Role of texture analysis in breast MRI as a cancer biomarker: A review

Diagnostic accuracy of MRI textural analysis in the classification of breast tumors

OBJECTIVE

To investigate whether textural analysis (TA) of MRI heterogeneity may play a role in the clinical assessment and classification of breast tumors.

MATERIALS AND METHODS

For this retrospective study, patients with breast masses ≥1 cm on contrast-enhanced MRI were obtained in 69 women (mean age: 51 years; range 21-78 years) with 77 masses (38 benign, 39 malignant) from 2006 to 2018. The selected single slice sagittal peak post-contrast T1-weighted image was analyzed with commercially available TA software [TexRAD Ltd., UK]. Eight histogram TA parameters were evaluated at various spatial scaling factors (SSF) including mean pixel intensity, standard deviation of the pixel histogram (SD), entropy, mean of the positive pixels (MPP), skewness, kurtosis, sigma, and Tx_sigma. Additional statistical tests were used to determine their predictiveness.

RESULTS

Entropy showed a significant difference between benign and malignant tumors at all textural scales (p < 0.0001) and kurtosis was significant at SSF = 0-5 (p = 0.0026-0.0241). The single best predictor was entropy at SSF = 4 with AUC = 0.80, giving a sensitivity of 95% and specificity of 53%. An AUC of 0.91 was found using a model combining entropy with sigma, which yielded better performance with a sensitivity of 92% and specificity of 79%.

CONCLUSION

TA of breast masses has the potential to assist radiologists in categorizing tumors as benign or malignant on MRI. Measurements of entropy, kurtosis, and entropy combined with sigma may provide the best predictability.

An MRI-Based Radiomic Prognostic Index Predicts Poor Outcome and Specific Genetic Alterations in Endometrial Cancer

Integrative tumor characterization linking radiomic profiles to corresponding gene expression profiles has the potential to identify specific genetic alterations based on non-invasive radiomic profiling in cancer. The aim of this study was to develop and validate a radiomic prognostic index (RPI) based on preoperative magnetic resonance imaging (MRI) and assess possible associations between the RPI and gene expression profiles in endometrial cancer patients. Tumor texture features were extracted from preoperative 2D MRI in 177 endometrial cancer patients. The RPI was developed using least absolute shrinkage and selection operator (LASSO) Cox regression in a study cohort (n = 95) and validated in an MRI validation cohort (n = 82). Transcriptional alterations associated with the RPI were investigated in the study cohort. Potential prognostic markers were further explored for validation in an mRNA validation cohort (n = 161). The RPI included four tumor texture features, and a high RPI was significantly associated with poor disease-specific survival in both the study cohort (p < 0.001) and the MRI validation cohort (p = 0.030). The association between RPI and gene expression profiles revealed 46 significantly differentially expressed genes in patients with a high RPI versus a low RPI (p < 0.001). The most differentially expressed genes, COMP and DMBT1, were significantly associated with disease-specific survival in both the study cohort and the mRNA validation cohort. In conclusion, a high RPI score predicts poor outcome and is associated with specific gene expression profiles in endometrial cancer patients. The promising link between radiomic tumor profiles and molecular alterations may aid in developing refined prognostication and targeted treatment strategies in endometrial cancer.

Texture Analysis in the Diagnosis of Primary Breast Cancer: Comparison of High-Resolution Dedicated Breast Positron Emission Tomography (dbPET) and Whole-Body PET/CT

Objective: This retrospective study aimed to compare the ability to classify tumor characteristics of breast cancer (BC) of positron emission tomography (PET)-derived texture features between dedicated breast PET (dbPET) and whole-body PET/computed tomography (CT).

Methods: Forty-four BCs scanned by both high-resolution ring-shaped dbPET and whole-body PET/CT were analyzed. The primary BC was extracted with a standardized uptake value (SUV) threshold segmentation method. On both dbPET and PET/CT images, 38 texture features were computed; their ability to classify tumor characteristics such as tumor (T)-category, lymph node (N)-category, molecular subtype, and Ki67 levels was compared. The texture features were evaluated using univariate and multivariate analyses following principal component analysis (PCA). AUC values were used to evaluate the diagnostic power of the computed texture features to classify BC characteristics.

Results: Some texture features of dbPET and PET/CT were different between Tis-1 and T2-4 and between Luminal A and other groups, respectively. No association with texture features was found in the N-category or Ki67 level. In contrast, receiver-operating characteristic analysis using texture features' principal components showed that the AUC for classification of any BC characteristics were equally good for both dbPET and whole-body PET/CT.

Conclusions: PET-based texture analysis of dbPET and whole-body PET/CT may have equally good classification power for BC.

Bone Structure Analysis of the Radius Using Ultrahigh Field (7T) MRI: Relevance of Technical Parameters and Comparison with 3T MRI and Radiography

Bone fractal signature analysis (FSA-also termed bone texture analysis) is a tool that assesses structural changes that may relate to clinical outcomes and functions. Our aim was to compare bone texture analysis of the distal radius in patients and volunteers using radiography and 3T and 7T magnetic resonance imaging (MRI)-a patient group (n = 25) and a volunteer group (n = 25) were included. Participants in the patient group had a history of chronic wrist pain with suspected or confirmed osteoarthritis and/or ligament instability. All participants had 3T and 7T MRI including T1-weighted turbo spin echo (TSE) sequences. The 7T MRI examination included an additional high-resolution (HR) T1 TSE sequence. Radiographs of the wrist were acquired for the patient group. When comparing patients and volunteers (unadjusted for gender and age), we found a statistically significant difference of horizontal and vertical fractal dimensions (FDs) using 7T T1 TSE-HR images in low-resolution mode (horizontal: p = 0.04, vertical: p = 0.01). When comparing radiography to the different MRI sequences, we found a statistically significant difference for low- and high-resolution horizontal FDs between radiography and 3T T1 TSE and 7T T1 TSE-HR. Vertical FDs were significantly different only between radiographs and 3T T1 TSE in the high-resolution mode; FSA measures obtained from 3T and 7T MRI are highly dependent on the sequence and reconstruction resolution used, and thus are not easily comparable between MRI systems and applied sequences.

Texture analysis in the characterization of parotid salivary gland lesions: A study on MR diffusion weighted imaging

BACKGROUND AND PURPOSE

Parotid lesions show overlaps of morphological findings, apparent diffusion coefficient (ADC) values and types of time/intensity curve. This research aimed to evaluate the role of diffusion weighted imaging texture analysis in differentiating between benign and malignant parotid lesions and in characterizing pleomorphic adenoma (PA), Warthin tumor (WT), epithelial malignancy (EM), and lymphoma (LY).

METHODS

Texture analysis of 54 parotid lesions (19 PA, 14 WT, 14 EM, and 7 LY) was performed on ADC map images. An ANOVA test was used to estimate both the difference between benign and malignant lesions and the texture feature differences among PA, WT, EM, and LY. A P-value≤0.01 was considered to be statistically significant. A cut-off value defined by ROC curve analysis was found for each statistically significant texture parameter. The diagnostic accuracy was obtained for each texture parameter with AUC ≥ 0.5. The agreement between each texture parameter and histology was calculated using the Cohen's kappa coefficient.

RESULTS

The mean kappa values were 0.61, 0.34, 0.26, 0.17, and 0.48 for LY, EM, WT, PA, and benign vs. malignant lesions respectively. Long zone emphasis cut-off values >1.870 indicated EM with an accuracy of 81 % and values >2.630 revealed LY with an accuracy of 93 %. Long run emphasis values >1.050 and >1.070 indicated EM and LY with a diagnostic accuracy of 79% and 93% respectively.

CONCLUSIONS

Long zone emphasis and long run emphasis texture parameters allowed the identification of LY and the differentiation between benign and malignant lesions. WT and PA were not accurately recognized.

Multi-marker quantitative radiomics for mass characterization in dedicated breast CT imaging

Purpose

To develop and evaluate the diagnostic performance of an algorithm for multi‐marker radiomic‐based classification of breast masses in dedicated breast computed tomography (bCT) images.

Methods

Over 1000 radiomic descriptors aimed at quantifying mass and border heterogeneity, morphology, and margin sharpness were developed and implemented. These included well‐established texture and shape feature descriptors, which were supplemented with additional approaches for contour irregularity quantification, spicule and lobe detection, characterization of degree of infiltration, and differences in peritumoral compartments. All descriptors were extracted from a training set of 202 bCT masses (133 benign and 69 malignant), and their individual diagnostic performance was investigated in terms of area under the receiver operating characteristics (ROC) curve (AUC) of single‐feature‐based linear discriminant analysis (LDA) classifiers. Subsequently, the most relevant descriptors were selected through a multiple‐step feature selection process (including stability analysis, statistical significance, evaluation of feature interaction, and dimensionality reduction), and used to develop a final LDA radiomic model for classification of benign and malignant masses, which was then tested on an independent test set of 82 cases (45 benign and 37 malignant).

Results

The majority of the individual radiomic descriptors showed, on the training set, an AUC value deriving from a linear decision boundary higher than 0.65, with the lower limit of the associated 95% confidence interval (C.I.) not overlapping with random chance (AUC = 0.5). The final LDA radiomic model resulted in a test set AUC of 0.90 (95% C.I. 0.80–0.96).

Conclusions

The proposed multi‐marker radiomic approach achieved high diagnostic accuracy in bCT mass classification, using a radiomic signature based on different feature types. While future studies with larger datasets are needed to further validate these results, quantitative radiomics applied to bCT shows potential to improve the breast cancer diagnosis pipeline.

Magnetic Resonance Texture Analysis in Alzheimer's disease

Texture analysis is an emerging field that allows mathematical detection of changes in MRI signals that are not visible among image pixels. Alzheimer's disease, a progressive neurodegenerative disease, is the most common cause of dementia. Recently, multiple texture analysis studies in patients with Alzheimer's disease have been performed. This review summarizes the main contributors to Alzheimer's disease-associated cognitive decline, presents a brief overview of texture analysis, followed by review of various MR imaging texture analysis applications in Alzheimer's disease. We also discuss the current challenges for widespread clinical utilization. MR texture analysis could potentially be applied to develop neuroimaging biomarkers for use in Alzheimer's disease clinical trials and diagnosis.

Feasibility of T1 mapping with histogram analysis for the diagnosis and staging of liver fibrosis: Preclinical results

PURPOSE

To compare the diagnostic accuracy of parameters derived from the histogram analysis of precontrast, 10-min hepatobiliary phase (HBP) and 20-min HBP T1 maps for staging liver fibrosis (LF).

METHODS

LF was induced in New Zealand white rabbits by subcutaneous injections of carbon tetrachloride for 4-16 weeks (n = 120), and 20 rabbits injected with saline served as controls. Precontrast, 10-min and 20-min HBP modified Look-Locker inversion recovery (MOLLI) T1 mapping was performed. Histogram analysis of T1 maps was performed, and the mean, median, skewness, kurtosis, entropy, inhomogeneity and 10th/25th/75th/90th percentiles of T1_native, T1_10min and T1_20min were derived. Quantitative histogram parameters were compared. For significant parameters, further receiver operating characteristic (ROC) analyses were performed to evaluate the potential diagnostic performance in differentiating LF stages.

RESULTS

Finally, 17, 20, 21, 21 and 20 rabbits were included for the F0, F1, F2, F3, and F4 pathological grades of fibrosis, respectively. The mean/75th of T1_native, entropy of T1_10min and entropy/mean/median/10th of T1_20min demonstrated a significant good correlation with the LF stage (|r| = 0.543-0.866, all P < 0.05). The 75th of T1_native, entropy_10min, and entropy_20min were the three most reliable imaging markers in reflecting the stage of LF. The area under the ROC curve of entropy_20min was larger than that of entropy_10min (P < 0.05 for LF ≥ F2, ≥F3, and ≥ F4) and the 75th of T1_native (P < 0.05 for LF ≥ F2 and ≥ F3) for staging LF.

CONCLUSION

Magnetic resonance histogram analysis of T1 maps, particularly the entropy derived from 20-min HBP T1 mapping, is promising for predicting the LF stage.

Multi-parametric MRI lesion heterogeneity biomarkers for breast cancer diagnosis

PURPOSE

To identify intra-lesion imaging heterogeneity biomarkers in multi-parametric Magnetic Resonance Imaging (mpMRI) for breast lesion diagnosis.

METHODS

Dynamic Contrast Enhanced (DCE) and Diffusion Weighted Imaging (DWI) of 73 female patients, with 85 histologically verified breast lesions were acquired. Non-rigid multi-resolution registration was utilized to spatially align sequences. Four (4) DCE (2^nd post-contrast frame, Initial-Enhancement, Post-Initial-Enhancement and Signal-Enhancement-Ratio) and one (1) DWI (Apparent-Diffusion-Coefficient) representations were analyzed, considering a representative lesion slice. 11 1^st-order-statistics and 16 texture features (Gray-Level-Co-occurrence-Matrix (GLCM) and Gray-Level-Run-Length-Matrix (GLRLM) based) were derived from lesion segments, provided by Fuzzy C-Means segmentation, across the 5 representations, resulting in 135 features. Least-Absolute-Shrinkage and Selection-Operator (LASSO) regression was utilized to select optimal feature subsets, subsequently fed into 3 classification schemes: Logistic-Regression (LR), Random-Forest (RF), Support-Vector-Machine-Sequential-Minimal-Optimization (SVM-SMO), assessed with Receiver-Operating-Characteristic (ROC) analysis.

RESULTS

LASSO regression resulted in 7, 6 and 7 features subsets from DCE, DWI and mpMRI, respectively. Best classification performance was obtained by the RF multi-parametric scheme (Area-Under-ROC-Curve, (AUC) ± Standard-Error (SE), AUC ± SE = 0.984 ± 0.025), as compared to DCE (AUC ± SE = 0.961 ± 0.030) and DWI (AUC ± SE = 0.938 ± 0.032) and statistically significantly higher as compared to DWI. The selected mpMRI feature subset highlights the significance of entropy (1^st-order-statistics and 2^nd-order-statistics (GLCM)) and percentile features extracted from 2^nd post-contrast frame, PIE, SER maps and ADC map.

CONCLUSION

Capturing breast intra-lesion heterogeneity, across mpMRI lesion segments with 1^st-order-statistics and texture features (GLCM and GLRLM based), offers a valuable diagnostic tool for breast cancer.

Harmonization of Quantitative Parenchymal Enhancement in T1 -Weighted Breast MRI

BACKGROUND

Differences in imaging parameters influence computer-extracted parenchymal enhancement measures from breast MRI.

PURPOSE

To investigate the effect of differences in dynamic contrast-enhanced MRI acquisition parameter settings on quantitative parenchymal enhancement of the breast, and to evaluate harmonization of contrast-enhancement values with respect to flip angle and repetition time.

STUDY TYPE

Retrospective.

PHANTOM/POPULATIONS

We modeled parenchymal enhancement using simulations, a phantom, and two cohorts (N = 398 and N = 302) from independent cancer centers.

SEQUENCE FIELD/STRENGTH

1.5T dynamic contrast-enhanced T1 -weighted spoiled gradient echo MRI. Vendors: Philips, Siemens, General Electric Medical Systems.

ASSESSMENT

We assessed harmonization of parenchymal enhancement in simulations and phantom by varying the MR parameters that influence the amount of T1 -weighting: flip angle (8°-25°) and repetition time (4-12 msec). We calculated the median and interquartile range (IQR) of the enhancement values before and after harmonization. In vivo, we assessed overlap of quantitative parenchymal enhancement in the cohorts before and after harmonization using kernel density estimations. Cohort 1 was scanned with flip angle 20° and repetition time 8 msec; cohort 2 with flip angle 10° and repetition time 6 msec.

STATISTICAL TESTS

Paired Wilcoxon signed-rank-test of bootstrapped kernel density estimations.

RESULTS

Before harmonization, simulated enhancement values had a median (IQR) of 0.46 (0.34-0.49). After harmonization, the IQR was reduced: median (IQR): 0.44 (0.44-0.45). In the phantom, the IQR also decreased, median (IQR): 0.96 (0.59-1.22) before harmonization, 0.96 (0.91-1.02) after harmonization. Harmonization yielded significantly (P < 0.001) better overlap in parenchymal enhancement between the cohorts: median (IQR) was 0.46 (0.37-0.58) for cohort 1 vs. 0.37 (0.30-0.44) for cohort 2 before harmonization (57% overlap); and 0.35 (0.28-0.43) vs. .0.37 (0.30-0.44) after harmonization (85% overlap).

DATA CONCLUSION

The proposed practical harmonization method enables an accurate comparison between patients scanned with differences in imaging parameters.

LEVEL OF EVIDENCE

3 TECHNICAL EFFICACY STAGE: 4.

Ultrasound-Based Radiomics Analysis for Preoperatively Predicting Different Histopathological Subtypes of Primary Liver Cancer

Background

Preoperative identification of hepatocellular carcinoma (HCC), combined hepatocellular–cholangiocarcinoma (cHCC-ICC), and intrahepatic cholangiocarcinoma (ICC) is essential for treatment decision making. We aimed to use ultrasound-based radiomics analysis to non-invasively distinguish histopathological subtypes of primary liver cancer (PLC) before surgery.

Methods

We retrospectively analyzed ultrasound images of 668 PLC patients, comprising 531 HCC patients, 48 cHCC-ICC patients, and 89 ICC patients. The boundary of a tumor was manually determined on the largest imaging slice of the ultrasound medicine image by ITK-SNAP software (version 3.8.0), and then, the high-throughput radiomics features were extracted from the obtained region of interest (ROI) of the tumor. The combination of different dimension-reduction technologies and machine learning approaches was used to identify important features and develop the moderate radiomics model. The comprehensive ability of the radiomics model can be evaluated by the area under the receiver operating characteristic curve (AUC).

Results

After digitally processing tumor ultrasound images, 5,234 high-throughput radiomics features were obtained. We used the Spearman + least absolute shrinkage and selection operator (LASSO) regression method for feature selection and logistics regression for modeling to develop the HCC-vs-non-HCC radiomics model (composed of 16 features). The Spearman + statistical test + random forest methods were used for feature selection, and logistics regression was applied for modeling to develop the ICC-vs-cHCC-ICC radiomics model (composed of 19 features). The overall performance of the radiomics model in identifying different histopathological types of PLC was moderate, with AUC values of 0.854 (training cohort) and 0.775 (test cohort) in the HCC-vs-non-HCC radiomics model and 0.920 (training cohort) and 0.728 (test cohort) in the ICC-vs-cHCC-ICC radiomics model.

Conclusion

Ultrasound-based radiomics models can help distinguish histopathological subtypes of PLC and provide effective clinical decision making for the accurate diagnosis and treatment of PLC.

Texture Analysis Based on Sagittal Fat-Suppression and Transverse T2-Weighted Magnetic Resonance Imaging for Determining Local Invasion of Rectal Cancer

Background: Accurate evaluation of local invasion (T-stage) of rectal cancer is essential for treatment planning. A search of PubMed database indicated that the correlation between texture features from T2-weighted magnetic resonance imaging (T2WI) (MRI) and T-stage has not been explored extensively.

Purpose: To evaluate the performance of texture analysis using sagittal fat-suppression combined with transverse T2WI for determining T-stage of rectal cancer.

Methods: One hundred and seventy-four rectal cancer cases who underwent preoperative MRI were retrospectively selected and divided into high (T3/4) and low (T1/2) T-stage groups. Texture features were, respectively, extracted from sagittal fat-suppression and transverse T2WI images. Univariate and multivariate analyses were conducted to determine T-stage. Discrimination performance was assessed by receiver operating characteristic (ROC) analysis.

Results: For univariate analysis, the best performance in differentiating T1/2 from T3/4 tumors was achieved from transverse T2WI, and the area under the ROC curve (AUC) was 0.740. For multivariate analysis, the logical regression model incorporating the independent predictors achieved an AUC of 0.789.

Conclusions: Texture features from sagittal fat-suppression combined with transverse T2WI presented moderate association with T-stage of rectal cancer. These findings may be valuable in selecting optimum treatment strategy.

Application of Texture Analysis Based on Sagittal Fat-Suppression and Oblique Axial T2-Weighted Magnetic Resonance Imaging to Identify Lymph Node Invasion Status of Rectal Cancer

Objective: To investigate the value of texture features derived from T2-weighted magnetic resonance imaging (T2WI) for predicting preoperative lymph node invasion (N stage) in rectal cancer.

Materials and Methods: One hundred and eighty-two patients with histopathologically confirmed rectal cancer and preoperative magnetic resonance imaging were retrospectively analyzed, who were divided into high (N1-2) and low N stage (N0). Texture features were calculated from histogram, gray-level co-occurrence matrix, and gray-level run-length matrix from sagittal fat-suppression and oblique axial T2WI. Independent sample t-test or Mann-Whitney U-test were used for statistical analysis. Multivariate logistic regression analysis was conducted to build the predictive models. Predictive performance was evaluated by receiver operating characteristic (ROC) analysis.

Results: Energy (ENE), entropy (ENT), information correlation (INC), long-run emphasis (LRE), and short-run low gray-level emphasis (SRLGLE) extracted from sagittal fat-suppression T2WI, and ENE, ENT, INC, low gray-level run emphasis (LGLRE), and SRLGLE from oblique axial T2WI were significantly different between stage N0 and stage N1-2 tumors. The multivariate analysis for features from sagittal fat-suppression T2WI showed that higher SRLGLE and lower ENE were independent predictors of lymph node invasion. The model reached an area under ROC curve (AUC) of 0.759. The analysis for features from oblique axial T2WI showed that higher INC and SRLGLE were independent predictors. The model achieved an AUC of 0.747. The analysis for all extracted features showed that lower ENE from sagittal fat-suppression T2WI and higher INC and SRLGLE from oblique axial T2WI were independent predictors. The model showed an AUC of 0.772.

Conclusions: Texture features derived from T2WI could provide valuable information for identifying the status of lymph node invasion in rectal cancer.

Value of Conventional MRI Texture Analysis in the Differential Diagnosis of Phyllodes Tumors and Fibroadenomas of the Breast

Background

There is substantial overlap in MRI findings between phyllodes tumors (PTs) and fibroadenomas (FAs). Our study was performed to investigate the value of conventional MRI texture analysis in the differential diagnosis of PTs and FAs.

Methods

Preoperative MRI data - including axial T1WI, T2WI_FS (T2WI with fat suppression), dynamic contrast-enhanced (DCE)-T1WI_2min and DCE-T1WI_7min (T1WI post-strengthened for 2 and 7 min, respectively, on DCE-MRI) - of 45 patients with PTs and 67 patients with FAs were retrospectively analyzed. MaZda 4.7 software was used to manually draw the maximum ROIs at the same lesion level of the above MRI images. The optimized feature selection methods included Fisher's coefficient, probability of classification error and average correction coefficient (POE + ACC), and mutual information (MI) as well as a combination of the above 3 methods (F + POE + ACC + MI [FPM]), respectively. The misclassification rates of PTs and FAs were compared between texture analysis and subjective diagnosis by radiologists.

Results

The DCE-T1WI_7min images had the lowest misclassification rate of 10.71% (12/112). The misclassification rate for the radiologists' analysis (31.25%, 35/112) was higher than that of all the texture analysis, and there was a statistically significant difference between the radiologists' misclassification rates and those from the FPM method in terms of the T2WI_FS and DCE-T1WI_2min images (all p < 0.05), and for the DCE-T1WI_7min images by using the Fisher and FPM methods (all p < 0.05).

Conclusion

Texture analysis of conventional MRI can be used as an assistant tool in providing a certain objective basis for differentiating PTs from FAs.

Current and Emerging Magnetic Resonance-Based Techniques for Breast Cancer

Breast cancer is the most commonly diagnosed cancer among women worldwide, and early detection remains a principal factor for improved patient outcomes and reduced mortality. Clinically, magnetic resonance imaging (MRI) techniques are routinely used in determining benign and malignant tumor phenotypes and for monitoring treatment outcomes. Static MRI techniques enable superior structural contrast between adipose and fibroglandular tissues, while dynamic MRI techniques can elucidate functional characteristics of malignant tumors. The preferred clinical procedure-dynamic contrast-enhanced MRI-illuminates the hypervascularity of breast tumors through a gadolinium-based contrast agent; however, accumulation of the potentially toxic contrast agent remains a major limitation of the technique, propelling MRI research toward finding an alternative, noninvasive method. Three such techniques are magnetic resonance spectroscopy, chemical exchange saturation transfer, and non-contrast diffusion weighted imaging. These methods shed light on underlying chemical composition, provide snapshots of tissue metabolism, and more pronouncedly characterize microstructural heterogeneity. This review article outlines the present state of clinical MRI for breast cancer and examines several research techniques that demonstrate capacity for clinical translation. Ultimately, multi-parametric MRI-incorporating one or more of these emerging methods-presently holds the best potential to afford improved specificity and deliver excellent accuracy to clinics for the prediction, detection, and monitoring of breast cancer.

Ependymoma and pilocytic astrocytoma: Differentiation using radiomics approach based on machine learning

Mandatory accurate and specific diagnosis demands have brought about increased challenges for radiologists in pediatric posterior fossa tumor prediction and prognosis. With the development of high-performance computing and machine learning technologies, radiomics provides increasing opportunities for clinical decision-making. Several studies have applied radiomics as a decision support tool in intracranial tumors differentiation. Here we seek to achieve preoperative differentiation between ependymoma (EP) and pilocytic astrocytoma (PA) using radiomics analysis method based on machine learning. A total of 135 Magnetic Resonance Imaging (MRI) slices are divided into training sets and validation sets. Three kinds of radiomics features, including Gabor transform, texture and wavelet transform based ones are used to obtain 300 multimodal features. Kruskal-Wallis test score (KWT) and support vector machines (SVM) are applied for feature selection and tumor differentiation. The performance is investigated via accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) analysis. Results show that the accuracy, sensitivity, specificity, and AUC of the selected feature set are 0.8775, 0.9292, 0.8000, and 0.8646 respectively, having no significantdifferencescomparedwiththe overall feature set. For different types of features, texture features yield the best differentiation performance and the significance analysis results are consistent with this. Our study demonstrates texture features perform better than the other features. The radiomics approach based on machine learning is efficient for pediatric posterior fossa tumors differentiation and could enhance the application of radiomics methods for assisted clinical diagnosis.

Performance evaluation of texture analysis based on kinetic parametric maps from breast DCE-MRI in classifying benign from malignant lesions

BACKGROUND AND OBJECTIVES

To investigate the performance of texture analysis based on enhancement kinetic parametric maps derived from breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in discriminating benign from malignant tumors.

METHODS

A total of 192 cases confirmed by histopathology were retrospectively selected from our Picture Archiving and Communication System, including 93 benign and 99 malignant tumors. Lesion areas were delineated semi-automatically, and six kinetic parametric maps reflecting the perfusion information were generated, namely the maximum slope of increase (MSI), slope of signal intensity (SI_slope ), initial percentage of peak enhancement (E_initial ), percentage of peak enhancement (E_peak ), early signal enhancement ratio (ESER), and second enhancement percentage (SEP) maps. A total of 286 texture features were extracted from those quantitative parametric maps. The Student t test or Mann-Whitney U test was used to select features that were statistically significantly different between the benign and malignant groups. A support vector machine was employed with a leave-one-out cross-validation method to establish the classification model. Classification performance was evaluated according to the receiver operating characteristic (ROC) theory.

RESULTS

The areas under ROC curves (AUCs) indicating the diagnostic performance were 0.925 for MSI, 0.854 for SI_slope , 0.756 for E_initial , 0.923 for E_peak , 0.871 for ESER and 0.881 for SEP. Significant differences in AUCs were found between E_initial vs MSI, E_initial vs E_peak and E_initial vs SEP (P < .05). There were no significant differences in other pairwise comparisons.

CONCLUSION

Texture analysis of the kinetic parametric maps derived from breast DCE-MRI can contribute to the discrimination between malignant and benign lesions. It can be considered as a supplementary tool for breast diagnosis.

Comparison of Breast MRI Tumor Classification Using Human-Engineered Radiomics, Transfer Learning From Deep Convolutional Neural Networks, and Fusion Methods

Digital image-based signatures of breast tumors may ultimately contribute to the design of patient-specific breast cancer diagnostics and treatments. Beyond traditional human-engineered computer vision methods, tumor classification methods using transfer learning from deep convolutional neural networks (CNNs) are actively under development. This article will first discuss our progress in using CNN-based transfer learning to characterize breast tumors for various diagnostic, prognostic, or predictive image-based tasks across multiple imaging modalities, including mammography, digital breast tomosynthesis, ultrasound (US), and magnetic resonance imaging (MRI), compared to both human-engineered feature-based radiomics and fusion classifiers created through combination of such features. Second, a new study is presented that reports on a comprehensive comparison of the classification performances of features derived from human-engineered radiomic features, CNN transfer learning, and fusion classifiers for breast lesions imaged with MRI. These studies demonstrate the utility of transfer learning for computer-aided diagnosis and highlight the synergistic improvement in classification performance using fusion classifiers.

Texture Analysis of Dynamic Contrast-Enhanced MRI in Evaluating Pathologic Complete Response (pCR) of Mass-Like Breast Cancer after Neoadjuvant Therapy

Objectives

MRI is the standard imaging method in evaluating treatment response of breast cancer after neoadjuvant therapy (NAT), while identification of pathologic complete response (pCR) remains challenging. Texture analysis (TA) on post-NAT dynamic contrast-enhanced (DCE) MRI was explored to assess the existence of pCR in mass-like cancer.

Materials and Methods

A primary cohort of 112 consecutive patients (40 pCR and 72 non-pCR) with mass-like breast cancers who received preoperative NAT were retrospectively enrolled. On post-NAT MRI, volumes of the residual-enhanced areas and TA first-order features (19 for each sequence) of the corresponding areas were achieved for both early- and late-phase DCE using an in-house radiomics software. Groups were divided according to the operational pathology. Receiver operating characteristic curves and binary logistic regression analysis were used to select features and achieve a predicting formula. Overall diagnostic abilities were compared between TA and radiologists' subjective judgments. Validation was performed on a time-independent cohort of 39 consecutive patients.

Results

TA features with high consistency (Cronbach's alpha >0.9) between 2 observers showed significant differences between pCR and non-pCR groups. Logistic regression using features selected by ROC curves generated a synthesized formula containing 3 variables (volume of residual enhancement, entropy, and robust mean absolute deviation from early-phase) to yield AUC = 0.81, higher than that of using radiologists' subjective judgment (AUC = 0.72), and entropy was an independent risk factor (P < 0.001). Accuracy and sensitivity for identifying pCR were 83.93% and 70.00%. AUC of the validation cohort was 0.80.

Conclusions

TA may help to improve the diagnostic ability of post-NAT MRI in identifying pCR in mass-like breast cancer. Entropy, as a first-order feature to depict residual tumor heterogeneity, is an important factor.

Precision diagnostics based on machine learning-derived imaging signatures

The complexity of modern multi-parametric MRI has increasingly challenged conventional interpretations of such images. Machine learning has emerged as a powerful approach to integrating diverse and complex imaging data into signatures of diagnostic and predictive value. It has also allowed us to progress from group comparisons to imaging biomarkers that offer value on an individual basis. We review several directions of research around this topic, emphasizing the use of machine learning in personalized predictions of clinical outcome, in breaking down broad umbrella diagnostic categories into more detailed and precise subtypes, and in non-invasively estimating cancer molecular characteristics. These methods and studies contribute to the field of precision medicine, by introducing more specific diagnostic and predictive biomarkers of clinical outcome, therefore pointing to better matching of treatments to patients.

Texture Analysis with 3.0-T MRI for Association of Response to Neoadjuvant Chemotherapy in Breast Cancer

Background Previous studies have suggested that texture analysis is a promising tool in the diagnosis, characterization, and assessment of treatment response in various cancer types. Therefore, application of texture analysis may be helpful for early prediction of pathologic response in breast cancer. Purpose To investigate whether texture analysis of features from MRI is associated with pathologic complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer. Materials and Methods This retrospective study included 136 women (mean age, 47.9 years; range, 31-70 years) who underwent NAC and subsequent surgery for breast cancer between January 2012 and August 2017. Patients were monitored with 3.0-T MRI before (pretreatment) and after (midtreatment) three or four cycles of NAC. Texture analysis was performed at pre- and midtreatment T2-weighted MRI, contrast material-enhanced T1-weighted MRI, diffusion-weighted MRI, and apparent diffusion coefficient (ADC) mapping by using commercial software. A random forest method was applied to build a predictive model for classifying those with pCR with use of texture parameters. Diagnostic performance for predicting pCR was assessed and compared with that of six other machine learning classifiers (adaptive boosting, decision tree, k-nearest neighbor, linear support vector machine, naive Bayes, and linear discriminant analysis) by using the Wald test and DeLong method. Results Forty of the 136 patients (29%) achieved pCR after NAC. In the prediction of pCR, the random forest classifier showed the lowest diagnostic performance with pretreatment ADC (area under the receiver operating characteristic curve [AUC], 0.53; 95% confidence interval: 0.44, 0.61) and the highest diagnostic performance with midtreatment contrast-enhanced T1-weighted MRI (AUC, 0.82; 95% confidence interval: 0.74, 0.88) among pre- and midtreatment T2-weighted MRI, contrast-enhanced T1-weighted MRI, diffusion-weighted MRI, and ADC mapping. Conclusion Texture parameters using a random forest method of contrast-enhanced T1-weighted MRI at midtreatment of neoadjuvant chemotherapy were valuable and associated with pathologic complete response in breast cancer. © RSNA, 2019 Online supplemental material is available for this article.

Diagnosis of Benign and Malignant Breast Lesions on DCE-MRI by Using Radiomics and Deep Learning With Consideration of Peritumor Tissue

BACKGROUND

Computer-aided methods have been widely applied to diagnose lesions detected on breast MRI, but fully-automatic diagnosis using deep learning is rarely reported.

PURPOSE

To evaluate the diagnostic accuracy of mass lesions using region of interest (ROI)-based, radiomics and deep-learning methods, by taking peritumor tissues into consideration.

STUDY TYPE

Retrospective.

POPULATION

In all, 133 patients with histologically confirmed 91 malignant and 62 benign mass lesions for training (74 patients with 48 malignant and 26 benign lesions for testing).

FIELD STRENGTH/SEQUENCE

3T, using the volume imaging for breast assessment (VIBRANT) dynamic contrast-enhanced (DCE) sequence.

ASSESSMENT

3D tumor segmentation was done automatically by using fuzzy-C-means algorithm with connected-component labeling. A total of 99 texture and histogram parameters were calculated for each case, and 15 were selected using random forest to build a radiomics model. Deep learning was implemented using ResNet50, evaluated with 10-fold crossvalidation. The tumor alone, smallest bounding box, and 1.2, 1.5, 2.0 times enlarged boxes were used as inputs.

STATISTICAL TESTS

The malignancy probability was calculated using each model, and the threshold of 0.5 was used to make a diagnosis.

RESULTS

In the training dataset, the diagnostic accuracy was 76% using three ROI-based parameters, 84% using the radiomics model, and 86% using ROI + radiomics model. In deep learning using the per-slice basis, the area under the receiver operating characteristic (ROC) was comparable for tumor alone, smallest and 1.2 times box (AUC = 0.97-0.99), which were significantly higher than 1.5 and 2.0 times box (AUC = 0.86 and 0.71, respectively). For per-lesion diagnosis, the highest accuracy of 91% was achieved when using the smallest bounding box, and that decreased to 84% for tumor alone and 1.2 times box, and further to 73% for 1.5 times box and 69% for 2.0 times box. In the independent testing dataset, the per-lesion diagnostic accuracy was also the highest when using the smallest bounding box, 89%.

DATA CONCLUSION

Deep learning using ResNet50 achieved a high diagnostic accuracy. Using the smallest bounding box containing proximal peritumor tissue as input had higher accuracy compared to using tumor alone or larger boxes.

LEVEL OF EVIDENCE

3 Technical Efficacy: Stage 2.

AI-based applications in hybrid imaging: how to build smart and truly multi-parametric decision models for radiomics

INTRODUCTION

The quantitative imaging features (radiomics) that can be obtained from the different modalities of current-generation hybrid imaging can give complementary information with regard to the tumour environment, as they measure different morphologic and functional imaging properties. These multi-parametric image descriptors can be combined with artificial intelligence applications into predictive models. It is now the time for hybrid PET/CT and PET/MRI to take the advantage offered by radiomics to assess the added clinical benefit of using multi-parametric models for the personalized diagnosis and prognosis of different disease phenotypes.

OBJECTIVE

The aim of the paper is to provide an overview of current challenges and available solutions to translate radiomics into hybrid PET-CT and PET-MRI imaging for a smart and truly multi-parametric decision model.

Machine learning in breast MRI

Machine-learning techniques have led to remarkable advances in data extraction and analysis of medical imaging. Applications of machine learning to breast MRI continue to expand rapidly as increasingly accurate 3D breast and lesion segmentation allows the combination of radiologist-level interpretation (eg, BI-RADS lexicon), data from advanced multiparametric imaging techniques, and patient-level data such as genetic risk markers. Advances in breast MRI feature extraction have led to rapid dataset analysis, which offers promise in large pooled multiinstitutional data analysis. The object of this review is to provide an overview of machine-learning and deep-learning techniques for breast MRI, including supervised and unsupervised methods, anatomic breast segmentation, and lesion segmentation. Finally, it explores the role of machine learning, current limitations, and future applications to texture analysis, radiomics, and radiogenomics. Level of Evidence: 3 Technical Efficacy Stage: 2 J. Magn. Reson. Imaging 2019. J. Magn. Reson. Imaging 2020;52:998-1018.