normalized with Z-score showed results comparable in both ICC (0.90 (0.82-0.98)) and CCC (0.82 (0.69-0.95)) distribution values, in terms of median and quartiles, with those extracted from the images normalized with WhiteStripe (0.89 (0.84-0.92)) and (0.80 (0.73-0.84)), respectively. CONCLUSION: Our findings underline the importance of, providing useful guidelines for, the intensity normalization of brain MRI prior to a longitudinal radiomic analysis.

Abstract:

Radiotherapy (RT) is a cancer treatment technique that involves exposing cells to ionizing radiation, including X-rays, electrons, or protons. RT offers promise to treat cancer, however, some inherent limitations can hamper its performance. Radio-resistance, whether innate or acquired, refers to the ability of tumor cells to withstand treatment, making it a key factor in RT failure. This perspective hypothesizes that nanoscale surface topography can impact on the topology of cancer cells network under radiation, and that this understanding can possibly advance the assessment of cell radio-resistance in RT applications. An experimental plan is proposed to test this hypothesis, using cancer cells exposed to various RT forms. By examining the influence of 2D surface and 3D scaffold nanoscale architecture on cancer cells, this approach diverges from traditional methodologies, such as clonogenic assays, offering a novel viewpoint that integrates fields such as tissue engineering, artificial intelligence, and nanotechnology. The hypotheses at the base of this perspective not only may advance cancer treatment but also offers insights into the broader field of structural biology. Nanotechnology and label-free Raman phenotyping of biological samples are lenses through which scientists can possibly better elucidate the structure-function relationship in biological systems.

Abstract:

BACKGROUND: Longitudinal Displacement (LD) is the relative motion of the intima-media upon adventitia of the arterial wall during the cardiac cycle, probably linked to atherosclerosis. It has a direction, physiologically first backward in its main components with respect to the arterial flow. Here, LD was investigated in various disease and in presence of a unilateral carotid stent. METHODS: Carotid acquisitions were performed by ultrasound imaging on both body sides of 75 participants (150 Arteries). LD was measured in its percent quantity and direction. RESULTS: Obesity (p = 0.001) and carotid plaques (p = 0.01) were independently associated to quantity decrease of LD in the whole population. In a subgroup analysis, it was respectively 143% in healthy (n = 48 carotids), 129% (n = 34) in presence of cardiovascular risk factors, 121% (n = 20) in MACE patients, 119% (n = 24) in the carotid contralateral to a stent, 110% (n = 24) in carotids with stents. Regarding the direction of LD, in a subgroup analysis an inverted movement was identified in aged (p = 0.001) and diseased (p = 0.001) participants who also showed less quantity of LD (p = 0.001), but independently with age only (p = 0.002) in the whole population. CONCLUSIONS: This observational study suggests that LD within carotid wall layers is lower additively with ageing, cardiovascular risk factors, cardiovascular diseases, and stent. Even if stent is surely beneficial, these data might shed some light on stent restenosis, emphasising the need for interventional studies.

Abstract:

Mitral regurgitation is a structural heart disease characterized by dysfunction of the mitral valve. It can be treated through minimally invasive transcatheter edge-to-edge repair procedures, such as those involving the MitraClip system. Despite its clinical efficacy, the procedure poses technical challenges, requiring interventional cardiologists to undergo multiple procedures to achieve instrumentation proficiency. Previous studies have demonstrated the efficacy of extended reality simulators in improving surgical skill training. The purpose of this study is to evaluate the effectiveness of visuo-haptic feedback within an extended reality simulator to accelerate and enhance the learning process for the MitraClip procedure, potentially offering a valuable tool in surgical training. In this study, based on the prototype of our previous simulator, we further implemented a novel cardiac dynamic environment, a visual feedback interface, an innovative haptic armband, and a new model of the catheter to enhance the simulator’s fidelity and educational usability. Experiment results demonstrated that users benefiting from visuo-haptic feedback exhibited significantly reduced completion times, reduced deviation from predefined trajectories, and presented safer navigation results. Moreover, users preferred the visuo-haptic feedback experience, highlighting its perceived value in enhancing training. These findings underscore the potential of adding visuo-haptic feedback to extended reality simulators in surgical training for complex procedures, improving proficiency and patient surgery outcomes.

Abstract:

Accurate pelvic alignment is essential for assessing femoroacetabular impingement (FAI) when using CT-based three-dimensional bone models, as the acquisition is performed in a supine rather than a standing position, which does not accurately reflect the patients kinematics. This study investigated whether reliable pelvic alignment could be achieved using a reduced, unilateral field of view (FOV) based solely on treatment-side anatomical landmarks,with the goal of reducing both radiation exposure and segmentation time.Correlation analyses of landmarks, found on the pelvis and femora, showed strong interdependence, with Pearson correlation coefficients reaching up to 97 %, enabling the selection of the femoral head center (FHC), anterior inferior iliac spine (AIIS), and pubic tubercle (PT) for alignment. Regression models were trained on these landmarks to accurately predict both treatment-side and contralateral anterior superior iliac spine (ASIS) positions, which are required for the standard alignment method, achieving mean absolute errors between 2.39 mm and 6.03 mm across all three cartesian axes. Error propagation analyses indicated that the alignment differences resulting from the predictions averaged by 2.8 →, which was expected to influence only the average anteversion measurement in less than 15 % of the cases. Isolated outliers were observed for Center Edge and Tönnigs Angle.Reducing the superior scan limit, from the ASIS to the AIIS, allowed for an average reduction of 21 % in scan length, translating into a proportional decrease in radiation exposure. Furthermore, the findings of this study support the hypothesis that unilateral landmark information is sufficient for reliable pelvic alignment and FAI evaluation, offering a practical approach to optimize segmentation time and reduce patient dose. Further validation on larger datasets is recommended to ensure robustness and generalizability.