Characterization of Breast Tumors from MR Images Using Radiomics and Machine Learning Approaches.

Determining histological subtypes, such as invasive ductal and invasive lobular carcinomas (IDCs and ILCs) and immunohistochemical markers, such as estrogen response (ER), progesterone response (PR), and the HER2 protein status is important in planning breast cancer treatment. MRI-based radiomic analysis is emerging as a non-invasive substitute for biopsy to determine these signatures. We explore the effectiveness of radiomics-based and CNN (convolutional neural network)-based classification models to this end. T1-weighted dynamic contrast-enhanced, contrast-subtracted T1, and T2-weighted MR images of 429 breast cancer tumors from 323 patients are used. Various combinations of input data and classification schemes are applied for ER+ vs. ER-, PR+ vs. PR-, HER2+ vs. HER2-, and IDC vs. ILC classification tasks. The best results were obtained for the ER+ vs. ER- and IDC vs. ILC classification tasks, with their respective AUCs reaching 0.78 and 0.73 on test data. The results with multi-contrast input data were generally better than the mono-contrast alone. The radiomics and CNN-based approaches generally exhibited comparable results. ER and IDC/ILC classification results were promising. PR and HER2 classifications need further investigation through a larger dataset. Better results by using multi-contrast data might indicate that multi-parametric quantitative MRI could be used to achieve more reliable classifiers.

Utility of Wavelet Denoising with Geometry Factor Weighting for Gadoxetic Acid-enhanced Hepatobiliary-phase MR Imaging.

PURPOSE: The wavelet denoising with geometry factor weighting (g-denoising) method can reduce the image noise by adapting to spatially varying noise levels induced by parallel imaging. The aim of this study was to investigate the clinical applicability of g-denoising on hepatobiliary-phase (HBP) images with gadoxetic acid. METHODS: We subjected 53 patients suspected of harboring hepatic neoplastic lesions to gadoxetic acid-enhanced HBP imaging with and without g-denoising (g+HBP and g-HBP). The matrix size was reduced for g+HBP images to avoid prolonging the scanning time. Two radiologists calculated the SNR, the portal vein-, and paraspinal muscle contrast-to-noise ratio (CNR) relative to the hepatic parenchyma (liver-to-portal vein- and liver-to-muscle CNR). Two other radiologists independently graded the sharpness of the liver edge, the visibility of intrahepatic vessels, the image noise, the homogeneity of liver parenchyma, and the overall image quality using a 5-point scale. Differences between g-HBP and g+HBP images were determined with the two-sided Wilcoxon signed-rank test. RESULTS: The liver-to-portal- and liver-to-muscle CNR and the SNR were significantly higher on g+HBP- than g-HBP images (P < 0.01), as was the qualitative score for the image noise, homogeneity of liver parenchyma, and overall image quality (P < 0.01). Although there were no significant differences in the scores for the sharpness of the liver edge or the score assigned for the visibility of intrahepatic vessels (P = 0.05, 0.43), with g+HBP the score was lower in three patients for the sharpness of the liver edge and in six patients for the visibility of intrahepatic vessels. CONCLUSION: At gadoxetic acid-enhanced HBP imaging, g-denoising yielded a better image quality than conventional HBP imaging although the anatomic details may be degraded.

Quantitative parameter mapping of contrast agent concentration and relaxivity and brain tumor extracellular pH.

In clinical magnetic resonance imaging, gadolinium-based contrast agents are commonly used for detecting brain tumors and evaluating the extent of malignancy. We present a new method to evaluate relaxivity (r1) and contrast agent concentration separately in contrast-enhanced lesions using quantitative parameter mapping (QPM). Furthermore, we also aimed to estimate the extracellular pH (pHe) of tumor lesions. We demonstrated that it is possible to evaluate pathophysiological tumor changes due to therapeutic efficacy by measuring r1 in contrast-enhanced lesions. In this study, the primary brain tumor group showed significantly higher r1 values than other brain disease groups (P < 0.001). Moreover, mean pHe value showed a trend for tumor malignancy having a lower pHe value and primary brain tumor having a significantly lower pHe than other brain diseases (P < 0.001). Our results might suggest that QPM can separately quantify r1 and CA concentration in brain tumors and that pHe brain tumor mapping could serve as a tumor biomarker. In conclusion, our method has potential clinical applications for assessing the treatment effects.

Quantitative Susceptibility Mapping versus R2*-based Histogram Analysis for Evaluating Liver Fibrosis: Preliminary Results.

PURPOSE: The staging of liver fibrosis is clinically important, and a less invasive method is preferred. Quantitative susceptibility mapping (QSM) has shown a great potential in estimating liver fibrosis in addition to R2* relaxometry. However, few studies have compared QSM analysis and liver fibrosis. We aimed to evaluate the feasibility of estimating liver fibrosis by using QSM and R2*-based histogram analyses by comparing it with ultrasound-based transient elastography and the stage of histologic fibrosis. METHODS: Fourteen patients with liver disease were enrolled. Data sets of multi-echo gradient echo sequence with breath-holding were acquired on a 3-Tesla scanner. QSM and R2* were reconstructed by water-fat separation method, and ROIs were analyzed for these images. Quantitative parameters with histogram features (mean, variance, skewness, kurtosis, and 1st, 10th, 50th, 90th, and 99th percentiles) were extracted. These data were compared with the elasticity measured by ultrasound transient elastography and histological stage of liver fibrosis (F0 to F4, based on the new Inuyama classification) determined by biopsy or hepatectomy. The correlation of histogram parameters with intrahepatic elasticity and histologically confirmed fibrosis stage was examined. Texture parameters were compared between subgroups divided according to fibrosis stage. Receiver operating characteristic (ROC) analysis was also performed. P < 0.05 indicated statistical significance. RESULTS: The six histogram parameters of both QSM and R2*were significantly correlated with intrahepatic elasticity. In particular, three parameters (variance, percentiles [90th and 99th]) of QSM showed high correlation (r = 0.818-0.844), whereas R2* parameters showed a moderate correlation with elasticity. Four parameters of QSM were significantly correlated with fibrosis stage (ρ = 0.637-0.723) and differentiated F2-4 from F0-1 fibrosis and F3-4 from F0-2 fibrosis with areas under the ROC curve of > 0.8, but those of R2* did not. CONCLUSION: QSM may serve as a promising surrogate indicator in detecting liver fibrosis.

Utility of Radial Scanning for the Identification of Arterial Hypervascularity of Hepatocellular Carcinoma on Gadoxetic Acid-Enhanced Magnetic Resonance Images.

OBJECTIVES: This study aimed to compare the accuracy of assessing the arterial hypervascularity of hepatocellular carcinoma (HCC) on dynamic computed tomography (CT) scans and gadoxetic acid (EOB)-enhanced magnetic resonance imaging (MRI) scans performed with radial sampling. METHODS: We studied the images of 40 patients with hypervascular HCC. A radiologist recorded the standard deviation of the attenuation (or the signal intensity [SI]) in subcutaneous fat tissue as the image noise (N) and calculated the contrast-to-noise ratio (CNR) as follows: (CNR) = (n-ROIT - n-ROIL)/N, where n-ROIT is the mean attenuation (or SI) of the tumor divided by the mean attenuation (or SI) of the aorta and n-ROIL is the mean attenuation (or SI) of the liver parenchyma divided by the mean attenuation (or SI) of the aorta. RESULTS: The CNR was significantly higher on EOB-enhanced MRI than on dynamic CT scans. CONCLUSIONS: For the assessment of HCC vascularity, EOB-enhanced MRI scans acquired with radial sampling were more accurate than dynamic CT images.

Quantitative susceptibility mapping of prostate with separate calculations for water and fat regions for reducing shading artifacts.

We propose a novel processing method for reducing shading artifacts in quantitative susceptibility mapping (QSM) for prostate imaging. In the conventional method, calculation errors in the boundary regions between water and fat cause shading artifacts that degrade the image quality for QSM. In the proposed method, water and fat regions are separated, and susceptibilities in these two regions are calculated separately and then combined. Susceptibility in the water regions is calculated by using the fat regions as a background susceptibility source to remove shading artifacts. Susceptibility in the fat regions is calculated by using the constraint that shading artifacts in the water regions are suppressed to improve accuracy. In quantitative evaluation of the method with a numerical simulation, calculation errors for the water and fat regions were reduced by 62% and 85%, respectively, compared with the conventional method. In visual evaluation using human prostate imaging, the proposed method also reduced the shading artifacts unlike the conventional method. The proposed method is expected to improve the performance of QSM in diagnosing such diseases as prostate cancer.