Differentiation of left atrial appendage thrombus from circulatory stasis using cardiac CT radiomics in patients with valvular heart disease.

OBJECTIVES: To determine whether quantitative radiomic features from cardiac CT could differentiate the left atrial appendage (LAA) thrombus from circulatory stasis in patients with valvular heart disease. METHODS: Ninety-five consecutive patients with valvular heart disease and filling defects in LAA on two-phase cardiac CT from March 2016 to August 2018 were retrospectively enrolled and classified as having thrombus or stasis by transesophageal echocardiography or cardiac surgery. The ratio of Hounsfield units in the filling defects to those in the ascending aorta (AA) was calculated on early- and late-phase CT (LAA/AAE and LAA/AAL, respectively). Radiomic features were extracted from semi-automated three-dimensional segmentation of the filling defect on early-phase CT. The diagnostic ability of radiomic features for differentiating thrombus from stasis was assessed and compared to LAA/AAE and LAA/AAL by comparing the AUC of ROC curves. Diagnostic performances of CT attenuation ratios and radiomic features were validated with an independent validation set. RESULTS: Thrombus was diagnosed in 25 cases and stasis in 70. Sixty-eight radiomic features were extracted. Values of 8 wavelet-transformed features were lower in thrombus than in stasis (p < 0.001). The AUC value of a radiomic feature, wavelet_LHL, for diagnosing thrombus was 0.78, which was higher than that of LAA/AAE (AUC = 0.54, p = 0.025) and similar to that of LAA/AAL (AUC = 0.76, p = 0.773). In the validation set, the AUC of wavelet_LHL was 0.71, which was higher than that of LAA/AAE (AUC = 0.57, p = 0.391) and similar to that of LAA/AAL (AUC = 0.75, p = 0.707). CONCLUSIONS: Quantitative radiomic features from the early phase of cardiac CT may help diagnose LAA thrombus in patients with valvular heart disease. KEY POINTS: • Wavelet-transformed grey-level non-uniformity values from radiomic analysis are significantly lower for LAA thrombus than for circulatory stasis. • Radiomic features may have an additional value for differentiating LAA thrombus from circulatory stasis when interpreting single-phase cardiac CT. • Radiomic features extracted from single-phase images may show similar diagnostic ability as conventional quantitative analysis from two-phase images.

Diagnosis of thyroid micronodules on ultrasound using a deep convolutional neural network.

To assess the performance of deep convolutional neural network (CNN) to discriminate malignant and benign thyroid nodules < 10 mm in size and compare the diagnostic performance of CNN with those of radiologists. Computer-aided diagnosis was implemented with CNN and trained using ultrasound (US) images of 13,560 nodules ≥ 10 mm in size. Between March 2016 and February 2018, US images of nodules < 10 mm were retrospectively collected at the same institution. All nodules were confirmed as malignant or benign from aspirate cytology or surgical histology. Diagnostic performances of CNN and radiologists were assessed and compared for area under curve (AUC), sensitivity, specificity, accuracy, positive predictive value, and negative predictive value. Subgroup analyses were performed based on nodule size with a cut-off value of 5 mm. Categorization performances of CNN and radiologists were also compared. A total of 370 nodules from 362 consecutive patients were assessed. CNN showed higher negative predictive value (35.3% vs. 22.6%, P = 0.048) and AUC (0.66 vs. 0.57, P = 0.04) than radiologists. CNN also showed better categorization performance than radiologists. In the subgroup of nodules ≤ 5 mm, CNN showed higher AUC (0.63 vs. 0.51, P = 0.08) and specificity (68.2% vs. 9.1%, P < 0.001) than radiologists. Convolutional neural network trained with thyroid nodules ≥ 10 mm in size showed overall better diagnostic performance than radiologists in the diagnosis and categorization of thyroid nodules < 10 mm, especially in nodules ≤ 5 mm.

Balloon Bronchoplasty for the Treatment of Bronchial Stenosis After Lung Transplantation: A Single-Center 10-Year Experience.

OBJECTIVE: To assess the safety and efficacy of balloon dilatation under dual guidance using fluoroscopy and bronchoscopy for treating bronchial stenosis following lung transplantation (LT), and to elucidate the factors associated with patency after the procedure. MATERIALS AND METHODS: From September, 2012, to April, 2021, 50 patients (mean age ± standard deviation, 54.4 ± 12.2 years) with bronchial stenosis among 361 recipients of LT were retrospectively analyzed. The safety of balloon dilatation was assessed by evaluating procedure-related complications. Efficacy was assessed by evaluating the technical success, primary patency, and secondary patency. Primary and secondary cumulative patency rates were calculated using the Kaplan-Meier method. The factors associated with patency after the procedure were evaluated using multivariable Cox hazard proportional regression analysis. RESULTS: In total, 65 bronchi were treated with balloon dilatation in 50 patients. The total number of treatment sessions was 277 and the technical success rate was 99.3% (275/277 sessions). No major procedure-related complications were noted. During the mean follow-up period of 34.6 ± 30.8 months, primary patency was achieved in 12 of 65 bronchi (18.5%). However, the patency rate improved to 76.9% (50 of 65 bronchi) after repeated balloon dilatation (secondary patency). The 6-month, 1-year, 3-year, and 5-year secondary patency rates were 95.4%, 90.8%, 83.1%, and 78.5%, respectively. The presence of clinical symptoms was a significant prognostic factor associated with reduced primary patency (adjusted hazard ratio [HR], 0.465; 95% confidence interval [CI], 0.220-0.987). Early-stage treatment ≤ 6 months (adjusted HR, 3.588; 95% CI, 1.093-11.780) and prolonged balloon dilatation > 5 min (adjusted HR, 3.285; 95% CI, 1.018-10.598) were associated with significantly higher secondary patency. CONCLUSION: Repeated balloon dilatation was determined to be safe and effective for treating bronchial stenosis following LT. Early-stage treatment and prolonged balloon dilatation could significantly promote long-term patency.

Deep learning-based reconstruction on cardiac CT yields distinct radiomic features compared to iterative and filtered back projection reconstructions.

We aimed to determine the effects of deep learning-based reconstruction (DLR) on radiomic features obtained from cardiac computed tomography (CT) by comparing with iterative reconstruction (IR), and filtered back projection (FBP). A total of 284 consecutive patients with 285 cardiac CT scans that were reconstructed with DLR, IR, and FBP, were retrospectively enrolled. Radiomic features were extracted from the left ventricular (LV) myocardium, and from the periprosthetic mass if patients had cardiac valve replacement. Radiomic features of LV myocardium from each reconstruction were compared using a fitting linear mixed model. Radiomics models were developed to diagnose periprosthetic abnormality, and the performance was evaluated using the area under the receiver characteristics curve (AUC). Most radiomic features of LV myocardium (73 of 88) were significantly different in pairwise comparisons between all three reconstruction methods (P < 0.05). The radiomics model on IR exhibited the best diagnostic performance (AUC 0.948, 95% CI 0.880-1), relative to DLR (AUC 0.873, 95% CI 0.735-1) and FBP (AUC 0.875, 95% CI 0.731-1), but these differences did not reach significance (P > 0.05). In conclusion, applying DLR to cardiac CT scans yields radiomic features distinct from those obtained with IR and FBP, implying that feature robustness is not guaranteed when applying DLR.