Development and validation of a deep learning signature for predicting lymph node metastasis in lung adenocarcinoma: comparison with radiomics signature and clinical-semantic model.

OBJECTIVE: To develop and validate a deep learning (DL) signature for predicting lymph node (LN) metastasis in patients with lung adenocarcinoma. METHODS: A total of 612 patients with pathologically-confirmed lung adenocarcinoma were retrospectively enrolled and were randomly divided into training cohort (n = 489) and internal validation cohort (n = 123). Besides, 108 patients were enrolled and constituted an independent test cohort (n = 108). Patients' clinical characteristics and CT semantic features were collected. The radiomics features were derived from contrast-enhanced CT images. The clinical-semantic model and radiomics signature were built to predict LN metastasis. Furthermore, Swin Transformer was adopted to develop a DL signature predictive of LN metastasis. Model performance was evaluated by area under the receiver operating characteristic curve (AUC), sensitivity, specificity, calibration curve, and decision curve analysis. The comparisons of AUC were conducted by the DeLong test. RESULTS: The proposed DL signature yielded an AUC of 0.948-0.961 across all three cohorts, significantly superior to both clinical-semantic model and radiomics signature (all p < 0.05). The calibration curves show that DL signature predicted probabilities fit well the actual observed probabilities of LN metastasis. DL signature gained a higher net benefit than both clinical-semantic model and radiomics signature. The incorporation of radiomics signature or clinical-semantic risk predictors failed to reveal an incremental value over the DL signature. CONCLUSIONS: The proposed DL signature based on Swin Transformer achieved a promising performance in predicting LN metastasis and could confer important information in noninvasive mediastinal LN staging and individualized therapeutic options. KEY POINTS: • Accurate prediction for lymph node metastasis is crucial to formulate individualized therapeutic options for patients with lung adenocarcinoma. • The deep learning signature yielded an AUC of 0.948-0.961 across all three cohorts in predicting lymph node metastasis, superior to both radiomics signature and clinical-semantic model. • The incorporation of radiomics signature or clinical-semantic risk predictors into deep learning signature failed to reveal an incremental value over deep learning signature.

Subtraction Improves the Accuracy of Coronary CT Angiography in Patients with Severe Calcifications in Identifying Moderate and Severe Stenosis: A Multicenter Study.

RATIONALE AND OBJECTIVES: To investigate the diagnostic accuracy of subtraction coronary computed tomographic angiography (CCTAsub) in identifying ≥ 50% and ≥ 70% coronary stenosis in patients with different degrees of calcification. MATERIALS AND METHODS: In this study, 180 patients with coronary calcified plaques who underwent both coronary CT angiography and invasive coronary angiography (ICA) were prospectively enrolled at five centers. Patients were divided into three groups according to the Agatston score: group A (low to moderate, < 400), group B (high, 400-999), and group C (very high, ≥ 1000). Diagnostic accuracies estimated by area under the receiver operating characteristic curve (AUC) were compared between conventional CCTA (CCTAcon) and CCTAsub, with ICA as a reference standard. RESULTS: There were 86 patients in group A, 44 in group B, and 50 in group C. In identifying ≥ 70% coronary stenosis, subtraction improved the diagnostic accuracies on a per-segment basis in group B (AUC: 0.80 vs 0.92, p = 0.001) and group C (AUC: 0.75 vs 0.84, p = 0.001) after subtraction. When identifying ≥ 50% coronary stenosis, the per-segment AUC of CCTAsub in group B and C were significantly higher than that in CCTAcon (group B: 0.81 vs 0.92, p < 0.001; group C: 0.77 vs 0.88, p < 0.001). However, no improvement was observed in group A. CONCLUSION: Subtraction achieved better diagnostic accuracy in patients with Agatston score ≥ 400, both in identifying ≥ 50% and ≥ 70% coronary stenosis, which was instructive for the application of subtraction in clinical practice.

Feasibility of Subtraction Coronary Computed Tomographic Angiography and Influencing Factor Analysis: a Retrospective Study.

OBJECTIVE: To investigate the feasibility of subtraction coronary computed tomographic (CT) angiography (SubCCTA) to decline calcium artifacts and improve diagnostic accuracy in the presence of coronary calcification and analyze the factors that influence SubCCTA. METHODS: A total of 294 patients suspected of having coronary artery diseases underwent coronary computed tomographic angiography (CCTA) and SubCCTA. Coronary stenoses were blindly evaluated by two experienced radiologists, which were compared with invasive coronary angiography (ICA). Multiple statistical indexes were adopted to analyze the value of SubCCTA for the diagnosis of calcium stenoses. RESULTS: The diagnosable rate of SubCCTA was 67.2% (n=197), and the non-diagnosable rate was 32.8% (n=97). Using SubCCTA, the false positive rate decreased from 56.5% to 17.4%, and the corresponding diagnostic accuracy was increased from 83.6% to 92.9%. Univariate logistic regression analysis showed that height (OR=1.029, 95% CI=1.001-1.058), weight (OR=1.025, 95% CI=1.004-1.046), left ventricular size (OR=1.018, 95% CI=1.007-1.030), cardiothoracic ratio (OR=39.917, 95% CI=1.244-1281.098), the average heart rate (OR=0.866, 95% CI=0.836-0.896) and heart rate range (OR=0.882, 95% CI=0.853-0.912) might be the factors influencing SubCCTA. CONCLUSION: This study suggested that SubCCTA could help improve diagnostic accuracy in the presence of calcium plaques. Moreover, several factors were discovered for the first time to possibly influence SubCCTA, which will be helpful in improving the subtracted image quality.

MR Virtual Endoscopy of the Fetal Limb Anomalies Using Three-Dimensional Fast Imaging Employing Steady-State Acquisition Sequence.

OBJECTIVE: To retrospectively investigate the feasibility of magnetic resonance virtual endoscopy (MRVE) to visualize the normal limbs and limb deformities Methods: MR sequences included two-dimensional (2D) single fast spin-echo sequence and 2D and 3D steady-state procession fast imaging sequences. MRVE reconstruction was retrospectively performed by 2 radiologists in 32 fetuses in 30 pregnant women. The correlation between the radiologists for the virtual endoscopy threshold of MRVE was determined. Image quality and limb segment visibility were independently rated. Area under the receiver operating characteristics curve (AUC) of 2D MRI and MRVE was calculated. RESULTS: The mean virtual endoscopy threshold required for the visualization of the limb was 991.93 ± 12.13 and 991.83 ± 12.26 for 2 radiologists, respectively. The correlation between the radiologists for virtual endoscopy threshold was excellent (r = 0.933). The weighted kappa statistic was 0.96 for the evaluation of image quality of limb segments, indicating excellent interobserver agreement. Compared to that of 2D MRI alone, a higher AUC of 2D MRI with MRVE was achieved in detection of both upper and lower limb deformities (0.91 vs. 0.69 and 0.83 vs. 0.71, respectively). CONCLUSION: MRVE may display normal and abnormal fetal limb orientation and structures from multiple perspectives and provide incremental information for obstetrics.

Preliminary study on identification of estrogen receptor-positive breast cancer subtypes based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) texture analysis.

BACKGROUND: Currently, breast cancer is divided into Luminal A, Luminal B, HER-2 overexpression (HER-2) and basal cell at genetic level. However, the differential diagnosis of estrogen receptor (ER)-positive breast cancer subtypes is rare. Therefore, we aimed to investigate the feasibility of identifying the ER-positive breast cancer subtypes based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) texture analysis. METHODS: A retrospective analysis was performed for clinical data of 51 patients with ER-positive breast invasive ductal carcinoma confirmed by surgery and pathology from January 20 to October 2018. FireVoxel texture analysis software was used to delineate the tumor boundary layer by layer. The differences in the above characteristics between Luminal A and Luminal B breast cancer were compared, and the diagnostic efficacy of statistically significant texture parameters for ER-positive breast cancer subtypes was analyzed. RESULTS: There were no significant differences in mean, standard deviation (SD), skewness and tumor size between Luminal A and Luminal B groups (P>0.05). The kurtosis, inhomogeneity and entropy could effectively distinguish between the two groups with statistically significant difference (P=0.001, P=0.000, and P=0.000). The area under the receiver operating characteristic (ROC) curve (AUC) of kurtosis, inhomogeneity and entropy diagnosed with malignant mass were 0.832, 0.859 and 0.891, respectively (P<0.01). In addition, the entropy was the best among the three indicators. When the entropy was ≤4.22, the sensitivity of the diagnosis Luminal B was 90.62% and the specificity was 78.95%. CONCLUSIONS: The texture analysis features based on DCE-MRI can help to identify ER-positive breast cancer subtypes. Entropy can be the best single texture indicator.