Rui Li

Hypertrophic cardiomyopathy (HCM) is a genetically and phenotypically heterogeneous myocardial disease associated with a broad spectrum of clinical outcomes, including heart failure, atrial fibrillation, ventricular arrhythmias, and sudden cardiac death. Cardiac magnetic resonance (CMR) plays a central role in the comprehensive assessment of HCM by enabling integrated evaluation of myocardial morphology, ventricular and atrial function, tissue characterization, myocardial deformation, and longitudinal disease progression within a single comprehensive imaging modality. However, conventional CMR interpretation remains limited by manual post-processing, interobserver variability, sequence-dependent measurements, and reliance on a relatively restricted set of predefined imaging parameters. These limitations may hinder the full exploitation of the rich multidimensional information contained in CMR datasets. Artificial intelligence (AI) approaches, particularly machine learning and deep learning, together with radiomics-based quantitative image analysis, provide new opportunities to extend the diagnostic and prognostic value of CMR to extract high-dimensional and quantitative information from cine imaging, late gadolinium enhancement, parametric mapping, and feature-tracking-based deformation analysis. Emerging evidence suggests that AI-enhanced CMR may facilitate automated image analysis, improve diagnostic classification, support differentiation of HCM from phenocopies, enable tissue- and deformation-based phenotypic characterization, and enhance individualized risk prediction. Although several studies have included independent or external testing, many remain retrospective, rely on selected populations or internal validation, and lack prospective assessment of clinical utility. High discrimination performance should not be interpreted as established clinical utility, because calibration, decision-curve analysis, clinically relevant thresholds, and prospective clinical impact remain insufficiently evaluated. In addition, interpretability and clinical generalizability remain important challenges for translation into routine practice. This review summarizes the current role of CMR in HCM and discusses emerging applications, methodological challenges, and future directions of AI-enhanced CMR for diagnosis, phenotypic characterization, and risk stratification.

Background: Dilated cardiomyopathy (DCM) is a major cause of heart failure and sudden cardiac death (SCD), with a 5-year survival rate of approximately 45%–50%. Current risk stratification is predominantly dependent on left ventricular ejection fraction (LVEF), which has limited sensitivity and specificity. Hence, more effective biomarkers should be used in late gadolinium enhancement (LGE)-negative patients. Methods: A total of 378 consecutive patients with LGE-negative DCM were enrolled from four hospitals between December 2016 and December 2022. Cardiac magnetic resonance imaging-derived strain parameters (global radial strain, global circumferential strain, and global longitudinal strain [GLS]) were assessed against the primary (SCD and related events) and secondary (heart failure, appropriate implantable cardioverter-defibrillator therapy) endpoints. Internal validation was performed using stratified bootstrap resampling with Harrell’s optimism correction to report the optimism-corrected C-index. Data were accessed for research purposes from 15/06/2023–30/12/2023, and all records were de-identified prior to analysis. Results: Over a median follow-up of 59.78 months, 35 (9.26%) and 72 (19.0%) patients presented with the primary and secondary endpoints, respectively. Based on the multivariate Cox analysis, GLS, LVEF, and age were independent prognostic factors. However, only GLS (HR = 1.37; P = 0.041) remained significant in the LVEF P < 0.001). This advantage persisted after bootstrap internal validation (B = 1000; optimism-corrected C-index 0.754 vs 0.711, Holm adjusted P = 0.048). Further inclusion of age and New York Heart Association (NYHA) classification enhanced the model’s performance (model 4): apparent C-index 0.801; optimism-corrected C-index 0.785). Conclusion: GLS is an independent predictor of SCD-related events in LGE-negative DCM. Incorporating GLS with conventional indicators such as age, NYHA classification, and LVEF significantly enhances prognostic discrimination and model robustness, indicating potential value for future clinical risk stratification.

In this letter, we comment on the recent article by Karangeli et al , which systematically examined the spatial relationship between the hyoid bone and the carotid arteries using computed tomography angiography. The authors revealed significant anatomical variability in the 224 patients in their study, with type VI (external carotid artery lateral to the hyoid) and type VIII (both carotid arteries lateral to the hyoid) being the most common non-null configurations. Notably, sex-based differences were identified, suggesting possible morphometric or biomechanical influences. These findings have important implications for radiologists and surgeons, as non-null variants - particularly types VIII and XI - might predispose to dynamic carotid compression, transient ischemic events, or intraoperative vascular injury. Recognizing these patterns during preoperative imaging can enhance surgical safety and improve the diagnostic evaluation of unexplained cerebrovascular symptoms. The study by Karangeli et al underscores the clinical importance of translating detailed anatomical mapping into preventive strategies that bridge imaging precision and vascular protection.

Objectives: Aiming to evaluate the predictive value of high-resolution computed tomography (HRCT) features for identifying acute/subacute progression in patients with polymyositis/dermatomyositis (PM/DM)-associated interstitial lung disease (ILD), and to develop a risk stratification algorithm based on clinico-radiologic parameters. Methods: This retrospective cohort study included 282 patients with PM/DM who underwent HRCT from January 2020 to December 2024. Baseline clinical data and HRCT imaging characteristics were systematically collected. Over time, 140 patients with PM/DM-ILD were followed. HRCT scores and imaging patterns were assessed, and cases of acute/subacute ILD progression were documented during the follow-up period. Penalized Cox regression (LASSO) was conducted to identify independent risk factors associated with disease progression and to develop a risk stratification method. The concordance index (C-index), net reclassification improvement (NRI), integrated discrimination improvement (IDI) and decision curve analysis (DCA) were used to evaluate the discriminative ability of this stratification. Algorithm performance was assessed using calibration plots to evaluate agreement between predicted and observed risks. Results: During a median follow-up duration of 5.69 months (IQR, 1.77–5.91 months), 56 (40.0%) patients experienced acute/subacute ILD progression. The HRCT score was considered an independent predictor of acute/subacute progression in patients with PM/DM-ILD. A newly developed risk stratification scheme, according to thresholds of HRCT score, imaging classification (organizing pneumonia [OP] vs. non-OP patterns), and anti-MDA5 antibody status, demonstrated good predictive ability for identifying patients at risk of progression. In combination with clinical parameters, the integrated predictive algorithm significantly outperformed traditional clinical risk algorithm, with significant enhancements in C-index (to 0.764). The incremental predicted value was demonstrated by improved NRI (0.470), IDI (0.218), and DCA metrics. Conclusion: A high HRCT score is an independent predictor of acute/subacute progression in patients with PM/DM-ILD. Incorporating clinical parameters into the imaging-based algorithm significantly improves its predictive accuracy for progressive disease.

BACKGROUND Type 2 diabetes mellitus (T2DM) substantially increases the risk of cardiovascular disease, including ischemia with non-obstructive coronary artery disease (INOCA). Coronary computed tomography angiography (CCTA) enables early detection of coronary abnormalities; however, it may fail to identify INOCA due to the absence of overt stenosis. Pericoronary adipose tissue attenuation (PCATa) values derived from CCTA in the proximal segments of the left anterior descending (LAD), left circumflex (LCX), and right coronary artery (RCA) serve as effective imaging biomarkers for coronary inflammation. However, its clinical use for identifying INOCA in patients with T2DM remains poorly defined. AIM To investigate PCATa differences and their diagnostic value for identifying INOCA in patients with T2DM. METHODS This retrospective study involved 228 T2DM patients underwent CCTA and 120 healthy individuals. The mean PCATa values within the proximal segments of the three major coronary arteries were compared between groups. Further subgroup analysis was performed to assess the differences in PCTAa and clinical characteristics between T2DM patients with and without INOCA. Logistic regression analysis was conducted to identify the independent risk factors for INOCA, and the receiver operating characteristic curves were generated to evaluate the diagnostic performance of each indicator. RESULTS Compared with controls, T2DM patients exhibited significantly higher PCATa values in all three major coronary arteries. Among them, those with concomitant INOCA showed further increases compared to those without INOCA (all P < 0.05). Multivariate logistic regression identified age; female sex; elevated glycated hemoglobin; and increased PCATa in the LAD, LCX, and RCA as independent risk factors for INOCA. Receiver operating characteristic analysis showed good diagnostic performance for PCATa [LAD area under the curve (AUC) = 0.809; LCX AUC = 0.777; RCA AUC = 0.758], outperforming traditional clinical indicators (AUC = 0.731). Combining PCATa with clinical parameters yielded the highest diagnostic accuracy (LAD AUC = 0.851; LCX AUC = 0.842; RCA AUC = 0.841). CONCLUSION Elevated proximal PCATa is an independent risk factor for INOCA in T2DM. Combining PCATa with clinical data improves diagnostic performance in this population.

Background/Objectives : Quantitative computed tomography (CT) metrics are widely used to assess pulmonary involvement and to predict short-term severity in coronavirus disease 2019 (COVID-19). However, it remains unclear whether baseline artificial intelligence (AI)-based quantitative high-resolution computed tomography (HRCT) metrics of pneumonia burden provide incremental prognostic value for in-hospital composite adverse outcomes beyond routine clinical factors, or whether these imaging-derived markers carry any exploratory signal for long-term severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) reinfection among hospitalized patients. Most existing imaging studies have focused on diagnosis and acute-phase prognosis, leaving a specific knowledge gap regarding AI-based quantitative HRCT correlates of early deterioration and subsequent reinfection in this population. To evaluate whether combining deep learning-derived, quantitative, HRCT features and clinical factors improve prediction of in-hospital composite adverse events and to explore their association with long-term reinfection in patients with COVID-19 pneumonia. Methods: In this single-center retrospective study, we analyzed 236 reverse-transcription polymerase chain reaction (RT-PCR)-confirmed COVID-19 patients who underwent baseline HRCT. Median follow-up durations were 7.65 days for in-hospital outcomes and 611 days for long-term outcomes. A pre-trained, adaptive, artificial-intelligence-based, prototype model (Siemens Healthineers) was used for pneumonia analysis. Inflammatory lung lesions were automatically segmented, and multiple quantitative metrics were extracted, including opacity score, volume and percentage of opacities and high-attenuation opacities, and mean Hounsfield units (HU) of the total lung and opacity. Patients were stratified based on receiver operating characteristic (ROC)-derived optimal thresholds, and multivariable Cox regression was used to identify predictors of the composite adverse outcome (intensive care unit [ICU] admission or all-cause death) and SARS-CoV-2 reinfection, defined as a second RT-PCR-confirmed episode of COVID-19 occurring ≥90 days after initial infection. Results: The composite adverse outcome occurred in 38 of 236 patients (16.1%). Higher AI-derived opacity burden was significantly associated with poorer outcomes; for example, opacity score cut-off of 5.5 yielded an area under the ROC curve (AUC) of 0.71 (95% confidence interval [CI] 0.62–0.79), and similar performance was observed for the volume and percentage of opacities and high-attenuation opacities (AUCs up to 0.71; all p < 0.05). After adjustment for age and comorbidities, selected HRCT metrics—including opacity score, percentage of opacities, and mean HU of the total lung (cut-off −662.38 HU; AUC 0.64, 95% CI 0.54–0.74)—remained independently associated with adverse events. Individual predictors demonstrated modest discriminatory ability, with C-indices of 0.59 for age, 0.57 for chronic obstructive pulmonary disease (COPD), 0.62 for opacity score, 0.63 for percentage of opacities, and 0.63 for mean total-lung HU, whereas a combined model integrating clinical and imaging variables improved prediction performance (C-index = 0.68, 95% CI: 0.57–0.80). During long-term follow-up, RT-PCR–confirmed reinfection occurred in 18 of 193 patients (9.3%). Higher baseline CT-derived metrics—particularly opacity score and both volume and percentage of high-attenuation opacities (percentage cut-off = 4.94%, AUC 0.69, 95% CI 0.60–0.79)—showed exploratory associations with SARS-CoV-2 reinfection. However, this analysis was constrained by the very small number of events (n = 18) and wide confidence intervals, indicating substantial statistical uncertainty. In this context, individual predictors again showed only modest C-indices (e.g., 0.62 for procalcitonin [PCT], 0.66 for opacity score, 0.66 for the volume and 0.64 for the percentage of high-attenuation opacities), whereas the combined model achieved an apparent C-index of 0.73 (95% CI 0.64–0.83), suggesting moderate discrimination in this underpowered exploratory reinfection sample that requires confirmation in external cohorts. Conclusions: Fully automated, deep learning-derived, quantitative HRCT parameters provide useful prognostic information for early in-hospital deterioration beyond routine clinical factors and offer preliminary, hypothesis-generating insights into long-term reinfection risk. The reinfection-related findings, however, require external validation and should be interpreted with caution given the small number of events and limited precision. In both settings, combining AI-based imaging and clinical variables yields better risk stratification than either modality alone.

This study aimed to evaluate the prognostic value of lesion-specific and proximal 40 mm pericoronary adipose tissue attenuation (PCATa) in predicting major adverse cardiovascular events (MACE) in patients with ischemic heart disease presenting with angina pectoris and to compare their predictive performance. This retrospective study included 213 patients with ≥ 50% coronary artery stenosis who underwent coronary computed tomography angiography and digital subtraction angiography between January 2020 and August 2023. MACE was assessed over a median follow-up period of 15 months. Cox proportional hazards models were used to identify MACE risk factors, and the incremental predictive value of PCATa was evaluated using the C-index and global Chi-square statistics. MACE occurred in 72 patients (33.80%). The proximal 40 mm PCATa (LAD: HR = 1.041, LCX: HR = 1.046, RCA: HR = 1.043; all P < 0.01) and lesion-specific PCATa (HR = 1.048, P = 0.001) were independent predictors of MACE, and their prognostic performance was comparable. In the secondary endpoint analysis, defined as a composite of all-cause mortality, sudden cardiac death, and acute myocardial infarction, only elevated RCA-PCATa (HR = 1.084, 95% CI: 1.026–1.146, P = 0.004) was independently associated with adverse outcomes. Risk stratification models integrating PCATa cutoff values and angina status demonstrated robust predictive ability for MACE. Incorporating PCATa into clinical risk models significantly improved the C-index and global Chi-square values for MACE prediction (all P < 0.01). These findings suggest that incorporating PCATa to clinical models significantly improves prognostic accuracy, with comparable performance between proximal and lesion-specific PCATa, while RCA-PCATa may provide superior prognostic value for the most severe cardiovascular outcomes.

Background: Chronic low-grade inflammation in nonalcoholic fatty liver disease (NAFLD) plays a critical role in the development of cardiovascular complications, particularly ischaemia with non-obstructive coronary arteries (INOCA). This study aimed to develop and evaluate models combining pericoronary adipose tissue (PCAT) radiomics, PCAT attenuation (PCATa), CCTA plaque parameters, and clinical risk factors to identify INOCA in NAFLD patients. Methods: This retrospective study included 159 patients with NAFLD who underwent CCTA. The patients were randomly divided into the training (70%) and validation (30%) cohorts. Clinical features, CCTA imaging indicators, and right coronary artery PCAT radiomic features were analyzed. Five models were constructed using logistic regression: Model 1 (PCATa model), Model 2 (radiomics model), Model 3 (clinical factors model), Model 4 (combined imaging model), and Model 5 (combined imaging-clinical model). The models’ diagnostic performance was assessed using the area under the curve, reclassification metrics, and decision curve analysis (DCA). Results: The PCAT radiomics model exhibited higher diagnostic efficacy than the PCATa model in identifying INOCA (training cohort: AUC 0.734 vs. 0.674; validation cohort: AUC 0.706 vs. 0.637). The combined imaging model showed improved performance over the clinical factors model (training AUC 0.830, validation AUC 0.813). The model integrating imaging and clinical factors achieved the highest diagnostic accuracy (AUCs of 0.873 and 0.824 in the training and validation cohorts, respectively), demonstrating incremental value based on improved NRI, IDI, and DCA metrics. Calibration analysis indicated good agreement between predicted and observed outcomes. Conclusions: The radiomics model provided better discrimination than the PCATa model for identifying INOCA among patients with NAFLD. Models incorporating radiomics and CCTA imaging parameters outperformed those based solely on clinical factors. The comprehensive imaging-clinical model achieved the best overall performance and may serve as a promising non-invasive approach for INOCA risk stratification in NAFLD, although external validation is still required.

Objective: To determine whether deep-learning–derived quantitative HRCT parameters, combined with clinical characteristics, can predict in-hospital adverse outcomes and long-term reinfection in patients with COVID-19 pneumonia. Methods: We retrospectively analyzed 236 RT-PCR–confirmed patients who underwent HRCT between November 2022 and January 2023 at the Affiliated Hospital of North Sichuan Medical College. Pulmonary inflammatory regions were automatically segmented, and quantitative metrics—including opacity score, lesion volume and percentage, high-attenuation lesion volume and percentage, and mean total-lung attenuation—were extracted using an artificial-intelligence pneumonia analysis prototype (Siemens Healthineers, Erlangen, Germany). Optimal thresholds derived from receiver-operating-characteristic curves for the composite endpoint of intensive-care-unit admission or all-cause death were applied to stratify patients. Cox proportional-hazards models were used to identify independent predictors of adverse outcomes and subsequent SARS-CoV-2 reinfection. Results: Adverse outcomes occurred in 16.1% of patients. Higher opacity scores, larger lesion burdens, greater proportions of high-attenuation opacities, and higher mean lung attenuation were all associated with poorer outcomes (all P < 0.05). After adjusting for age and chronic obstructive pulmonary disease, an opacity score ≥ 5.5 (HR = 3.02), lesion percentage ≥ 18.85% (HR = 2.33), and mean attenuation ≥ −662.4 HU (HR = 2.20) remained independent predictors. During a median follow-up of 603 days, opacity score ≥ 5.5 (HR = 5.32), high-attenuation volume ≥ 140.37 ml (HR = 3.81), and high-attenuation percentage ≥ 4.94% (HR = 3.39) independently predicted reinfection (all P ≤ 0.027). Conclusions: Deep-learning–based quantitative HRCT metrics provide incremental prognostic information for risk stratification of both acute adverse outcomes and long-term reinfection among COVID-19 patient.