3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;48:1570-1577.
In radiomics studies, researchers usually need to develop a supervised machine learning model to map image features onto the clinical conclusion. A classical machine learning pipeline consists of several steps, including normalization, feature selection, and classification. It is often tedious to find an optimal pipeline with appropriate combinations. We designed an open-source software package named FeAture Explorer (FAE). It was programmed with Python and used NumPy, pandas, and scikit-learning modules. FAE can be used to extract image features, preprocess the feature matrix, develop different models automatically, and evaluate them with common clinical statistics. FAE features a user-friendly graphical user interface that can be used by radiologists and researchers to build many different pipelines, and to compare their results visually. To prove the effectiveness of FAE, we developed a candidate model to classify the clinical-significant prostate cancer (CS PCa) and non-CS PCa using the PROSTATEx dataset. We used FAE to try out different combinations of feature selectors and classifiers, compare the area under the receiver operating characteristic curve of different models on the validation dataset, and evaluate the model using independent test data. The final model with the analysis of variance as the feature selector and linear discriminate analysis as the classifier was selected and evaluated conveniently by FAE. The area under the receiver operating characteristic curve on the training, validation, and test dataset achieved results of 0.838, 0.814, and 0.824, respectively. FAE allows researchers to build radiomics models and evaluate them using an independent testing dataset. It also provides easy model comparison and result visualization. We believe FAE can be a convenient tool for radiomics studies and other medical studies involving supervised machine learning.
Background and Purpose: Mean apparent propagator (MAP) MRI is a novel diffusion imaging method to map tissue microstructure. The purpose of this study was to evaluate the diagnostic value of the MAP MRI in Parkinson's disease (PD) in comparison with conventional diffusion tensor imaging (DTI). Methods: 23 PD patients and 22 age-and gender-matched healthy controls were included. MAP MRI and DTI were performed on a 3T MR scanner with a 20-channel head coil. The MAP metrics including mean square displacement (MSD), return to the origin probability (RTOP), return to the axis probability (RTAP), and return to the plane probability (RTPP), and DTI metrics including fractional anisotropy (FA), and mean diffusivity (MD), were measured in subcortical gray matter and compared between the two groups. The receiver operating characteristic (ROC) curve was used to analyze the diagnostic performance of all the metrics. The association between the diffusion metrics and disease severity was assessed by Pearson correlation analysis. Results: For MAP MRI, the mean values of MSD in the bilateral caudate, pallidum, putamen, thalamus and substantia nigra (SN) were higher in PD patients than in healthy controls (p FDR ≤ 0.001); the mean values of the zero displacement probabilities (RTOP, RTAP, and RTPP) in the bilateral caudate, pallidum, putamen and thalamus were lower in PD patients (p FDR < 0.001). For DTI, only FA in the bilateral SN was significantly higher in PD patients than those in the controls (p FDR < 0.001). ROC analysis showed that the areas under the curves of MAP MRI metrics (MSD, RTOP, RTAP, and RTPP) in the bilateral caudate, pallidum, putamen and thalamus (range, 0.85-0.94) were greater than those of FA and MD of DTI (range, 0.55-0.69) in discriminating between PD patients and September 2020 | Volume 12 | Article 563595 Le et al. MAP MRI in PD healthy controls. RTAP in the ipsilateral pallidum (r = −0.56, p FDR = 0.027), RTOP in the bilateral and contralateral putamen (r = −0.58, p FDR = 0.019; r = −0.57, p FDR = 0.024) were negatively correlated with UPDRS III motor scores. Conclusion: MAP MRI outperformed the conventional DTI in the diagnosis of PD and evaluation of the disease severity.
Background To compare the diagnostic performance of neurite orientation dispersion and density imaging (NODDI), mean apparent propagator magnetic resonance imaging (MAP-MRI), diffusion kurtosis imaging (DKI), diffusion tensor imaging (DTI) and diffusion-weighted imaging (DWI) in distinguishing high-grade gliomas (HGGs) from solitary brain metastases (SBMs). Methods Patients with previously untreated, histopathologically confirmed HGGs (n = 20) or SBMs (n = 21) appearing as a solitary and contrast-enhancing lesion on structural MRI were prospectively recruited to undergo diffusion-weighted MRI. DWI data were obtained using a q-space Cartesian grid sampling procedure and were processed to generate parametric maps by fitting the NODDI, MAP-MRI, DKI, DTI and DWI models. The diffusion metrics of the contrast-enhancing tumor and peritumoral edema were measured. Differences in the diffusion metrics were compared between HGGs and SBMs, followed by receiver operating characteristic (ROC) analysis and the Hanley and McNeill test to determine their diagnostic performances. Results NODDI-based isotropic volume fraction (Viso) and orientation dispersion index (ODI); MAP-MRI-based mean-squared displacement (MSD) and q-space inverse variance (QIV); DKI-generated radial, mean diffusivity and fractional anisotropy (RDk, MDk and FAk); and DTI-generated radial, mean diffusivity and fractional anisotropy (RD, MD and FA) of the contrast-enhancing tumor were significantly different between HGGs and SBMs (p < 0.05). The best single discriminative parameters of each model were Viso, MSD, RDk and RD for NODDI, MAP-MRI, DKI and DTI, respectively. The AUC of Viso (0.871) was significantly higher than that of MSD (0.736), RDk (0.760) and RD (0.733) (p < 0.05). Conclusion NODDI outperforms MAP-MRI, DKI, DTI and DWI in differentiating between HGGs and SBMs. NODDI-based Viso has the highest performance.
Image denoising has a profound impact on the precision of estimated parameters in diffusion kurtosis imaging (DKI). This work first proposes an approach to constructing a DKI phantom that can be used to evaluate the performance of denoising algorithms in regard to their abilities of improving the reliability of DKI parameter estimation. The phantom was constructed from a real DKI dataset of a human brain, and the pipeline used to construct the phantom consists of diffusion-weighted (DW) image filtering, diffusion and kurtosis tensor regularization, and DW image reconstruction. The phantom preserves the image structure while minimizing image noise, and thus can be used as ground truth in the evaluation. Second, we used the phantom to evaluate three representative algorithms of non-local means (NLM). Results showed that one scheme of vector-based NLM, which uses DWI data with redundant information acquired at different b-values, produced the most reliable estimation of DKI parameters in terms of Mean Square Error (MSE), Bias and standard deviation (Std). The result of the comparison based on the phantom was consistent with those based on real datasets.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.