Citations

Abstract Background Radiomics extracts quantitative imaging features from computed tomography (CT) data for clinical decision‐making. However, variations in acquisition parameters—particularly x‐ray tube voltage (kV)—introduce non‐biological variability in attenuation values, limiting the reproducibility of radiomic features across scanners, protocols, and institutions. Purpose To develop and evaluate a CT dAta harmoNiZAtion framework based on deep learNed vIrTual monoEnergetic imaging (TANZANITE), which leverages the keV flexibility of virtual monoenergetic images (VMIs) to enable cross‐kV scan translation and radiomics harmonization. Methods TANZANITE is a model hub consisting of multiple pre‐trained convolutional neural networks (CNNs), each designed to translate VMIs from 1 keV level to another. Phantom‐based calibration was first used to determine energy‐equivalent (Eff_E) keV levels corresponding to each tube potential (e.g., Eff_E(A) keV for source kV and Eff_E(B) keV for target kV). A CNN trained using 69 120 patches from seven patient cases to map VMIs from Eff_E(A) to Eff_E(B) was selected from the TANZANITE hub and applied directly to clinical CT images acquired at the source kV. This harmonized the images to match the attenuation characteristics of the target kV setting. Evaluation was conducted on independent dual‐energy CT datasets acquired at 100/Sn150 kV. Regions of interest (ROIs) were placed in the kidney, liver, and spine to assess CT number consistency and radiomic feature reproducibility. The concordance correlation coefficient (CCC) was calculated across 93 non‐shape radiomic features. Results After TANZANITE processing with 100 kV images, CT numbers in evaluated organs closely matched the Sn150 kV reference values in four testing patient cases. For example, mean kidney CT numbers changed from 320 HU (100 kV) to 156 HU (TANZANITE), approximating the Sn150 kV value of 160 HU. Similar changes were observed in the liver (157–105 HU vs. 104 HU reference) and spine (45–22 HU vs. 19 HU reference). Radiomic reproducibility improved substantially across organs: mean CCC increased from 0.590 to 0.995 in the liver, 0.300 to 0.970 in the kidney, and 0.630 to 0.968 in the spine. Post‐TANZANITE, over 98% of features exceeded the stability threshold (CCC ≥ 0.900) in all three representative organs. Conclusion TANZANITE provides a flexible, image‐domain harmonization framework by learning the keV‐to‐keV translation in the VMI domain and applying pre‐trained CNNs to clinical kV images. It improves CT number consistency and organ‐specific radiomic reproducibility without requiring raw projection data or scanner‐specific training. This approach supports consistent quantitative imaging across multiple‐kV acquisition protocols, enhancing radiomics reliability in clinical settings.

A deep learning–based neural network enabled virtual monoenergetic imaging at energy-integrating detector CT, reduced blooming artifacts, and improved stenosis assessment in heavily calcified coronary plaques, with performance comparable to ultrahigh-resolution photon-counting detector CT.

Abstract Background Aortic valve calcification (AVC), as measured by gold-standard computed tomography (CT) Agatston score, provides an anatomic assessment of aortic stenosis (AS) severity and is a key predictor of AS progression and need for valve replacement. AVC detection and quantification from transthoracic echocardiography (TTE) could expand AS early diagnosis and risk stratification, currently limited by CT availability and radiation exposure. Methods A multi-view video-based deep learning framework was developed using 1166 TTE aortic valve videos from 187 TTE studies acquired in 110 patients with available AVC score by CT. EchoAVC architecture includes a feature extraction model followed by a quality-aware model that aggregates video-level information to obtain patient-level predictions for AVC detection and quantification (score). The framework was validated internally with 173 TTE studies from 86 patients across seven centres, and externally using 430 TTE studies from 280 patients across four different centres. The associations between EchoAVC estimations and AS severity, progression, and need for aortic valve replacement were examined. Results EchoAVC demonstrated excellent performance in AVC detection (AUROC 0.98, accuracy 94.4%), and quantification (R = 0.64) in the external multi-centre testing set. EchoAVC score was correlated with echocardiographic AS severity descriptors, including aortic valve mean pressure gradient (ρ = 0.749) and peak velocity (ρ = 0.757), and predicted future increase in mean pressure gradient (ρ = 0.382), peak velocity (ρ = 0.433) and calcium score by CT (ρ = 0.650). In 361 patients followed for a median of 3.8 years, 139 underwent aortic valve replacement. Baseline presence and extent of AVC as predicted by EchoAVC showed strong risk-stratification power for aortic valve replacement, remarkably in line with those obtained by CT, and incremental over TTE AS descriptors. EchoAVC was further tested in routine clinical practice images, confirming strong associations with AS severity and progression, including stratification for incident AS in previously unaffected individuals (p<0.001). Conclusions EchoAVC enables accurate and non-invasive detection and quantification of AVC, offering substantial diagnostic and prognostic value for aortic stenosis progression and need for valve replacement. This technique holds promise as a scalable tool for early detection and clinical management of aortic valve stenosis. Clinical Perspective What Is New? * Aortic valve calcification can be detected and quantified on standard, two-dimensional transthoracic echocardiography * EchoAVC score is associated with echocardiographic metrics of aortic stenosis severity * EchoAVC predicts progressive increase in aortic stenosis severity and need for aortic valve replacement What Are the Clinical Implications? * This openly available deep learning framework can be used to estimate aortic valve calcium in patients at risk of or presenting aortic stenosis to predict aortic stenosis progression * EchoAVC may help identify patients likely to have high aortic valve calcium and prioritize them for confirmatory CT assessment, thereby supporting earlier identification of calcific aortic valve disease in clinical pathways.

Background: Lung cancer is classified into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC – including adenocarcinoma, squamous cell, and large cell carcinoma). The incidence of major adverse cardiovascular events (MACE)—defined as acute myocardial infarction, acute coronary syndrome/ischemic heart disease requiring revascularization, stroke, and cardiovascular death—have not been well defined in lung cancer patients following treatment. Coronary artery calcifications (CAC) and aortic calcifications (CA) are known cardiovascular risk markers in non-cancer populations. Purpose: To determine whether the presence of CAC or CA, as identified on baseline PET-CT (Positron Emission Tomography–Computed Tomography) or chest CT imaging, is predictive of MACE and all-cause mortality (ACM) in patients with lung cancer. Methods: We conducted a retrospective cohort study of patients diagnosed with SCLC or NSCLC who received chemotherapy, between January 1, 2010, and December 31, 2024. The baseline PET-CT or chest CT scans were reviewed for the presence of CAC and/or CA. Patients were excluded if they lacked baseline imaging or had experienced a MACE before treatment. Univariate and multivariate adjusted Cox regression models were employed to assess associations between CAC/CA and MACE or ACM. The Kaplan-Meier survival analyses was performed to evaluate event-free survival. Results: A total of 385 patients were included (mean age: 74 years, IQR: 67–79), predominantly female (158 [55%]) and white (279 [96.2%]). NSCLC was found in 255 (88%) and SCLC in 35 (12%). CAC was identified in 172 (60%) and CA in 87 (30%). Among patients with NSCLC, the presence of CAC was significantly associated with MACE (HR: 2.84 [95% CI: 1.6–4.8], p=0.0001) (Fig. a), particularly in those with adenocarcinoma (Fig. b). In SCLC, (HR: 1.56 [95% CI: 0.1–14.0], p=ns). On multivariable analysis, CAC remained the strongest predictor of MACE after adjusting for age, hypertension, diabetes, dyslipidemia, and smoking status (HR: 2.6 [95% CI: 1.5–4.4], p=0.0008) (Fig. d). No significant association was found between CA and MACE in either group (p>0.05). ACM was similar for SCLC, squamous cell, and NSCLC (Fig. c). Conclusion: In patients with NSCLC especially lung adenocarcinoma, the presence of CAC in the baseline PET/CT was an independent predictor of MACE. This finding supports the value of CAC as a non-invasive cardiovascular risk marker.

Abstract Background To utilize high spatial resolution reconstructions for cardiac imaging at energy‐integrating detector CT (EID)‐CT with comparable noise to similar reconstructions at photon‐counting detector (PCD)‐CT, methods to control EID‐CT image noise are needed. Supervised convolutional neural networks (CNN) have shown promise for denoising, but a challenge remains to efficiently create high‐quality and unbiased estimates of noise without access to dedicated software or proprietary information, such that natural noise texture is retained in CNN‐denoised CT images. Purpose This study aims to develop and test image‐based noise estimation methods that can be used to train a CNN model, and to evaluate denoising performance and noise texture preservation for EID‐CT coronary CT angiography (cCTA) images reconstructed with high‐resolution kernels. Methods U‐net CNN models were trained for denoising. To supervise training, noise‐only images were estimated directly from high‐resolution kernel (Bv59) reconstructed EID‐CT (HR EID‐CT) patient images using two different methods: subtraction of low‐ and high‐strength iterative reconstruction (IR); subtraction of adjacent image slices with the same IR strength. The noise estimates from these methods contain differing noise texture and anatomical information. Networks were trained and validated separately for three data sets: the training data from each of the two noise‐estimation methods, and a 50%–50% partition of training data between the two methods. The trained models were applied to two sets of testing data: CT images of a uniform water phantom to measure noise power spectra (NPS), and an independent cohort of seven patient cCTA HR EID‐CT exams. The denoised patient images were compared to standard resolution EID‐CT reconstructions (Bv40). As a low‐noise reference, patient images acquired on the same day with a PCD‐CT and reconstructed using a similar kernel as HR EID‐CT were used for comparison. Results Models trained with each noise‐image estimation method denoised the HR EID‐CT images by 74%–79% to achieve a comparable noise magnitude to the HR PCD‐CT images. The peak, average, and 10% peak frequencies of the NPS of the input images (6.08, 6.24, and 12.0 cm −1 ) were better approximated by the model trained on adjacent slice subtraction (6.56, 5.87, and 11.5 cm −1 ) than by the model trained on subtraction of low‐ and high‐IR images (4.64, 5.44, and 11.3 cm −1 ). In cCTA images, the IR subtraction model images retained anatomic structures from input images but resulted in undesirable salt‐and‐pepper noise texture and CT number bias. The model trained on adjacent slice subtraction images had more natural texture and no significant bias, but the model sometimes removed small anatomic structures. The model trained on the mixed training dataset preserved both noise texture and anatomy from the model inputs and enabled visualization of small structures seen in PCD‐CT images that were previously unresolved by EID‐CT. Conclusions The noise texture and anatomical accuracy in CT images denoised with an image‐based supervised CNN are greatly influenced by the characteristics and partitioning of training data. With higher‐resolution reconstructions and noise texture‐preserving deep learning denoising, the quality of cCTA images from EID‐CT can be enhanced to enable resolvability of subtle anatomy similar to PCD‐CT.

Background Right ventricular (RV) maladaptation to elevated pulmonary afterload is the primary determinant of outcomes in pulmonary artery (PA) hypertension; however, the pathobiological mechanisms underlying RV decompensation remain poorly understood. Methods We performed global untargeted metabolomics on plasma from 55 patients who underwent gold‐standard RV‐PA coupling measurements using multibeat pressure volume loop assessment in a single‐center cohort and from 1027 patients with coupling surrogate measurements in a larger multicenter cohort, the PVDOMICS (Pulmonary Vascular Disease Phenomics) study. Age and sex‐adjusted linear regression was performed to identify associations between metabolites and coupling metrics. Additionally, we performed a metabolic flux analysis using gene expression data from RV tissue in an independent cohort of 32 patients. Partial least squares–discriminant analysis was used to identify metabolites and reactions characteristic of the decompensated RV. Results RV‐PA coupling was negatively associated with tricarboxylic acid (TCA) cycle intermediate levels. Specifically, plasma α‐ketoglutarate and fumarate were significantly associated with all coupling metrics in both cohorts. Metabolic flux analysis indicated that decompensated RVs exhibited aberrant TCA cycle activity, including reduced acetyl coenzyme A entry and increased lactate elimination, suggesting a shift from the TCA cycle toward glycolysis at the RV tissue level. Conclusions We identify an association between circulating TCA cycle intermediate levels and RV‐PA uncoupling in 2 independent cohorts, and dysregulated TCA cycle metabolism in decompensated PA hypertension RVs, suggesting that aberrant TCA cycle metabolism could represent a hallmark of RV maladaptation in PA hypertension. Further study of this pathway is warranted to develop novel biomarkers of RV function or RV‐targeted therapies.

Background: Major adverse cardiovascular events (MACE), including myocardial infarction, acute coronary syndrome, ischemic heart disease requiring revascularization, stroke, and heart failure, have been documented as significant contributors to mortality in cancer survivors. Coronary artery calcium (CAC) can predict MACE in non-cancer populations, while calcium in the aorta (CA) has not been evaluated as a prognostic marker. Every patient diagnosed with cancer undergoes a standard-of-care Positron Emission Tomography-Computed Tomography (PET-CT) or a chest CT before the initiation of chemotherapy. Hypothesis: To determine whether the CAC or CA of patients with Hodgkin's or non-Hodgkin's lymphoma, derived from standard-of-care PET-CT/chest CT, can predict the incidence of MACE. Methods: Patients treated with anthracycline-based chemotherapy, diagnosed and followed from January 1, 2013, through June 30, 2023, were included. Patients who did not undergo a PET-CT or CT, and/or developed MACE before treatment initiation were excluded. Univariate and multivariate adjusted Cox regression models were employed to assess whether the presence of CAC, CA, or CAC-CA was associated with the development of MACE. Calcium was retrospectively quantified using TeraRecon software (Durham NC) and categorized as: 0, 1-99, and >100. Outcome analyses was estimated using the Kaplan-Meier method. Results: 326 patients were included, mean age of 55 years (range: 52-60), predominantly male 201 (61%) and white 314 (96%), CAC was found in 89 patients and CA in 140. In the univariate regression model, a statistically significant association was found with values >100 for CA, CAC and CAC-CA with the risk of MACE. (Fig 1a/b/c). CAC equal to 0 demonstrated a significant protective effect against MACE. (Fig 1a). In the multivariable analysis, these associations persisted even after adjusting for comorbidities. (Table 1). Conclusion: CAC, CA and CAC/CA >100 in the standard-of-care CT/PET CT are predictors of MACE in lymphoma patients undergoing anthracycline treatment, a CAC equal to 0 has protective effect, these relationships remained statistically significant after adjusting for comorbidities.

Abstract Background Major adverse cardiovascular events (MACE), commonly defined as acute myocardial infarction, acute coronary syndrome/ischemic heart disease requiring revascularization, stroke, and heart failure, represent a significant cause of mortality among patients with cancer. With increased rates of both cancer and cancer survivorship, the identification of new predictive markers for the development of MACE in this population is crucial for implementing effective preventive and risk-mitigation strategies. Purpose To determine whether the presence of calcifications in the coronary arteries (CAC) or aorta (CA) of lymphoma patients, as reported in standard-of-care PET-CT (Positron Emission Tomography-Computed Tomography) or chest CT (Computed Tomography) scan, can predict the occurrence of MACE. Methods A Retrospective analysis was conducted on a cohort of patients diagnosed with Hodgkin or non-Hodgkin lymphoma and treated with anthracycline-based chemotherapy. Patients were diagnosed from January 1, 2013, and followed through to June 30, 2023. The reference point for the study was the radiologist’s report of the PET-CT or chest CT conducted before the initiation of anthracycline therapy. Patients who did not undergo a PET-CT or CT before the start of treatment, and/or developed MACE before treatment initiation were excluded. Univariate and multivariate adjusted Cox regression models were employed to assess whether the presence of CA and CAC was associated with the development of MACE. Associations with outcomes were evaluated using the Kaplan-Meier method and Cox regression. Results 326 patients were included with a mean age of 55 years (range: 52-60), predominantly male 201 (61%) and white 314 (96%), CAC was reported in 75 patients (23%) and CA in 18 (6%). In the univariate regression model, a statistically significant association was found between the presence of both CA and CAC with the risk of MACE. CAC showed a stronger association with the development of MACE than CA (HR: 3.7 [1-7.9], p=0.0005 vs HR: 2.9 [1- 8.5], p=0.050). In multivariable analysis CAC emerged as the strongest predictor of MACE (HR: 2.9 [1.3 – 6.4], p=0.01), after accounting for hypertension, diabetes, dyslipidemia, and GFR &lt;60 (Table 1), while the association with CA attenuated (HR: 2.01 [0.7-6.5], p=0.2010) and was no longer significant (Table 2). Conclusion CAC in the standard-of-care PET/-CT scans was a predictor of MACE in lymphoma patients treated with anthracyclines, while the association with CA weakened in multivariable analyses and was likely underpowered given the small number of patients with CA. If replicated, this finding warrants closer surveillance of this group of patients and can be considered as a non-invasive cardiovascular risk marker.

Abstract Background The relative contribution of calcific versus noncalcific aortic valve tissue volume to aortic stenosis (AS) in bicuspid aortic valve (BAV) is unclear. Purpose We aimed to assess the association of aortic valve tissue volume quantified by contrast-enhanced CT with echocardiographic AS severity, and diagnostic implication of noncalcific tissue volume for severe AS. Methods We retrospectively identified 234 patients with BAV who had echocardiographic AS severity and contrast-enhanced CT evaluation within six months and left ventricular ejection ≥50%, including 150 severe AS (aortic valve area≤1 cm2 or indexed valve area≤0.6cm2/m2 plus mean grdient≥40mmHg or peak velocity≥4 m/s), 84 nonsevere AS (61 with at least mild AS: peak veloicty≥2.0 m/s plus mean gradient ≥10mmHg, and 23 with normal aortic valve function). Contrast-enhanced CT aortic valve calcific volume and noncalcific (fibrosis) volume were quantified using patient-specific Hounsfield unit thresholds (Figure). Noncontrast CT aortic valve calcium score (AVCscore) was quantified by the Agaston method. CT measurements were indexed to echocardiographic aortic valve annulus area and square-root transformed for the analysis. Results Of 234 patients, the mean age was 64±11 years, and 64% were men. There were positive correlations of indexed calcific volume, noncalcific volume, and inverse correlations of percentages of noncalcific volume (noncalcific volume divided by the sum of calcific and noncalcific volume) with peak velocity and mean gradient (|r|=0.30 to 0.68). Among 150 patients with severe AS, 28 had an AVCscore below the severe AVC threshold (1200 AU in women and 2000 AU in men) while exhibited a higher percentage of noncalcific volume compared to 122 patients with severe AVCscore (67±15% vs. 39±15%, p&lt;0.001). After excluding 122 patients with severe AS and severe AVCscore, the indexed noncalcific volume was associated with severe AS (β±S.E: 0.18±0.07, p=0.005) with an area under the curve of 0.71; in contrast, AVCscore was not (p=0.55). Conclusion Both calcific and noncalcific aortic valve tissue volume contribute to AS severity in patients with BAV. Notably, noncalcific tissue volume is a key factor in determining severe AS in patients without severe AVCscore, suggesting its potential diagnostic value.Figure

Abstract Introduction Aortic stenosis (AS) presents diagnostic challenges, especially in patients with discordant low gradient AS (LGAS), i.e., small aortic valve area (AVA) and low Doppler mean gradient (MG). LGAS is associated with worse outcomes compared to concordant AS. Aortic valve calcium score (AVCS) is a measure of AS severity independent of flow and helps diagnose severe AS. It is unknown if the degree of discordance in AVCS and MG is of any clinical significance. A high AVCS/MG ratio may indicate impaired ventricular function and response to pressure overload despite preserved LVEF, potentially contributing to worse outcomes. Purpose This study aims to elucidate the prognostic significance of the AVCS/MG ratio in patients with severe AS and preserved left ventricular ejection fraction (LVEF) undergoing TAVR. We hypothesized a higher AVCS/MG ratio correlates with increased mortality. Methods Patients who underwent transcatheter aortic valve replacement (TAVR) from 2010 to 2020 for native valve severe AS with LVEF≥50% were identified from an institutional database. AVCS were obtained within 90 days of pre-TAVR echocardiography. Patients were divided into tertiles based on their AVCS/MG ratio. Primary outcome was overall mortality. Results The analysis included 1067 patients (mean age 79±9 years, 43% female); overall AVA was 0.80±0.14 cm², LVEF 63±6%, MG 46±12 mmHg, AVCS 2630±1528 AU, and AVCS/MG ratio 58±31. Patients stratified by AVCS/MG tertiles showed no difference in overall comorbidity index across tertiles (Figure 1). However, the group in the highest tertile was the oldest (80±8 vs 79±8 vs 77±9 years), predominantly male (87% vs 61% vs 24%), with the highest prevalent atrial fibrillation (44% vs 28% vs 25%) and stroke volume index &lt;35 mL/m² (16% vs 11% vs 5%). 9-year survival was 15% vs 25% vs 36% (p&lt;0.001, Figure 2). In multivariate analysis adjusted for age, sex, comorbidities, LVEF and AVA, AVCS/MG ratio was independently associated with increased risk of mortality (HR 1.005; 95% CI [1.001-1.008], p=0.005). Conclusion Higher AVCS/MG discordance is associated with increased risk of mortality following TAVR. Serial AVCS measurements during progressive AS may help identify patients at higher risk who may potentially benefit from earlier aortic valve intervention.