Background and aims:This article aims to develop and validate a methodology for realistic anatomical reconstruction of stented segments from optical coherence tomography (OCT) and angiographic images and the computation of local haemodynamic forces using computational fluid dynamics, towards a more realistic haemodynamic characterisation in coronary atherosclerotic lesions post-PCI. Methods:Three-dimensional anatomical models of stented vessels (6 silicone models in vitro and 16 patient models) were reconstructed from OCT and angiography, using a new methodology called REFINED, which relies on the separate reconstruction of the lumen and the stent; these two are then fused in a single model. Blood flow simulations were performed using the anatomical models to calculate the endothelial shear stress (ESS) and shear rate (SR). The models were geometrically and haemodynamically evaluated against those reconstructed using conventional models. Results:The reconstruction error distance was smaller with the REFINED approach than with the conventional approach, in both the silicone and patient models [8 (2-15) μm vs 23 (10-54) μm, p < 0.001, and 19 (9-32) μm vs 19 (7-36) μm, p < 0.001]. The REFINED method depicted higher ESS at the top of the struts in both the silicone (8.01 (4.21-14.92) Pa vs 5.54 (2.96-9.63) Pa; p < 0.001) and patient models [2.41 (1.41-3.65) Pa vs 1.44 (0.77-2.52) Pa; p < 0.001], and a more spatially heterogeneous SR compared to the conventional approach (p < 0.001 for all analyses). Conclusions:The REFINED approach outperforms the conventional method in reconstructing stent geometry and has been shown to be effective in facilitating a high-fidelity computation of the highly disturbed local haemodynamic environment. This demonstrates its potential role in examining device failure.

Abstract Accurate, reproducible interpretation of kidney allograft biopsies is critical for the diagnosis of graft injury and for informing prognosis and clinical management. The international Banff classification is a consensus diagnostic system based on semiquantitative histological lesion scoring according to either lesion extent or severity in kidney transplant biopsies. However, pathologist scoring is limited by interobserver variability, constrained scalability, and the inherent nature of the scoring system itself. Here we present BanffNET, a weakly supervised, probabilistic deep learning framework that combines self-supervised feature extraction with a novel Bayesian multiple-instance learning framework to predict (continuously) the full spectrum of Banff lesion scores directly from whole-slide images (WSIs). Using lesion-specific aggregation functions tailored to localized (modeling severity) and diffuse histological lesions (modeling extent), BanffNET generates interpretable, patch-level probability maps and calibrated slide-level scores. BanffNET’s performance was assessed relative to consensus, biological correlates of rejection and clinical outcome, demonstrating superior consistency, transportability and generalization. Trained on 7,533 WSIs from three cohorts, BanffNET demonstrates consistent performance on 12,687 WSIs across five external validation cohorts, matching or surpassing individual expert pathologists across lesion assessments. BanffNET scores align more closely than pathologist Banff scores with molecular profiles of rejection, offering an objective, transparent, biologically grounded framework for computational pathology with relevance beyond kidney transplantation.

Invasive coronary angiography (ICA) is the reference standard for diagnosing coronary artery disease and guiding percutaneous coronary intervention, yet clinical interpretation relies largely on visual assessment, which is variable and often requires additional invasive testing to assess functional significance. Artificial intelligence (AI)-based analysis of ICA images has emerged as a potential solution to automate interpretation, improve reproducibility, and extract anatomical and physiological information directly from angiograms. We conducted a systematic review of AI applications for ICA image analysis, registered in PROSPERO and reported according to PRISMA guidelines. A total of 134 studies were included, covering tasks across the ICA workflow, including automated frame selection, vessel segmentation, lesion detection and quantification, prediction of invasive physiological indices, coronary anatomy labelling, image registration and reconstruction, outcome prediction and left ventricular function estimation. Most studies focused on vessel segmentation and lesion assessment, generally demonstrating high internal performance but marked heterogeneity in datasets, reference standards, evaluation metrics, and validation strategies. While earlier work relied predominantly on single-centre retrospective validation, more recent studies increasingly incorporate multi-centre data, external validation and prospective evaluation. AI-based prediction of invasive physiological indices appears particularly promising for reducing reliance on wire-based measurements, though robust prospective evaluation remains limited. Overall, AI-based ICA analysis has progressed from technical feasibility studies towards clinically oriented applications. However, challenges in generalizability, methodological standardization, and workflow integration must be addressed to enable reliable clinical adoption.

BACKGROUND:The incorporation of side branches in vessel geometry influences wall shear stress (WSS) distribution. However, complete vessel reconstruction is time-consuming, and there is no evidence that its WSS estimations better predict atherosclerotic disease progression compared with the output of the conventional single-vessel reconstruction (SVR). METHODS:Patients who had baseline and 1-year follow-up intravascular ultrasound imaging (n=40 vessels), and patients with neoatherosclerotic lesions (n=13 vessels) on optical coherence tomography were included. All the studied vessels had at least one side branch with a diameter >1 mm; 3-dimensional complete vessel reconstruction and SVR were performed, and the time-averaged WSS and multidirectional WSS were computed. The performance of both methods in predicting disease progression in intravascular ultrasound and optical coherence tomography models was assessed. RESULTS:The incorporation of side branches in 3-dimensional geometry resulted in lower minimum predominant time-averaged WSS in the intravascular ultrasound (1.09 versus 1.58 Pa, P<0.001) and optical coherence tomography-based reconstructions (0.68 versus 1.33 Pa, P<0.001) and influenced the multidirectional WSS distribution. In native segments, complete vessel reconstruction-derived WSS metrics demonstrated superior predictive performance for disease progression-defined as lumen area reduction and plaque burden increase-compared with SVR, as evidenced by improved out-of-sample accuracy (leave-one-out information criterion: 429 versus 551), discrimination (C statistic: 0.725 versus 0.651), calibration (Brier score: 0.172 versus 0.226), and explained variance (27.8% versus 20.7%). Consistent findings were observed in stented segments, where complete vessel reconstruction-derived WSS metrics more accurately predicted neointimal proliferation than SVR-derived metrics. CONCLUSIONS:Incorporating side branches into vessel reconstruction influences WSS distribution and enables more accurate prediction of atherosclerotic disease progression in native and stented segments than SVR.

Intravascular ultrasound (IVUS) and optical coherence tomography (OCT) are complementary imaging modalities to assess atherosclerosis in vivo. Combining both modalities in a single imaging system has been shown to improve the characterization of vulnerable plaques that are likely to cause acute coronary events. However, fundamental differences in tissue sensitivities and acquisition protocols make the registration of sequences challenging. Anatomical landmarks used to align IVUS and OCT sequences can be masked or lack visual similarity between modalities which renders manual alignment time-consuming and prone to observer variability, limiting its clinical use. Existing methods impose strict frame-level correspondences leading to instability in low information regions, and rely on a two-step registration process that compounds alignment errors. We propose IntraCross, a novel graph matching framework that learns partial assignments between landmarks rather than enforcing rigid frame-by-frame matching, enabling flexible correspondences while rejecting unmatchable landmarks. This is the first method to perform both temporal and rotational registration simultaneously, aligning with clinical workflows. We extend existing partial matching techniques from 2D to 3D sequences and incorporate a temporal prior to regularize the matching process. Testing in 77 vessels from 22 patients showed a high agreement with expert analysts (Williams Index=1.1; p=0.62, 0.89, 0.07) and our approach outperforms existing methods reported in the literature for circumferential registration (p=0.01, 0.04).

BACKGROUND:Invasive intracoronary imaging represents the gold standard for identifying vulnerable coronary plaques, but it is not suitable for widespread clinical use. Coronary computed tomography angiography (CTA) may offer a noninvasive alternative. OBJECTIVES:This study aims to integrate coronary CTA-derived plaque morphology, pericoronary inflammation, and plaque burden into a unified morphology-inflammation-burden (MIB) score and to evaluate its association with plaque vulnerability and clinical outcomes. METHODS:Patients undergoing coronary CTA followed by optical coherence tomography (OCT) and intravascular ultrasound (IVUS) were followed for a median of 31 months. High-risk plaque, pericoronary adipose tissue attenuation, and total plaque burden (TPB) were quantified and compared with invasive imaging. A vulnerable lesion was defined as ≥2 vulnerability features on OCT. RESULTS:A total of 438 patients (median age 67 years) and 1,038 plaques were included; 45.4% presented with non-ST-segment elevation acute coronary syndrome. High-risk plaque, elevated pericoronary adipose tissue attenuation, and high TPB were independently associated with OCT-defined vulnerability (P < 0.05 for all). TPB correlated with IVUS percent atheroma volume (Pearson's r = 0.69; P < 0.001). The MIB score demonstrated a stepwise increase in vulnerability, exceeding a predicted risk of 90% in the highest category. Vulnerable patients, defined by the presence of ≥1 untreated lesion with a high MIB score, had a significantly higher rate of cardiac death, acute coronary syndrome, or revascularization (15.3% vs 4.4%; P < 0.001). CONCLUSIONS:A coronary CTA-derived MIB score correlates with plaque vulnerability by intracoronary imaging and identifies patients at increased risk for adverse events. These findings support the value of coronary CTA for noninvasive risk stratification in clinical practice. (Massachusetts General Hospital and Tsuchiura Kyodo General Hospital Coronary Imaging Collaboration; NCT04523194).

BACKGROUND AND AIMS:To clarify the relationship between haemodynamic milieu and lipid core plaques among culprit and non-culprit coronary vessels. METHODS:A total of 45 vessels from 20 patients with acute coronary syndrome who underwent invasive coronary angiography were prospectively enrolled for additional near-infrared spectroscopy intravascular ultrasound (NIRS-IVUS) imaging to quantify lipid plaque content. Haemodynamics assessments between culprit (n = 19, one excluded due to suboptimal angiography) and non-culprit (n = 25) vessels were performed using three-dimensional arterial reconstructions derived from fused NIRS-IVUS and quantitative coronary angiography imaging. RESULTS:Culprit vessels were characterised by a higher probability of lipid core containing coronary plaques (0.00 [interquartile range, IQR: 0.00-0.278] vs. 0.00 [0.00-0.119], p ≪ 0.05). The greatest haemodynamics differentiation between the two groups, in descending order, was observed in transverse endothelial shear stress (transESS), oscillatory shear index (OSI), and elevated blood viscosity (EBV). In mixed logistic regression, after adjusting for other haemodynamic metrics, culprit vessels exhibited decreasing odds of moderate-to-high OSI (odds ratio [OR] 0.484, 95% confidence interval [CI] 0.304-0.770, p = 0.002 and OR 0.578, 95% CI 0.339-0.985, p = 0.044, respectively), as was moderate transESS (OR 0.440, 95% CI 0.276-0.702, p < 0.001), but increasing odds of moderate-to-high EBV (OR 2.030, 95% CI 1.180-3.491, p = 0.010 and OR 4.373, 95% CI 2.017-9.479, p < 0.001, respectively). CONCLUSIONS:This study observed elevated blood viscosity within culprit vessels, as a potentially underexplored feature of plaque vulnerability. Our findings suggest that blood viscosity is associated with vessel-specific differences in lipid core plaque, though the small sample size means these findings should be considered hypothesis-generating.

BACKGROUND:The resorbable fibrillated scaffold (RFS) is a novel electrospun, polylactide-based endoluminal scaffold developed for peripheral arterial applications. Its porous microfibre architecture is intended to support host cell infiltration and vascular restoration. Although its near-wall hemodynamic behaviour has been characterised, its serial anatomical and virtual physiological evolution after implantation has not previously been examined. AIMS:To characterise the serial anatomical changes of the RFS after implantation in peripheral arterial models, and to assess the hemodynamic significance of luminal narrowing during follow-up using image-derived virtual flow indices. METHODS:Two preclinical studies were conducted in rabbit and mini-pig peripheral arterial models. The RFS was implanted bilaterally in the external iliac arteries of three rabbits and in the profunda femoris arteries of six mini-pigs. Serial follow-up to 3 months was performed using invasive angiography and intravascular optical coherence tomography (OCT) for anatomical assessment. Virtual physiology was assessed using angiography-derived Murray-law-based quantitative flow ratio (μFR) and OCT-derived flow ratio (OFR). RESULTS:Implantation was technically successful in all but one case. In rabbits, between post-implantation and 3 months, reference vessel diameter increased from 2.08 to 2.54 mm (p = 0.03), while minimum lumen diameter remained stable; OCT-derived area stenosis increased from 28.9% to 56.1% (p < 0.01), scaffold length shortened from 10.40 to 8.63 mm (p < 0.01), and μFR decreased from 0.99 to 0.92 (p = 0.04), whereas OFR remained unchanged. In mini-pigs, the principal changes occurred between post-implantation and 1 month, with reductions in minimum lumen diameter, minimum lumen area, μFR, and OFR (all p ≤ 0.03), accompanied by increased stenosis and relative stabilization thereafter. Both μFR and OFR declined non-linearly with increasing OCT-derived %AS, with similar model fit (R2 = 0.57). CONCLUSION:This preclinical study supports the feasibility of RFS implantation in peripheral arteries and demonstrates that the device undergoes early structural evolution after deployment. Virtual physiological assessment showed that these anatomical changes were accompanied by measurable reductions in flow indices, though most values remained above the 0.80 reference threshold. Together, these findings provide a basis for further development of the technology and for future studies incorporating both anatomical and virtual physiological assessment.

Percutaneous thermal ablation is a minimally invasive treatment for hepatocellular carcinoma. Evaluating the treatment success depends on the accurate quantification of the margin achieved between the ablation zone and the tumor. However, manual delineation of the ablation zone is labor-intensive, motivating the development of automated approaches. Existing deep learning-based ablation zone segmentation methods adopt voxel-wise segmentation. Voxel-wise models often underperform in noisy or low-contrast cases with poorly visible ablation zone borders, producing underestimated and fragmented masks that require complex post-processing. In addition, existing ablation zone segmentation models are trained primarily on tumor segmentation datasets, or rely on user interaction for mask correction. Contour-based segmentation methods improve boundary delineation, but their performance typically depends on the visibility of the borders, and produce overly-smooth contours. In this work, we present a fully automated deep learning model for ablation zone segmentation that addresses these limitations by predicting contours directly in the Fourier domain. Fourier contour embeddings enable precise modeling of curved shapes, and produce continuous segmentation masks. Thus, eliminating the need for extensive post-processing or manual correction. To avoid overly-smooth contours, we introduce a multiscale deep supervision strategy with dynamic loss weighting, encouraging the model to capture high-frequency boundary features. We train and evaluate our method on a dedicated ablation zone dataset specifically annotated for this task. Our results demonstrate improved prediction accuracy in Dice score and distance-based metrics compared to existing models. In particular, a detailed contrast-to-noise ratio (CNR) analysis shows that our model consistently outperforms existing approaches across all CNR levels.

Background: This study compared changes in percentage atheroma volume (PAV) using an end-diastolic (ED) intravascular ultrasound (IVUS) segmentation approach vs. the conventional 1-mm interval analysis in serial IVUS data from the PACMAN-AMI trial. Methods and results: IVUS data from the PACMAN-AMI study were analyzed by 2 core laboratories: one with 1-mm segmentation and the other with an ED-based approach. The same arterial segments were assessed at baseline and at the 52-week follow-up in patients receiving alirocumab or placebo. Changes in segment length, lumen, vessel, total atheroma volume (TAV), and PAV between baseline and follow-up were compared between methods. Biomarkers associated with atherosclerotic progression were measured and correlated with TAV and PAV changes. In all, 387 segments were analyzed. Agreement between conventional and ED volumetric analysis was excellent (intraclass coefficient >0.891, P<0.001). TAV and PAV were larger in both groups in the ED analysis than with the conventional approach; however, changes between treatment arms were similar for the conventional and ED analyses (TAV: 14.34 vs. 14.64 mm(3), respectively [P=0.823]; PAV: 1.29% vs. 1.25%, respectively [P=0.911]). Biomarker correlations with TAV and PAV changes did not differ between approaches. Conclusions: ED- and 1-mm-based analyses demonstrated comparable treatment effects of alirocumab on plaque regression in PACMAN-AMI. These findings support the use of the less time-consuming 1-mm segmentation method in serial IVUS studies.

Aims:To develop a deep-learning (DL) framework that enables fully automated longitudinal and circumferential co-registration of intravascular ultrasound (IVUS) and optical coherence tomography (OCT) images. Methods and results:Data from 230 patients (714 vessels) with acute myocardial infarction that underwent near-infrared spectroscopy IVUS and OCT imaging in their non-infarct related vessels were analysed. Experts annotated the lumen borders (61 655 IVUS and 62 334 OCT frames), the side branches and the calcific tissue (10 000 IVUS and 10 000 OCT frames each). This information was used to train DL models that extracted these features that were then used by a dynamic time warping algorithm to co-registered longitudinally the IVUS and OCT images. The circumferential registration of IVUS and OCT was performed through a rotation cost matrix and dynamic programming. On a test set of 22 patients (77 vessels), the DL method showed high concordance with the expert analysts for the longitudinal and circumferential co-registration of the two datasets (concordance correlation coefficient >0.99 and >0.90, respectively). The Williams Index was 0.96 for longitudinal and 0.97 for circumferential alignment, indicating a comparable performance of the proposed framework to the analysts. The time needed for the DL pipeline to process imaging data from a vessel was <90 s. Conclusion:A fully automated, DL-based framework for IVUS-OCT co-registration demonstrated both speed and accuracy, with performance comparable to that of expert analysts. These features enable its application in research using large-scale data incorporating multimodality imaging.

Aims:Segmental pressure gradients post-percutaneous coronary intervention (PCI) can detect residual disease and optimization targets. Ultrasonic flow ratio (UFR) or optical flow ratio (OFR) offer simultaneous physiological and morphological assessment using a single imaging catheter. This study evaluated the utility of UFR and OFR in identifying residual disease post-PCI. Methods and results:The study include patients from the Acetyl Salicylic Elimination Trial JAPAN Pilot study with complete intravascular imaging pullback data, where UFR or OFR was obtained post-PCI. Anatomical focal lesions distal and proximal to the stent were analysed in segments ≥5 mm long. UFR or OFR virtual pullback curves assessed intra-stent pressure gradients, defining physiological focal or diffuse by segmental pressure drops ≥0.05 over lengths <10 or ≥10 mm, respectively. The median post-PCI UFR/OFR was 0.93 (0.88-0.96) with 35.4% (69/195) vessels having a UFR/OFR < 0.91. There were significantly more focal lesions, both anatomical and physiological, proximal and distal to the stent in vessels with UFR/OFR < 0.91 compared with those ≥0.91. Agreement between anatomical and physiological focal lesions was moderate proximally (kappa = 0.553, P < 0.001) and fair distally (kappa = 0.219, P = 0.002). The in-stent gradient poorly predicted significant stent under-expansion. However, the virtual fractional flow reserve gradient performed well in detecting proximal or distal focal disease (area under the curve = 0.835 and 0.877, respectively). Conclusion:UFR/OFR effectively identifies sub-optimal vessel physiology post-PCI and locates precise anatomical issues, validated by intravascular imaging. Trial registration:The ASET JAPAN ClinicalTrials.gov reference: NCT05117866.

Liver-vessel segmentation is an essential task in the pre-operative planning of liver resection. State-of-the-art 2D or 3D convolution-based methods focusing on liver vessel segmentation on 2D CT cross-sectional views, which do not take into account the global liver-vessel topology. To maintain this global vessel topology, we rely on the underlying physics used in the CT reconstruction process, and apply this to liver-vessel segmentation. Concretely, we introduce the concept of top-k maximum intensity projections, which mimics the CT reconstruction by replacing the integral along each projection direction, with keeping the top-k maxima along each projection direction. We use these top-k maximum projections to condition a diffusion model and generate 3D liver-vessel trees. We evaluate our 3D liver-vessel segmentation on the 3D-ircadb-01 dataset, and achieve the highest Dice coefficient, intersection-over-union (IoU), and Sensitivity scores compared to prior work.

BACKGROUND:Endothelial shear stress (ESS) is an instigator of vulnerable plaque formation and destabilization. Traditionally, their computation is performed in models reconstructed from the fusion of intravascular imaging and angiography. Three-dimensional quantitative coronary angiography (3D-QCA) and computed tomography coronary angiography (CCTA) have emerged as alternative approaches to assess flow patterns, however, there is limited evidence about their performance. METHODS:We analysed data from 27 patients (38 vessels) that underwent coronary angiography, CCTA and near-infrared spectroscopy intravascular ultrasound (NIRS-IVUS) imaging. In each vessel, four reconstruction models were generated: 3D-QCA, CCTA and two fusion models - NIRS-IVUS with 3D-QCA (Angio-NIRS-IVUS) and CCTA (CCTA-NIRS-IVUS). In these models, the minimum and maximum predominant ESS were computed in 3 mm segments and the detected lesions, and their estimations were compared using the CCTA-NIRS-IVUS as the reference standard. RESULTS:In the 3 mm analysis, the Angio-NIRS-IVUS and CCTA estimations had a higher correlation with CCTA-NIRS-IVUS than 3D-QCA for the minimum (intraclass correlation coefficient, ICC: 0.822 vs 0.704 vs 0.581, p < 0.001) and maximum ESS (ICC: 0.852 vs 0.758 vs 0.634, p < 0.001). In lesion-level analysis, only the CCTA-NIRS-IVUS and Angio-NIRS-IVUS (ICC: 0.606, p < 0.001) estimations for the minimum ESS were correlated, while for the maximum ESS, there was a stronger correlation between CCTA-NIRS-IVUS and Angio-NIRS-IVUS and CCTA compared to the 3D QCA (ICC: 0.898 vs 0.836 vs 0.742, p < 0.001). CONCLUSIONS:A strong association was noted for the ESS estimated in the hybrid NIRS-IVUS-based reconstructions with CCTA appearing as the 2nd best modality for assessing the local hemodynamic milieu.