Machine learning-based radiomics analysis for predicting local recurrence of primary dermatofibrosarcoma protuberans after surgical treatment.

BACKGROUND AND PURPOSE: Dermatofibrosarcoma protuberans (DFSP) is characterized by locally invasive growth patterns and high local recurrence rates. Accurately identifying patients with high local recurrence risk may benefit patients during follow-up and has potential value for making treatment decisions. This study aimed to investigate whether machine learning-based radiomics models could accurately predict the local recurrence of primary DFSP after surgical treatment. MATERIALS AND METHODS: This retrospective study included a total of 146 patients with DFSP who underwent MRI scans between 2010 and 2016 from two different institutions: institution 1 (n = 104) for the training set and institution 2 (n = 42) for the external test set. Three radiomics random survival forest (RSF) models were developed using MRI images. Additionally, the performance of the Ki67 index was compared with the three RSF models in the external validation set. RESULTS: The average concordance index (C-index) scores of the RSF models based on fat-saturation T2W (FS-T2W) images, fat-saturation T1W with gadolinium contrast (FS-T1W + C) images, and both FS-T2W and FS-T1W + C images from 10-fold cross-validation in the training set were 0.855 (95% CI: 0.629, 1.00), 0.873 (95% CI: 0.711, 1.00), and 0.875 (95% CI: 0.688, 1.00), respectively. In the external validation set, the C-indexes of the three trained RSF models were higher than that of the Ki67 index (0.838, 0.754, and 0.866 vs. 0.601, respectively). CONCLUSION: Random survival forest models developed using radiomics features derived from MRI images were proven helpful for accurate prediction of local recurrence of primary DFSP after surgical treatment and showed better predicting performance than the Ki67 index.

Machine learning-based prediction of invisible intraprostatic prostate cancer lesions on 68 Ga-PSMA-11 PET/CT in patients with primary prostate cancer.

PURPOSE: 68 Ga-PSMA PET/CT has high specificity and sensitivity for the detection of both intraprostatic tumor focal lesions and metastasis. However, approximately 10% of primary prostate cancer are invisible on PSMA-PET (exhibit no or minimal uptake). In this work, we investigated whether machine learning-based radiomics models derived from PSMA-PET images could predict invisible intraprostatic lesions on 68 Ga-PSMA-11 PET in patients with primary prostate cancer. METHODS: In this retrospective study, patients with or without prostate cancer who underwent 68 Ga-PSMA PET/CT and presented negative on PSMA-PET image at either of two different institutions were included: institution 1 (between 2017 and 2020) for the training set and institution 2 (between 2019 and 2020) for the external test set. Three random forest (RF) models were built using selected features extracted from standard PET images, delayed PET images, and both standard and delayed PET images. Then, subsequent tenfold cross-validation was performed. In the test phase, the three RF models and PSA density (PSAD, cut-off value: 0.15 ng/ml/ml) were tested with the external test set. The area under the receiver operating characteristic curve (AUC) was calculated for the models and PSAD. The AUCs of the radiomics model and PSAD were compared. RESULTS: A total of 64 patients (39 with prostate cancer and 25 with benign prostate disease) were in the training set, and 36 (21 with prostate cancer and 15 with benign prostate disease) were in the test set. The average AUCs of the three RF models from tenfold cross-validation were 0.87 (95% CI: 0.72, 1.00), 0.86 (95% CI: 0.63, 1.00), and 0.91 (95% CI: 0.69, 1.00), respectively. In the test set, the AUCs of the three trained RF models and PSAD were 0.903 (95% CI: 0.830, 0.975), 0.856 (95% CI: 0.748, 0.964), 0.925 (95% CI:0.838, 1.00), and 0.662 (95% CI: 0.510, 0.813). The AUCs of the three radiomics models were higher than that of PSAD (0.903, 0.856, and 0.925 vs. 0.662, respectively; P = .007, P = .045, and P = .005, respectively). CONCLUSION: Random forest models developed by 68 Ga-PSMA-11 PET-based radiomics features were proven useful for accurate prediction of invisible intraprostatic lesion on 68 Ga-PSMA-11 PET in patients with primary prostate cancer and showed better diagnostic performance compared with PSAD.

Assessment of quantitative dynamic contrast-enhanced MRI in distinguishing different histologic grades of breast phyllode tumor.

OBJECTIVES: To investigate whether quantitative DCE-MRI (qDCE-MRI) could help distinguish breast phyllodes tumor (PT) grades. MATERIALS AND METHODS: This retrospective study included 67 breast PTs (26 benign lesions, 25 borderline lesions, and 16 malignant lesions) from April 2016 to July 2020. MRI was performed with a 1.5-T MR system. Perfusion parameters (Ktrans, kep, ve, iAUC60) derived from qDCE-MRI, tumor size, and the mean ADC value were correlated with histologic grades using Spearman's rank correlation coefficient. Ktrans, kep, ve, and iAUC60 of three histologic grades were also calculated and compared. RESULTS: The Spearman correlation coefficient with histologic grade of the tumor size was 0.578 (p < 0.001); the ADC value was not correlated with histologic grades of breast PT (p = 0.059). The Ktrans, kep, ve, and iAUC60 of benign breast PTs were significantly lower than those of borderline breast PTs (p < 0.001) and lower than those of malignant breast PTs (p < 0.001). In comparison, the Ktrans, ve, and iAUC60 of borderline breast PTs were significantly lower than those of malignant breast PTs (p < 0.001, p < 0.001, p = 0.007, respectively). For ROC analysis, AUCs of Ktrans, ve, and iAUC60 were higher than tumor size and ADC value for differentiating three PT grades. CONCLUSION: Quantitative and semi-quantitative perfusion parameters (Ktrans, ve, and iAUC60, especially Ktrans) derived from qDCE-MRI showed better diagnosis efficiency than tumor size and ADC for grading breast PTs. Therefore, qDCE-MRI may be helpful for preoperative differentiating breast PT grades. KEY POINTS: • Quantitative dynamic contrast-enhanced MRI can be used as a complementary noninvasive method to improve the differential diagnosis of breast PT. • Ktrans, ve, and iAUC60 derived from qDCE-MRI showed better diagnosis efficiency than tumor size and ADC for grading breast PTs.

Plasma D-dimer as a predictor of intraluminal thrombus burden and progression of abdominal aortic aneurysm.

AIM: Intraluminal thrombus (ILT) is presented in most abdominal aortic aneurysms (AAAs) and is suggested to promote AAA expansion. D-dimer, a breakdown product in the thrombus remodeling, may have prognostic value for AAA. This study investigated the interrelation between plasma D-dimer level, ILT volume, AAA size and progression. MAIN METHODS: This was a retrospective observational study that involved 181 patients with infra-renal AAA. They were divided into small and large AAA groups according to AAA diameter. 24 of them had repeated abdominal computed tomography angiography (CTA) scan and were divided into slow-growing and fast-growing AAA groups according to the median value of AAA growth rate. Baseline and follow-up plasma D-dimer level, maximum diameter of AAA, total infra-renal aortic volume and ILT volume were analyzed. KEY FINDINGS: Plasma D-dimer level was positively correlated with ILT volume (R = 0.382, P < 0.001) and maximum diameter of AAA (R = 0.442, P < 0.001). Increasing value of plasma D-dimer was positively associated with the accelerated growth rate of AAA (R = 0.720, P < 0.01). ILT volume showed positive correlation with maximum diameter (R = 0.859, P < 0.001) and growth rate of AAA (R = 0.490, P < 0.05). After adjusting the baseline ILT volume, the positive correlations remained to be statistically significant between plasma D-dimer level and AAA size (R = 0.200, P < 0.05), as well as increasing value of plasma D-dimer and growth rate of AAA (R = 0.642, P < 0.05). SIGNIFICANCE: Plasma D-dimer level reflected ILT burden in AAAs. Plasma D-dimer level and ILT volume were positively correlated with AAA size. Increasing value of plasma D-dimer and baseline ILT volume could be predictors of AAA progression.

Discrimination between benign and malignant breast lesions using volumetric quantitative dynamic contrast-enhanced MR imaging.

OBJECTIVE: To determine the diagnostic performance of volumetric quantitative dynamic contrast-enhanced MRI (qDCE-MRI) in differentiation between malignant and benign breast lesions. METHODS: DCE-MRI was performed in 124 patients with 136 breast lesions. Quantitative pharmacokinetic parameters Ktrans, Kep, Ve, Vp and semi-quantitative parameters TTP, MaxCon, MaxSlope, AUC were obtained by using a two-compartment extended Tofts model and three-dimensional volume of interest. Morphologic features (lesion size, margin, internal enhancement pattern) and time-signal intensity curve (TIC) type were also assessed. Logistic regression analysis was used to determine predictors of malignancy, followed by receiver operating characteristics (ROC) analysis to evaluate the diagnostic performance. RESULTS: qDCE parameters (Ktrans, Kep, Vp, TTP, MaxCon, MaxSlope and AUC), morphological parameters and TIC type were significantly different between malignant and benign lesions (P≤0.001). Multivariate logistic regression analyses showed that Ktrans, Kep, MaxSlope, size, margin and TIC type were independent predictors of malignancy. The diagnostic accuracy of logistic models based on qDCE parameters alone, morphological features plus TIC type, and all parameters combined was 94.9%, 89.0%, and 95.6% respectively. CONCLUSION: qDCE-MRI can be used to improve diagnostic differentiation between benign and malignant breast lesions in relation to morphology and kinetic analysis. KEY POINTS: • qDCE-MRI parameters are useful for discriminating between malignant and benign breast lesions. • K trans , K ep and MaxSlope were independent predictors of breast malignancy. • qDCE-MRI has a better diagnostic ability than morphology and kinetic analysis. • qDCE-MRI can be used to improve the diagnostic accuracy of breast malignancy.

Assessment of Nonalcoholic Fatty Liver Disease in Rats Using Quantitative Dynamic Contrast-Enhanced MRI.

PURPOSE: To evaluate the usefulness of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in the assessment of nonalcoholic fatty liver disease (NAFLD) severity. MATERIALS AND METHODS: Liver DCE-MRI at 3.0T was performed in 36 adult Sprague-Dawley rats with methionine choline-deficient diet-induced NAFLD and 10 untreated control rats. Pharmacokinetic parameters of DCE-MRI including Ktrans , Kep , Ve , Vp , and hepatic portal index (HPoI) were measured using the dual-input extended Tofts model. Animals were categorized as normal (n = 10), simple steatosis (SS, n = 11), borderline nonalcoholic steatohepatitis (bNASH, n = 20), and NASH (n = 5) subgroups according to the NAFLD activity score system, and classified into F0 (n = 24), F1 (n = 11), F2 (n = 7), and F3 (n = 4) subgroups according to an established scoring system. DCE-MRI parameters were compared. Receiver operating characteristic analyses were performed to assess the diagnostic performance of various DCE-MRI parameters in grading NAFLD activity and staging liver fibrosis. RESULTS: Ktrans and HPoI were elevated with increasing severity of NAFLD activity and increased fibrosis stage. The areas under the receiver operating characteristic curve (AUROCs) of HPoI ranged from 0.895-0.951 for discriminating between different grades of NAFLD activity, and the AUROC was 0.852 for discriminating F0 stage from overall F1-F3 stages. The AUROC of Ktrans for discriminating non-NASH from bNASH and NASH groups was 0.968, and 0.898 for discriminating between normal and overall fibrosis groups. CONCLUSION: DCE-MRI may play a role in assessing NAFLD severity. LEVEL OF EVIDENCE: 1 J. MAGN. RESON. IMAGING 2017;45:1485-1493.