Texture-based classification of focal liver lesions on MRI at 3.0 Tesla: a feasibility study in cysts and hemangiomas

A machine learning-based approach to identify peripheral artery disease using texture features from contrast-enhanced magnetic resonance imaging

Diagnosing and assessing the risk of peripheral artery disease (PAD) has long been a focal point for medical practitioners. The impaired blood circulation in PAD patients results in altered microvascular perfusion patterns in the calf muscles which is the primary location of intermittent claudication pain. Consequently, we hypothesized that changes in perfusion and increase in connective tissue could lead to alterations in the appearance or texture patterns of the skeletal calf muscles, as visualized with non-invasive imaging techniques. We designed an automatic pipeline for textural feature extraction from contrast-enhanced magnetic resonance imaging (CE-MRI) scans and used the texture features to train machine learning models to detect the heterogeneity in the muscle pattern among PAD patients and matched controls. CE-MRIs from 36 PAD patients and 20 matched controls were used for preparing training and testing data at a 7:3 ratio with cross-validation (CV) techniques. We employed feature arrangement and selection methods to optimize the number of features. The proposed method achieved a peak accuracy of 94.11% and a mean testing accuracy of 84.85% in a 2-class classification approach (controls vs. PAD). A three-class classification approach was performed to identify a high-risk PAD sub-group which yielded an average test accuracy of 83.23% (matched controls vs. PAD without diabetes vs. PAD with diabetes). Similarly, we obtained 78.60% average accuracy among matched controls, PAD treadmill exercise completers, and PAD exercise treadmill non-completers. Machine learning and imaging-based texture features may be of interest in the study of lower extremity ischemia.

Evidence Supporting Diagnostic Value of Liver Imaging Reporting and Data System for CT- and MR Imaging-based Diagnosis of Hepatocellular Carcinoma: A Systematic Review and Meta-analysis

Background

Based on the Liver Imaging Data and Reporting System (LI-RADS) guidelines, Hepatocellular Carcinoma (HCC) can be diagnosed using imaging criteria in patients at risk of HCC.

Objective

This study aimed to assess the diagnostic value of LI-RADS in high-risk patients with HCC.

Material and Methods

This systematic review is conducted on international databases, including Google Scholar, Web of Science, PubMed, Embase, PROQUEST, and Cochrane Library, with appropriate keywords. Using the binomial distribution formula, the variance of each study was calculated, and all the data were analyzed using STATA version 16. The pooled sensitivity and specificity were determined using a random-effects meta-analysis approach. Also, we used the chi-squared test and I2 index to calculate heterogeneity among studies, and Funnel plots and Egger tests were used for evaluating publication bias.

Results

The pooled sensitivity was estimated at 0.80 (95% CI 0.76-0.84). According to different types of Liver Imaging Reporting and Data Systems (LI-RADS), the highest pooled sensitivity was in version 2018 (0.83 (95% CI 0.79-0.87) (I2: 80.6%, P of chi 2 test for heterogeneity <0.001 and T2: 0.001). The pooled specificity was estimated as 0.89 (95% CI 0.87-0.92). According to different types of LI-RADS, the highest pooled specificity was in version 2014 (93.0 (95% CI 89.0-96.0) (I2: 81.7%, P of chi 2 test for heterogeneity <0.001 and T2: 0.001).

Conclusion

LI-RADS can assist radiologists in achieving the required sensitivity and specificity in high-risk patients suspected to have HCC. Therefore, this strategy can serve as an appropriate tool for identifying HCC.

Focal Lesions of the Liver and Radiomics: What Do We Know?

Despite differences in pathological analysis, focal liver lesions are not always distinguishable in contrast-enhanced magnetic resonance imaging (MRI), contrast-enhanced computed tomography (CT), and positron emission tomography (PET). This issue can cause problems of differential diagnosis, treatment, and follow-up, especially in patients affected by HBV/HCV chronic liver disease or fatty liver disease. Radiomics is an innovative imaging approach that extracts and analyzes non-visible quantitative imaging features, supporting the radiologist in the most challenging differential diagnosis when the best-known methods are not conclusive. The purpose of this review is to evaluate the most significant CT and MRI texture features, which can discriminate between the main benign and malignant focal liver lesions and can be helpful to predict the response to pharmacological or surgical therapy and the patient's prognosis.

Using Diffusion Tensor Imaging to Explore the Changes in the Microstructure of Canine Vocal Fold Scar Tissue

OBJECTIVE

To apply diffusion tensor imaging (DTI) in measurement of the diffusion characteristics of water molecules in vocal fold scar tissue, combined with the analysis of textural characteristics of collagen fibers in the cover layer of the vocal folds to explore the feasibility of DTI in the qualitative and quantitative diagnosis of vocal fold scars and the evaluation of microstructural changes of vocal fold scar tissue.

METHODS

A unilateral injury was created using micro-cup forceps in the left vocal fold of six beagles. The contralateral normal vocal fold was used as a self-control. Five months postinjury, the larynges were excised and placed into a magnetic resonance imaging (MRI) system (9.4T BioSpec MRI, Bruker, German) for scanning and extraction of the diffusion parameters, fractional anisotropy (FA) and tensor trace in the anterior, middle, and posterior portions of the vocal fold cover layer. These parameters were then analyzed for statistical significance between the scarred vocal fold and the normal vocal fold. After MRI scanning, the tissue of the vocal folds was divided into anterior, middle, and posterior parts for sectioning and staining with hematoxylin and eosin, and samples were subsequently digitally scanned for texture analysis. The irregularity parameters, energy, contrast, correlation, and homogeneity, of collagen fibers of the vocal folds and the mean gray value of collagen fibers were calculated by the gray-level co-occurrence matrix (GLCM) texture analysis method. The differences in the mean value of the two sides of the vocal fold were compared. In addition, Pearson correlation analysis was performed between DTI parameters and irregularity parameters.

RESULTS

The FA of the left vocal fold cover layer was significantly lower compared to the self-control group (P = 0.0366), and the tensor trace value on the left vocal fold cover layer was significantly higher compared to the self-control group (P = 0.0353). The FA was significantly higher in the anterior part of the right vocal fold cover layer compared to the middle and posterior parts of the same side (P = 0.0352), and the tensor trace was significantly lower in the anterior part of the right vocal fold cover layer compared to the middle and posterior parts of the same side (P = 0.0298). There were no significant differences in FA and tensor trace between the middle and posterior parts of the vocal fold cover layer. The mean gray value of the left vocal folds cover layer was significantly smaller than the right vocal fold cover layer (P = 0.0219), the energy of the left vocal fold cover layer was significantly smaller than that of the right vocal fold cover layer (P < 0.0001), the contrast of the left vocal folds cover layer was significantly larger than that of the right vocal fold cover layer (P = 0.0002), the correlation of the left vocal folds cover layer was significantly smaller than the right vocal fold cover layer (P = 0.0002), and the homogeneity of the left vocal folds cover layer was significantly smaller than the right vocal fold cover layer (P = 0.0003). Pearson correlation analysis yielded values of r = 0.926, P = 0.000 between the FA and mean gray value; r = -0.918, P = 0.000 between FA and energy; r = -0.924, P = 0.000 between the FA and homogeneity, r = -0.949, P = 0.000 between tensor trace and mean gray value; r = 0.893, P = 0.000 between the tensor trace and energy; and r = 0.929, P = 0.000 between the tensor trace and homogeneity.

CONCLUSION

FA and tensor trace can be used as effective parameters to reflect microstructural changes in vocal fold scars. DTI is an objective and quantitative method of analyzing vocal fold scarring, and it noninvasively evaluates the microstructure of vocal fold collagen fibers.

Performance of Machine Learning Methods Based on Multi-Sequence Textural Parameters Using Magnetic Resonance Imaging and Clinical Information to Differentiate Malignant and Benign Soft Tissue Tumors

RATIONALE AND OBJECTIVES

To evaluate the performance of a machine learning method to differentiate malignant from benign soft tissue tumors based on textural features on multiparametric magnetic resonance imaging (mpMRI).

MATERIALS AND METHODS

We enrolled 163 patients with soft tissue tumors whose diagnosis was pathologically proven (71 malignant, 92 benign). All patients underwent mpMRI. Twelve histographic and textural parameters were assessed on T1-weighted imaging (T1WI), T2-weighted imaging, diffusion-weighted imaging, apparent diffusion coefficient maps, and contrast-enhanced T1WI imaging. We compared mean signals of all sequences from the malignant and benign tumors using Welch's t-test. Prediction models were developed via a machine learning technique (support vector machine) using textural features of each sequence, clinical information (sex + age + tumor size), and the combined model incorporating all features. Areas under the receiver operating characteristic curves (AUCs) of these models were calculated using fivefold cross validation.

RESULTS

The diagnostic ability of clinical information model (AUC 0.85) was not inferior to the model with textural features of each sequence (AUC 0.79-0.84). The combined model showed the highest diagnostic ability (AUC 0.89). The AUC of the combined model (0.89) was comparable to those of two board-certified radiologists (0.89 and 0.87).

CONCLUSIONS

Machine learning methods based on textural features on mpMRI and clinical information offer adequate diagnostic performance to differentiate between malignant and benign soft tissue tumors.

Texture analysis of contrast-enhanced magnetic resonance imaging predicts microvascular invasion in hepatocellular carcinoma

BACKGROUND

Microvascular invasion is one of the important risk factors of postoperative recurrence of hepatocellular carcinoma. Texture analysis uses mathematical methods to analyze the gray's quantitative value and distribution of images, for quantifying the heterogeneity of tissues.

PURPOSE

To investigate the feasibility of predicting MVI in HCC by analyzing the texture features of hepatic MR-enhanced images.

METHODS

110 patients with HCC who underwent MR-enhanced examinations were included in this study, were divided into MVI-positive group (n = 52) and MVI-negative group (n = 58) according to postoperative pathology. Clinical, pathological data and MR imaging features were collected. 11 texture parameters were selected from the gray histogram and gray level co-occurrence matrix (GLCM). Texture parameters of MR-enhanced images were calculated for statistical analysis.

RESULTS

There were statistically significant differences in tumor size, location, degree of differentiation, AFP level, signal, pseudocapsule, margin, peritumoral enhancement and intratumoral artery between MVI-positive group and MVI-negative group (P < 0.05). The AUC value of combining MR image features in prediction of MVI was 0.693(sensitivity and specificity: 53.8 %, 82.8 %, respectively). There were statistically significant differences in the texture parameters of GLCM between two groups (P < 0.05). The AUC value of combining texture parameters in prediction of MVI was 0.797 (sensitivity and specificity: 88.2 %, 62.7 %, respectively).

CONCLUSION

MR image features and texture analysis have certain predictive effect on MVI, which are mutually verified and complementary. The texture parameters of GLCM could reflect tumor heterogeneity, which have great potential to help with preoperative decision. The combination of MR image features and texture analysis may improve the efficiency in prediction of MVI.

Early Osteogenic Marker Expression in hMSCs Cultured onto Acid Etching-Derived Micro- and Nanotopography 3D-Printed Titanium Surfaces

Polyetheretherketone (PEEK) titanium composite (PTC) is a novel interbody fusion device that combines a PEEK core with titanium alloy (Ti6Al4V) endplates. The present study aimed to investigate the in vitro biological reactivity of human bone-marrow-derived mesenchymal stem cells (hBM-MSCs) to micro- and nanotopographies produced by an acid-etching process on the surface of 3D-printed PTC endplates. Optical profilometer and scanning electron microscopy were used to assess the surface roughness and identify the nano-features of etched or unetched PTC endplates, respectively. The viability, morphology and the expression of specific osteogenic markers were examined after 7 days of culture in the seeded cells. Haralick texture analysis was carried out on the unseeded endplates to correlate surface texture features to the biological data. The acid-etching process modified the surface roughness of the 3D-printed PTC endplates, creating micro- and nano-scale structures that significantly contributed to sustaining the viability of hBM-MSCs and triggering the expression of early osteogenic markers, such as alkaline phosphatase activity and bone-ECM protein production. Finally, the topography of 3D-printed PTC endplates influenced Haralick’s features, which in turn correlated with the expression of two osteogenic markers, osteopontin and osteocalcin. Overall, these data demonstrate that the acid-etching process of PTC endplates created a favourable environment for osteogenic differentiation of hBM-MSCs and may potentially have clinical benefit.

Development and validation of novel radiomics-based nomograms for the prediction of EGFR mutations and Ki-67 proliferation index in non-small cell lung cancer

BACKGROUND

We developed and validated novel radiomics-based nomograms to identify epidermal growth factor receptor (EGFR) mutations and the Ki-67 proliferation index of non-small cell lung cancer.

METHODS

We enrolled 132 patients with histologically verified non-small cell lung cancer from four hospital institutions who underwent computed tomography (CT) scans. EGFR mutations and the Ki-67 proliferation index were measured from tumor tissues. A total of 1,287 radiomic features were extracted, and a three-stage feature selection method was implemented to acquire the most valuable radiomic features. Finally, the radiomic scores and nomograms of the two tasks were established and tested. Receiver operating characteristic curves, calibration curves, and decision curves were used to evaluate their prediction performance and clinical utility.

RESULTS

In task [1], smoking status and histological type were significantly associated with EGFR mutations. After feature selection, 10 features were used to establish radiomic score, which showed good performance [area under the curve (AUC) =0.800] in the validation cohort. The radiomic nomogram had an AUC of 0.798 (95% CI: 0.664 to 0.931) with a C-index of 0.798 in the validation cohort. In task [2], gender, smoking status, histological type, and stage showed a significant correlation with Ki-67 proliferation index expression. A total of 28 features were selected to develop a radiomic score, with an AUC of 0.820 in the validation cohort. The final nomogram showed an AUC of 0.828 (95% CI: 0.703 to 0.953) with a C-index of 0.828 in the validation cohort.

CONCLUSIONS

EGFR mutations and Ki-67 proliferation index in non-small cell lung cancer can be predicted efficiently by the novel radiomic scores and nomograms.

Haralick's texture analysis to predict cellular proliferation on randomly oriented electrospun nanomaterials

Using a computer vision approach we have extracted the Haralick's texture features of randomly oriented electrospun nanomaterials in order to predict the proliferative behavior of cells which were subsequently seeded onto the nanosurfaces.

CT in the Differentiation of Gliomas from Brain Metastases: The Radiomics Analysis of the Peritumoral Zone

Due to their similar imaging features, high-grade gliomas (HGGs) and solitary brain metastases (BMs) can be easily misclassified. The peritumoral zone (PZ) of HGGs develops neoplastic cell infiltration, while in BMs the PZ contains pure vasogenic edema. As the two PZs cannot be differentiated macroscopically, this study investigated whether computed tomography (CT)-based texture analysis (TA) of the PZ can reflect the histological difference between the two entities. Thirty-six patients with solitary brain tumors (HGGs, n = 17; BMs, n = 19) that underwent CT examinations were retrospectively included in this pilot study. TA of the PZ was analyzed using dedicated software (MaZda version 5). Univariate, multivariate, and receiver operating characteristics analyses were used to identify the best-suited parameters for distinguishing between the two groups. Seven texture parameters were able to differentiate between HGGs and BMs with variable sensitivity (56.67–96.67%) and specificity (69.23–100%) rates. Their combined ability successfully identified HGGs with 77.9–99.2% sensitivity and 75.3–100% specificity. In conclusion, the CT-based TA can be a useful tool for differentiating between primary and secondary malignancies. The TA features indicate a more heterogenous content of the HGGs’ PZ, possibly due to the local infiltration of neoplastic cells.

Texture appearance model, a new model-based segmentation paradigm, application on the segmentation of lung nodule in the CT scan of the chest

Lung cancer causes more than one million deaths worldwide each year. Averages of 5-year survival rate of patients with Non-small cell lung cancer (NSCLC), which is the most common type of lung cancer, is 15%. Computer-Aided Detection (CAD) is a very important tool for identifying lung lesions in medical imaging. In general, the process line of a CAD system can be divided into four main stages: preprocessing, localization, feature extraction, and classification. As segmentation is required for localization in computer vision and medical image analysis, this step has become a major and challenging problem, and much research has been done on new segmentation techniques. In recent years, interest in model-based segmentation methods has increased, and the reason for this is even if some object information is lost, such gaps can be filled by using the previous information in the model. This paper proposed Texture Appearance Model (TAM), which is a new model-based method and segments all types of nodule areas accurately and efficiently, including juxta-pleural nodules, without separating the lung from the surrounding area in a CT scan of the lung. In this method, Texture Representation of Image (TRI) is obtained using tissue texture feature extraction and feature selection algorithms. The proposed method has been evaluated in 85 nodules of the dataset, received from the Iranian hospital, in which the ground-truth annotation by physicians and CT imaging data were provided. The results show that the proposed algorithm has an encouraging performance for distinguishing different types of nodules, including pleural, cavity and non-solid nodules, achieving an average dice similarity coefficient (DSC) of 84.75%.

Development of computer-aided model to differentiate COVID-19 from pulmonary edema in lung CT scan: EDECOVID-net

The efforts made to prevent the spread of COVID-19 face specific challenges in diagnosing COVID-19 patients and differentiating them from patients with pulmonary edema. Although systemically administered pulmonary vasodilators and acetazolamide are of great benefit for treating pulmonary edema, they should not be used to treat COVID-19 as they carry the risk of several adverse consequences, including worsening the matching of ventilation and perfusion, impaired carbon dioxide transport, systemic hypotension, and increased work of breathing. This study proposes a machine learning-based method (EDECOVID-net) that automatically differentiates the COVID-19 symptoms from pulmonary edema in lung CT scans using radiomic features. To the best of our knowledge, EDECOVID-net is the first method to differentiate COVID-19 from pulmonary edema and a helpful tool for diagnosing COVID-19 at early stages. The EDECOVID-net has been proposed as a new machine learning-based method with some advantages, such as having simple structure and few mathematical calculations. In total, 13 717 imaging patches, including 5759 COVID-19 and 7958 edema images, were extracted using a CT incision by a specialist radiologist. The EDECOVID-net can distinguish the patients with COVID-19 from those with pulmonary edema with an accuracy of 0.98. In addition, the accuracy of the EDECOVID-net algorithm is compared with other machine learning methods, such as VGG-16 (Acc = 0.94), VGG-19 (Acc = 0.96), Xception (Acc = 0.95), ResNet101 (Acc = 0.97), and DenseNet20l (Acc = 0.97).

Contrast-Enhanced Mammography and Radiomics Analysis for Noninvasive Breast Cancer Characterization: Initial Results

PURPOSE

To investigate the potential of contrast-enhanced mammography (CEM) and radiomics analysis for the noninvasive differentiation of breast cancer invasiveness, hormone receptor status, and tumor grade.

PROCEDURES

This retrospective study included 100 patients with 103 breast cancers who underwent pretreatment CEM. Radiomics analysis was performed using MAZDA software. Lesions were manually segmented. Radiomic features were derived from first-order histogram (HIS), co-occurrence matrix (COM), run length matrix (RLM), absolute gradient, autoregressive model, the discrete Haar wavelet transform (WAV), and lesion geometry. Fisher, probability of error and average correlation (POE+ACC), and mutual information (MI) coefficients informed feature selection. Linear discriminant analysis followed by k-nearest neighbor classification (with leave-one-out cross-validation) was used for pairwise texture-based separation of tumor invasiveness and hormone receptor status using histopathology as the standard of reference.

RESULTS

Radiomics analysis achieved the highest accuracies of 87.4 % for differentiating invasive from noninvasive cancers based on COM+HIS/MI, 78.4 % for differentiating HR positive from HR negative cancers based on COM+HIS/Fisher, 97.2 % for differentiating human epidermal growth factor receptor 2 (HER2)-positive/HR-negative from HER2-negative/HR-positive cancers based on RLM+WAV/MI, 100 % for differentiating triple-negative from triple-positive breast cancers mainly based on COM+WAV+HIS/POE+ACC, and 82.1 % for differentiating triple-negative from HR-positive cancers mainly based on WAV+HIS/Fisher. Accuracies for differentiating grade 1 vs. grades 2 and 3 cancers were 90 % for invasive cancers (based on COM/MI) and 100 % for noninvasive cancers (almost entirely based on COM/MI).

CONCLUSIONS

Radiomics analysis with CEM has potential for noninvasive differentiation of tumors with different degrees of invasiveness, hormone receptor status, and tumor grade.

Quantitative magnetic resonance imaging for focal liver lesions: bridging the gap between research and clinical practice

Magnetic resonance imaging (MRI) is highly important for the detection, characterization, and follow-up of focal liver lesions. Several quantitative MRI-based methods have been proposed in addition to qualitative imaging interpretation to improve the diagnostic work-up and prognostics in patients with focal liver lesions. This includes DWI with apparent diffusion coefficient measurements, intravoxel incoherent motion, perfusion imaging, MR elastography, and radiomics. Multiple research studies have reported promising results with quantitative MRI methods in various clinical settings. Nevertheless, applications in everyday clinical practice are limited. This review describes the basic principles of quantitative MRI-based techniques and discusses the main current applications and limitations for the assessment of focal liver lesions.

Ultrasonography in the Diagnosis of Adnexal Lesions: The Role of Texture Analysis

The classic ultrasonographic differentiation between benign and malignant adnexal masses encounters several limitations. Ultrasonography-based texture analysis (USTA) offers a new perspective, but its role has been incompletely evaluated. This study aimed to further investigate USTA’s capacity in differentiating benign from malignant adnexal tumors, as well as comparing the workflow and the results with previously-published research. A total of 123 adnexal lesions (benign, 88; malignant, 35) were retrospectively included. The USTA was performed on dedicated software. By applying three reduction techniques, 23 features with the highest discriminatory potential were selected. The features’ ability to identify ovarian malignancies was evaluated through univariate, multivariate, and receiver operating characteristics analyses, and also by the use of the k-nearest neighbor (KNN) classifier. Three parameters were independent predictors for ovarian neoplasms (sum variance, and two variations of the sum of squares). Benign and malignant lesions were differentiated with 90.48% sensitivity and 93.1% specificity by the prediction model (which included the three independent predictors), and with 71.43–80% sensitivity and 87.5–89.77% specificity by the KNN classifier. The USTA shows statistically significant differences between the textures of the two groups, but it is unclear whether the parameters can reflect the true histopathological characteristics of adnexal lesions.

Differentiation of small (≤ 3 cm) hepatocellular carcinomas from benign nodules in cirrhotic liver: the added additive value of MRI-based radiomics analysis to LI-RADS version 2018 algorithm

Background

Accurate characterization of small nodules in a cirrhotic liver is challenging. We aimed to determine the additive value of MRI-based radiomics analysis to Liver Imaging Reporting and Data System version 2018 (LI-RADS v 2018) algorithm in differentiating small (≤ 3 cm) hepatocellular carcinomas (HCCs) from benign nodules in cirrhotic liver.

Methods

In this retrospective study, 150 cirrhosis patients with histopathologically confirmed small liver nodules (HCC, 112; benign nodules, 44) were evaluated from January 2013 to October 2018. Based on the LI-RADS algorithm, a LI-RADS category was assigned for each lesion. A radiomics signature was generated based on texture features extracted from T1-weighted, T2W, and apparent diffusion coefficient (ADC) images by using the least absolute shrinkage and selection operator regression model. A nomogram model was developed for the combined diagnosis. Diagnostic performance was assessed using receiver operating characteristic curve (ROC) analysis.

Results

A radiomics signature consisting of eight features was significantly associated with the differentiation of HCCs from benign nodules. Both LI-RADS algorithm (area under ROC [A_z] = 0.898) and the MRI-Based radiomics signature (A_z = 0.917) demonstrated good discrimination, and the nomogram model showed a superior classification performance (A_z = 0.975). Compared with LI-RADS alone, the combined approach significantly improved the specificity (97.7% vs 81.8%, p = 0.030) and positive predictive value (99.1% vs 92.9%, p = 0.031) and afforded comparable sensitivity (97.3% vs 93.8%, p = 0.215) and negative predictive value (93.5% vs 83.7%, p = 0.188).

Conclusions

MRI-based radiomics analysis showed additive value to the LI-RADS v 2018 algorithm for differentiating small HCCs from benign nodules in the cirrhotic liver.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12876-021-01710-y.

Radiomic Signatures Derived from Diffusion-Weighted Imaging for the Assessment of Breast Cancer Receptor Status and Molecular Subtypes

Purpose

To compare annotation segmentation approaches and to assess the value of radiomics analysis applied to diffusion-weighted imaging (DWI) for evaluation of breast cancer receptor status and molecular subtyping.

Procedures

In this IRB-approved HIPAA-compliant retrospective study, 91 patients with treatment-naïve breast malignancies proven by image-guided breast biopsy, (luminal A, n = 49; luminal B, n = 8; human epidermal growth factor receptor 2 [HER2]-enriched, n = 11; triple negative [TN], n = 23) underwent multiparametric magnetic resonance imaging (MRI) of the breast at 3 T with dynamic contrast-enhanced MRI, T2-weighted and DW imaging. Lesions were manually segmented on high b-value DW images and segmentation ROIS were propagated to apparent diffusion coefficient (ADC) maps. In addition in a subgroup (n = 79) where lesions were discernable on ADC maps alone, these were also directly segmented there. To derive radiomics signatures, the following features were extracted and analyzed: first-order histogram (HIS), co-occurrence matrix (COM), run-length matrix (RLM), absolute gradient, autoregressive model (ARM), discrete Haar wavelet transform (WAV), and lesion geometry. Fisher, probability of error and average correlation, and mutual information coefficients were used for feature selection. Linear discriminant analysis followed by k-nearest neighbor classification with leave-one-out cross-validation was applied for pairwise differentiation of receptor status and molecular subtyping. Histopathologic results were considered the gold standard.

Results

For lesion that were segmented on DWI and segmentation ROIs were propagated to ADC maps the following classification accuracies > 90% were obtained: luminal B vs. HER2-enriched, 94.7 % (based on COM features); luminal B vs. others, 92.3 % (COM, HIS); and HER2-enriched vs. others, 90.1 % (RLM, COM). For lesions that were segmented directly on ADC maps, better results were achieved yielding the following classification accuracies: luminal B vs. HER2-enriched, 100 % (COM, WAV); luminal A vs. luminal B, 91.5 % (COM, WAV); and luminal B vs. others, 91.1 % (WAV, ARM, COM).

Conclusions

Radiomic signatures from DWI with ADC mapping allows evaluation of breast cancer receptor status and molecular subtyping with high diagnostic accuracy. Better classification accuracies were obtained when breast tumor segmentations could be performed on ADC maps.

Preoperative classification of primary and metastatic liver cancer via machine learning-based ultrasound radiomics

OBJECTIVE

To investigate the application of machine learning-based ultrasound radiomics in preoperative classification of primary and metastatic liver cancer.

METHODS

Data of 114 consecutive histopathologically confirmed patients with liver cancer from January 2018 to November 2019 were retrospectively analyzed. All patients underwent liver ultrasonography within 1 week before hepatectomy or fine-needle biopsy. The liver lesions were manually segmented by two experts using ITK-SNAP software. Seven categories of radiomics features, including first-order, two-dimensional shape, gray-level co-occurrence matrices, gray-level run-length matrix, gray-level size-zone matrix, neighboring gray tone difference matrix, and gray-level dependence matrix, were extracted on the Pyradiomics platform. Fourteen filters were applied to the original images, and derived images were obtained. Then, the dimensions of radiomics features were reduced by least absolute shrinkage and selection operator (Lasso) method. Finally, k-nearest neighbor (KNN), logistic regression (LR), multilayer perceptron (MLP), random forest (RF), and support vector machine (SVM) were employed to distinguish primary liver cancer from metastatic liver cancer by a fivefold cross-validation strategy. The performance of the established model was mainly evaluated by the area under the receiver operating characteristic (ROC) curve (AUC) and accuracy.

RESULTS

One thousand four hundred nine radiomics features were extracted from the original images and/or derived images for each patient. The mentioned five machine learning classifiers were able to differentiate primary liver cancer from metastatic liver cancer. LR outperformed other classifiers, with the accuracy of 0.843 ± 0.078 (AUC, 0.816 ± 0.088; sensitivity, 0.768 ± 0.232; specificity, 0.880 ± 0.117).

CONCLUSIONS

Machine learning-based ultrasound radiomics features are able to non-invasively distinguish primary liver tumors from metastatic liver tumors.

KEY POINTS

• Ultrasound-based radiomics was initially used for preoperative classification of primary versus metastatic liver cancer. • Multiple machine learning-based algorithms with cross-validation strategy were applied to extract machine learning-based ultrasound radiomics features. • Distinction between primary and metastatic tumors was obtained with a sensitivity of 0.768 and a specificity of 0.880.

Recent advances in medical image processing for the evaluation of chronic kidney disease

Assessment of renal function and structure accurately remains essential in the diagnosis and prognosis of Chronic Kidney Disease (CKD). Advanced imaging, including Magnetic Resonance Imaging (MRI), Ultrasound Elastography (UE), Computed Tomography (CT) and scintigraphy (PET, SPECT) offers the opportunity to non-invasively retrieve structural, functional and molecular information that could detect changes in renal tissue properties and functionality. Currently, the ability of artificial intelligence to turn conventional medical imaging into a full-automated diagnostic tool is widely investigated. In addition to the qualitative analysis performed on renal medical imaging, texture analysis was integrated with machine learning techniques as a quantification of renal tissue heterogeneity, providing a promising complementary tool in renal function decline prediction. Interestingly, deep learning holds the ability to be a novel approach of renal function diagnosis. This paper proposes a survey that covers both qualitative and quantitative analysis applied to novel medical imaging techniques to monitor the decline of renal function. First, we summarize the use of different medical imaging modalities to monitor CKD and then, we show the ability of Artificial Intelligence (AI) to guide renal function evaluation from segmentation to disease prediction, discussing how texture analysis and machine learning techniques have emerged in recent clinical researches in order to improve renal dysfunction monitoring and prediction. The paper gives a summary about the role of AI in renal segmentation.

Radiomics of high-resolution computed tomography for the differentiation between cholesteatoma and middle ear inflammation: effects of post-reconstruction methods in a dual-center study

Objectives

To evaluate the performance of radiomic features extracted from high-resolution computed tomography (HRCT) for the differentiation between cholesteatoma and middle ear inflammation (MEI), and to investigate the impact of post-reconstruction harmonization and data resampling.

Methods

One hundred patients were included in this retrospective dual-center study: 48 with histology-proven cholesteatoma (center A: 23; center B: 25) and 52 with MEI (A: 27; B: 25). Radiomic features (co-occurrence and run-length matrix, absolute gradient, autoregressive model, Haar wavelet transform) were extracted from manually defined 2D-ROIs. The ten best features for lesion differentiation were selected using probability of error and average correlation coefficients. A multi-layer perceptron feed-forward artificial neural network (MLP-ANN) was used for radiomics-based classification, with histopathology serving as the reference standard (70% of cases for training, 30% for validation). The analysis was performed five times each on (a) unmodified data and on data that were (b) resampled to the same matrix size, and (c) corrected for acquisition protocol differences using ComBat harmonization.

Results

Using unmodified data, the MLP-ANN classification yielded an overall median area under the receiver operating characteristic curve (AUC) of 0.78 (0.72–0.84). Using original data from center A and resampled data from center B, an overall median AUC of 0.88 (0.82–0.99) was yielded, while using ComBat harmonized data, an overall median AUC of 0.89 (0.79–0.92) was revealed.

Conclusion

Radiomic features extracted from HRCT differentiate between cholesteatoma and MEI. When using multi-centric data obtained with differences in CT acquisition parameters, data resampling and ComBat post-reconstruction harmonization clearly improve radiomics-based lesion classification.

Key Points

• Unenhanced high-resolution CT coupled with radiomics analysis may be useful for the differentiation between cholesteatoma and middle ear inflammation.

• Pooling of data extracted from inhomogeneous CT datasets does not appear meaningful without further post-processing.

• When using multi-centric CT data obtained with differences in acquisition parameters, post-reconstruction harmonization and data resampling clearly improve radiomics-based soft-tissue differentiation.

Tumor heterogeneity for differentiation between liver tumors and normal liver tissue in 18F-FDG PET/CT

AIM

 Malignancies show higher spatial heterogeneity than normal tissue. We investigated, if textural parameters from FDG PET describing the heterogeneity function as tool to differentiate between tumor and normal liver tissue.

METHODS

 FDG PET/CT scans of 80 patients with liver metastases and 80 patients with results negative upper abdominal organs were analyzed. Metastases and normal liver tissue were analyzed drawing up to three VOIs with a diameter of 25 mm in healthy liver tissue of the tumoral affected and results negative liver, whilst up to 3 metastases per patient were delineated. Within these VOIs 30 different textural parameters were calculated as well as SUV. The parameters were compared in terms of intra-patient and inter-patient variability (2-sided t test). ROC analysis was performed to analyze predictive power and cut-off values.

RESULTS

 28 textural parameters differentiated healthy and pathological tissue (p < 0.05) with high sensitivity and specificity. SUV showed ability to differentiate but with a lower significance. 15 textural parameters as well as SUV showed a significant variation between healthy tissues out of tumour infested and negative livers. Mean intra- and inter-patient variability of metastases were found comparable or lower for 6 of the textural features than the ones of SUV. They also showed good values of mean intra- and inter-patient variability of VOIs drawn in liver tissue of patients with metastases and of results negative ones.

CONCLUSION

 Heterogeneity parameters assessed in FDG PET are promising to classify tissue and differentiate malignant lesions usable for more personalized treatment planning, therapy response evaluation and precise delineation of tumors for target volume determination as part of radiation therapy planning.

Differentiation of Endometriomas from Ovarian Hemorrhagic Cysts at Magnetic Resonance: The Role of Texture Analysis

Background and Objectives: To assess ovarian cysts with texture analysis (TA) in magnetic resonance (MRI) images for establishing a differentiation criterion for endometriomas and functional hemorrhagic cysts (HCs) that could potentially outperform their classic MRI diagnostic features. Materials and Methods: Forty-three patients with known ovarian cysts who underwent MRI were retrospectively included (endometriomas, n = 29; HCs, n = 14). TA was performed using dedicated software based on T2-weighted images, by incorporating the whole lesions in a three-dimensional region of interest. The most discriminative texture features were highlighted by three selection methods (Fisher, probability of classification error and average correlation coefficients, and mutual information). The absolute values of these parameters were compared through univariate, multivariate, and receiver operating characteristic analyses. The ability of the two classic diagnostic signs ("T2 shading" and "T2 dark spots") to diagnose endometriomas was assessed by quantifying their sensitivity (Se) and specificity (Sp), following their conventional assessment on T1-and T2-weighted images by two radiologists. Results: The diagnostic power of the one texture parameter that was an independent predictor of endometriomas (entropy, 75% Se and 100% Sp) and of the predictive model composed of all parameters that showed statistically significant results at the univariate analysis (100% Se, 100% Sp) outperformed the ones shown by the classic MRI endometrioma features ("T2 shading", 75.86% Se and 35.71% Sp; "T2 dark spots", 55.17% Se and 64.29% Sp). Conclusion: Whole-lesion MRI TA has the potential to offer a superior discrimination criterion between endometriomas and HCs compared to the classic evaluation of the two lesions' MRI signal behaviors.

Differentiating High-Grade Gliomas from Brain Metastases at Magnetic Resonance: The Role of Texture Analysis of the Peritumoral Zone

High-grade gliomas (HGGs) and solitary brain metastases (BMs) have similar imaging appearances, which often leads to misclassification. In HGGs, the surrounding tissues show malignant invasion, while BMs tend to displace the adjacent area. The surrounding edema produced by the two cannot be differentiated by conventional magnetic resonance (MRI) examinations. Forty-two patients with pathology-proven brain tumors who underwent conventional pretreatment MRIs were retrospectively included (HGGs, n = 16; BMs, n = 26). Texture analysis of the peritumoral zone was performed on the T2-weighted sequence using dedicated software. The most discriminative texture features were selected using the Fisher and the probability of classification error and average correlation coefficients. The ability of texture parameters to distinguish between HGGs and BMs was evaluated through univariate, receiver operating, and multivariate analyses. The first percentile and wavelet energy texture parameters were independent predictors of HGGs (75–87.5% sensitivity, 53.85–88.46% specificity). The prediction model consisting of all parameters that showed statistically significant results at the univariate analysis was able to identify HGGs with 100% sensitivity and 66.7% specificity. Texture analysis can provide a quantitative description of the peritumoral zone encountered in solitary brain tumors, that can provide adequate differentiation between HGGs and BMs.

Cervical spine osteoradionecrosis or bone metastasis after radiotherapy for nasopharyngeal carcinoma? The MRI-based radiomics for characterization

Background

To develop and validate an MRI-based radiomics nomogram for differentiation of cervical spine ORN from metastasis after radiotherapy (RT) in nasopharyngeal carcinoma (NPC).

Methods

A radiomics nomogram was developed in a training set that comprised 46 NPC patients after RT with 95 cervical spine lesions (ORN, n = 51; metastasis, n = 44), and data were gathered from January 2008 to December 2012. 279 radiomics features were extracted from the axial contrast-enhanced T1-weighted image (CE-T1WI). A radiomics signature was created by using the least absolute shrinkage and selection operator (LASSO) algorithm. A nomogram model was developed based on the radiomics scores. The performance of the nomogram was determined in terms of its discrimination, calibration, and clinical utility. An independent validation set contained 25 consecutive patients with 47 lesions (ORN, n = 25; metastasis, n = 22) from January 2013 to December 2015.

Results

The radiomics signature that comprised eight selected features was significantly associated with the differentiation of cervical spine ORN and metastasis. The nomogram model demonstrated good calibration and discrimination in the training set [AUC, 0.725; 95% confidence interval (CI), 0.622–0.828] and the validation set (AUC, 0.720; 95% CI, 0.573–0.867). The decision curve analysis indicated that the radiomics nomogram was clinically useful.

Conclusions

MRI-based radiomics nomogram shows potential value to differentiate cervical spine ORN from metastasis after RT in NPC.

Radiomics for Tumor Characterization in Breast Cancer Patients: A Feasibility Study Comparing Contrast-Enhanced Mammography and Magnetic Resonance Imaging

The aim of our intra-individual comparison study was to investigate and compare the potential of radiomics analysis of contrast-enhanced mammography (CEM) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) of the breast for the non-invasive assessment of tumor invasiveness, hormone receptor status, and tumor grade in patients with primary breast cancer. This retrospective study included 48 female patients with 49 biopsy-proven breast cancers who underwent pretreatment breast CEM and MRI. Radiomics analysis was performed by using MaZda software. Radiomics parameters were correlated with tumor histology (invasive vs. non-invasive), hormonal status (HR+ vs. HR-), and grading (low grade G1 + G2 vs. high grade G3). CEM radiomics analysis yielded classification accuracies of up to 92% for invasive vs. non-invasive breast cancers, 95.6% for HR+ vs. HR- breast cancers, and 77.8% for G1 + G2 vs. G3 invasive cancers. MRI radiomics analysis yielded classification accuracies of up to 90% for invasive vs. non-invasive breast cancers, 82.6% for HR+ vs. HR- breast cancers, and 77.8% for G1+G2 vs. G3 cancers. Preliminary results indicate a potential of both radiomics analysis of DCE-MRI and CEM for non-invasive assessment of tumor-invasiveness, hormone receptor status, and tumor grade. CEM may serve as an alternative to MRI if MRI is not available or contraindicated.

A rapid volume of interest-based approach of radiomics analysis of breast MRI for tumor decoding and phenotyping of breast cancer

Background

Recently, radiomics has emerged as a non-invasive, imaging-based tissue characterization method in multiple cancer types. One limitation for robust and reproducible analysis lies in the inter-reader variability of the tumor annotations, which can potentially cause differences in the extracted feature sets and results. In this study, the diagnostic potential of a rapid and clinically feasible VOI (Volume of Interest)-based approach to radiomics is investigated to assess MR-derived parameters for predicting molecular subtype, hormonal receptor status, Ki67- and HER2-Expression, metastasis of lymph nodes and lymph vessel involvement as well as grading in patients with breast cancer.

Methods

A total of 98 treatment-naïve patients (mean 59.7 years, range 28.0–89.4) with BI-RADS 5 and 6 lesions who underwent a dedicated breast MRI prior to therapy were retrospectively included in this study. The imaging protocol comprised dynamic contrast-enhanced T1-weighted imaging and T2-weighted imaging. Tumor annotations were obtained by drawing VOIs around the primary tumor lesions followed by thresholding. From each segmentation, 13.118 quantitative imaging features were extracted and analyzed with machine learning methods. Validation was performed by 5-fold cross-validation with 25 repeats.

Results

Predictions for molecular subtypes obtained AUCs of 0.75 (HER2-enriched), 0.73 (triple-negative), 0.65 (luminal A) and 0.69 (luminal B). Differentiating subtypes from one another was highest for HER2-enriched vs triple-negative (AUC 0.97), followed by luminal B vs triple-negative (0.86). Receptor status predictions for Estrogen Receptor (ER), Progesterone Receptor (PR) and Hormone receptor positivity yielded AUCs of 0.67, 0.69 and 0.69, while Ki67 and HER2 Expressions achieved 0.81 and 0.62. Involvement of the lymph vessels could be predicted with an AUC of 0.8, while lymph node metastasis yielded an AUC of 0.71. Models for grading performed similar with an AUC of 0.71 for Elston-Ellis grading and 0.74 for the histological grading.

Conclusion

Our preliminary results of a rapid approach to VOI-based tumor-annotations for radiomics provides comparable results to current publications with the perks of clinical suitability, enabling a comprehensive non-invasive platform for breast tumor decoding and phenotyping.

A radiomics signature to identify malignant and benign liver tumors on plain CT images

BACKGROUND

In regular examinations, it may be difficult to visually identify benign and malignant liver tumors based on plain computed tomography (CT) images. RCAD (radiomics-based computer-aided diagnosis) has proven to be helpful and provide interpretability in clinical use.

OBJECTIVE

This work aims to develop a CT-based radiomics signature and investigate its correlation with malignant/benign liver tumors.

METHODS

We retrospectively analyzed 168 patients of hepatocellular carcinoma (malignant) and 117 patients of hepatic hemangioma (benign). Texture features were extracted from plain CT images and used as candidate features. A radiomics signature was developed from the candidate features. We performed logistic regression analysis and used a multiple-regression coefficient (termed as R) to assess the correlation between the developed radiomics signature and malignant/benign liver tumors. Finally, we built a logistic regression model to classify benign and malignant liver tumors.

RESULTS

Thirteen features were chosen from 1223 candidate features to constitute the radiomics signature. The logistic regression analysis achieved an R = 0.6745, which was much larger than Rα = 0.3703 (the critical value of R at significant level α = 0.001). The logistic regression model achieved an average AUC of 0.87.

CONCLUSIONS

The developed radiomics signature was statistically significantly correlated with malignant/benign liver tumors (p < 0.001). It has potential to help enhance physicians' diagnostic abilities and play an important role in RCADs.

The feasibility of differentiating colorectal cancer from normal and inflammatory thickening colon wall using CT texture analysis

To investigate the diagnostic value of texture analysis (TA) for differentiating between colorectal cancer (CRC), colonic lesions caused by inflammatory bowel disease (IBD), and normal thickened colon wall (NTC) on computed tomography (CT) and assess which scanning phase has the highest differential diagnostic value. In all, 107 patients with CRC, 113 IBD patients with colonic lesions, and 96 participants with NTC were retrospectively enrolled. All subjects underwent multiphase CT examination, including pre-contrast phase (PCP), arterial phase (AP), and portal venous phase (PVP) scans. Based on these images, classification by TA and visual classification by radiologists were performed to discriminate among the three tissue types. The performance of TA and visual classification was compared. Precise TA classification results (error, 2.03–12.48%) were acquired by nonlinear discriminant analysis for CRC, IBD and NTC, regardless of phase or feature selection. PVP images showed a better ability to discriminate the three tissues by comprising the three scanning phases. TA showed significantly better performance in discriminating CRC, IBD and NTC than visual classification for residents, but there was no significant difference in classification between TA and experienced radiologists. TA could provide useful quantitative information for the differentiation of CRC, IBD and NTC on CT, particularly in PVP images.

Improving the malignancy characterization of hepatocellular carcinoma using deeply supervised cross modal transfer learning for non-enhanced MR

The malignancy characterization of hepatocellular carcinoma (HCC) is remarkably significant in clinical practice. In this work, we propose a deeply supervised cross modal transfer learning method to remarkably improve the malignancy characterization of HCC based on non-enhanced MR. First, we use samples of non-enhanced and contrast-enhanced MR for pre-training a deep learning network to learn the cross modal relationship between the non-enhanced modal and enhanced modal. Then, the parameters of the pre-trained across modal representation are transferred to a second deep learning model for fine-tuning based only on non-enhanced MR for malignancy characterization of HCC. Specifically, a deeply supervised network is designed to enhance the stability of the second deep learning model and further improve the performance of lesion characterization. Importantly, only non-enhanced MR of HCC is required for the malignancy characterization in the training and test phase of the second deep learning model. Experiments of one hundred and fifteen clinical HCCs demonstrate that the proposed deeply supervised cross modal transfer learning method can significantly improve the malignancy characterization of HCC based on non-enhanced MR.

Differentiation of Small Hepatocellular Carcinoma From Dysplastic Nodules in Cirrhotic Liver: Texture Analysis Based on MRI Improved Performance in Comparison Over Gadoxetic Acid-Enhanced MR and Diffusion-Weighted Imaging

Background: Accurate characterization of small (3 cm) hepatocellular carcinoma (sHCC) and dysplastic nodules (DNs) in cirrhotic liver is challenging. We aimed to investigate whether texture analysis (TA) based on T2-weighted images (T2WI) is superior to qualitative diagnosis using gadoxetic acid-enhanced MR imaging (Gd-EOB-MRI) and diffusion-weighted imaging (DWI) for distinguishing sHCC from DNs in cirrhosis.

Materials and methods: Sixty-eight patients with 73 liver nodules (46 HCCs, 27 DNs) pathologically confirmed by operation were included. For imaging diagnosis, three sets of images were reviewed by two experienced radiologists in consensus: a Gd-EOB-MRI set, a DWI set, and a combined set (combination of Gd-EOB-MRI and DWI). For TA, 279 texture features resulting from T2WI were extracted for each lesion. The performance of each approach was evaluated by a receiver operating characteristic analysis. The area under the receiver operating characteristic curve (A_z), sensitivity, specificity, and accuracy were determined.

Results: The performance of TA (A_z = 0.96) was significantly higher than that of imaging diagnosis using Gd-EOB-MRI set (A_z = 0.86) or DWI set (A_z = 0.80) alone in differentiation of sHCC from DNs (P = 0.008 and 0.025, respectively). The combination of Gd-EOB-MRI and DWI showed a greater sensitivity (95.6%) but reduced specificity (66.7%). The specificity of TA (92.6%) was significantly higher than that of the combined set (P < 0.001), but no significant difference was observed in sensitivity (97.8 vs. 95.6%, P = 0.559).

Conclusion: TA-based T2WI showed a better classification performance than that of qualitative diagnosis using Gd-EOB-MRI and DW imaging in differentiation of sHCCs from DNs in cirrhotic liver. TA-based MRI may become a potential imaging biomarker for the early differentiation HCCs from DNs in cirrhosis.

Machine Learning to Differentiate T2-Weighted Hyperintense Uterine Leiomyomas from Uterine Sarcomas by Utilizing Multiparametric Magnetic Resonance Quantitative Imaging Features

RATIONALE AND OBJECTIVE

Uterine leiomyomas with high signal intensity on T2-weighted imaging (T2WI) can be difficult to distinguish from sarcomas. This study assessed the feasibility of using machine learning to differentiate uterine sarcomas from leiomyomas with high signal intensity on T2WI on multiparametric magnetic resonance imaging.

MATERIALS AND METHODS

This retrospective study included 80 patients (50 with benign leiomyoma and 30 with uterine sarcoma) who underwent pelvic 3 T magnetic resonance imaging examination for the evaluation of uterine myometrial smooth muscle masses with high signal intensity on T2WI. We used six machine learning techniques to develop prediction models based on 12 texture parameters on T1WI and T2WI, apparent diffusion coefficient maps, and contrast-enhanced T1WI, as well as tumor size and age. We calculated the areas under the curve (AUCs) using receiver-operating characteristic analysis for each model by 10-fold cross-validation and compared these to those for two board-certified radiologists.

RESULTS

The eXtreme Gradient Boosting model gave the highest AUC (0.93), followed by the random forest, support vector machine, multilayer perceptron, k-nearest neighbors, and logistic regression models. Age was the most important factor for differentiation (leiomyoma 44.9 ± 11.1 years; sarcoma 58.9 ± 14.7 years; p < 0.001). The AUC for the eXtreme Gradient Boosting was significantly higher than those for both radiologists (0.93 vs 0.80 and 0.68, p = 0.03 and p < 0.001, respectively).

CONCLUSION

Machine learning outperformed experienced radiologists in the differentiation of uterine sarcomas from leiomyomas with high signal intensity on T2WI.

Radiomic signatures with contrast-enhanced magnetic resonance imaging for the assessment of breast cancer receptor status and molecular subtypes: initial results

BACKGROUND

To evaluate the diagnostic performance of radiomic signatures extracted from contrast-enhanced magnetic resonance imaging (CE-MRI) for the assessment of breast cancer receptor status and molecular subtypes.

METHODS

One hundred and forty-three patients with biopsy-proven breast cancer who underwent CE-MRI at 3 T were included in this IRB-approved HIPAA-compliant retrospective study. The training dataset comprised 91 patients (luminal A, n = 49; luminal B, n = 8; HER2-enriched, n = 11; triple negative, n = 23), while the validation dataset comprised 52 patients from a second institution (luminal A, n = 17; luminal B, n = 17; triple negative, n = 18). Radiomic analysis of manually segmented tumors included calculation of features derived from the first-order histogram (HIS), co-occurrence matrix (COM), run-length matrix (RLM), absolute gradient (GRA), autoregressive model (ARM), discrete Haar wavelet transform (WAV), and lesion geometry (GEO). Fisher, probability of error and average correlation (POE + ACC), and mutual information coefficients were used for feature selection. Linear discriminant analysis followed by k-nearest neighbor classification (with leave-one-out cross-validation) was used for pairwise radiomic-based separation of receptor status and molecular subtypes. Histopathology served as the standard of reference.

RESULTS

In the training dataset, radiomic signatures yielded the following accuracies > 80%: luminal B vs. luminal A, 84.2% (mainly based on COM features); luminal B vs. triple negative, 83.9% (mainly based on GEO features); luminal B vs. all others, 89% (mainly based on COM features); and HER2-enriched vs. all others, 81.3% (mainly based on COM features). Radiomic signatures were successfully validated in the separate validation dataset for luminal A vs. luminal B (79.4%) and luminal B vs. triple negative (77.1%).

CONCLUSIONS

In this preliminary study, radiomic signatures with CE-MRI enable the assessment of breast cancer receptor status and molecular subtypes with high diagnostic accuracy. These results need to be confirmed in future larger studies.

Two-dimensional Texture Analysis Based on CT Images to Differentiate Pancreatic Lymphoma and Pancreatic Adenocarcinoma: A Preliminary Study

OBJECTIVE

To retrospectively assess the diagnostic performance of texture analysis and characteristics of CT images for the discrimination of pancreatic lymphoma (PL) from pancreatic adenocarcinoma (PA).

METHODS

Fifteen patients with pathologically proved PL were compared with 30 age-matched controls with PA in a 1:2 ratio. Patients underwent a CT scan with three phases including the precontrast phase, the arterial phase, and the portal vein phase. The regions of interest of PA and PL were drawn and analyzed to derive texture parameters with MaZda software. Texture features and CT characteristics were selected for the discrimination of PA and PL by the least absolute shrinkage and selection operator and logistic regression analysis. Receiver operating characteristic analysis was performed to assess the diagnostic performance of texture analysis and characteristics of CT images.

RESULTS

Sixty texture features were obtained by MaZda. Of these, four texture features were selected by least absolute shrinkage and selection operator. Following this, three texture features and nine CT characteristics were excluded by logistic regression analysis. Finally, "S(5, -5)SumAverg" (texture feature) and "Size" (CT characteristic) were selected for the receiver operating characteristic analysis. The AUC of "S(5, -5)SumAverg" and "Size" were to be 0.704 and 0.821, respectively, with no significance between them (p = 0.3064).

CONCLUSION

Two-dimensional texture analysis is a quantitative method for differential diagnosis of PL from PA. The diagnostic performance of both texture analysis and CT characteristics was similar.

Grading of hepatocellular carcinoma based on diffusion weighted images with multiple b-values using convolutional neural networks

PURPOSE

To effectively grade hepatocellular carcinoma (HCC) based on deep features derived from diffusion weighted images (DWI) with multiple b-values using convolutional neural networks (CNN).

MATERIALS AND METHODS

Ninety-eight subjects with 100 pathologically confirmed HCC lesions from July 2012 to October 2018 were included in this retrospective study, including 47 low-grade and 53 high-grade HCCs. DWI was performed for each subject with a 3.0T MR scanner in a breath-hold routine with three b-values (0,100, and 600 s/mm² ). First, logarithmic transformation was performed on original DWI images to generate log maps (logb0, logb100, and logb600). Then, a resampling method was performed to extract multiple 2D axial planes of HCCs from the log map to increase the dataset for training. Subsequently, 2D CNN was used to extract deep features of the log map for HCCs. Finally, fusion of deep features derived from three b-value log maps was conducted for HCC malignancy classification. Specifically, a deeply supervised loss function was devised to further improve the performance of lesion characterization. The data set was split into two parts: the training and validation set (60 HCCs) and the fixed test set (40 HCCs). Four-fold cross validation with 10 repetitions was performed to assess the performance of deep features extracted from single b-value images for HCC grading using the training and validation set. Receiver operating characteristic curve (ROC) and area under the curve (AUC) values were used to assess the characterization performance of the proposed deep feature fusion method to differentiate low-grade and high-grade in the fixed test set.

RESULTS

The proposed fusion of deep features derived from logb0, logb100, and logb600 with deeply supervised loss function generated the highest accuracy for HCC grading (80%), thus outperforming the method of deep feature derived from the ADC map directly (72.5%), the original b0 (65%), b100 (68%), and b600 (70%) images. Furthermore, AUC values of the deep features of the ADC map, the deep feature fusion with concatenation, and the proposed deep feature fusion with deeply supervised loss function were 0.73, 0.78, and 0.83, respectively.

CONCLUSION

The proposed fusion of deep features derived from the logarithm of the three b-value images yields high performance for HCC grading, thus providing a promising approach for the assessment of DWI in lesion characterization.

Automatic classification of focal liver lesions based on MRI and risk factors

OBJECTIVES

Accurate classification of focal liver lesions is an important part of liver disease diagnostics. In clinical practice, the lesion type is often determined from the abdominal MR examination, which includes T2-weighted and dynamic contrast enhanced (DCE) MR images. To date, only T2-weighted images are exploited for automatic classification of focal liver lesions. In this study additional MR sequences and risk factors are used for automatic classification to improve the results and to make a step forward to a clinically useful aid for radiologists.

MATERIALS AND METHODS

Clinical MRI data sets of 95 patients with in total 125 benign lesions (40 adenomas, 29 cysts and 56 hemangiomas) and 88 malignant lesions (30 hepatocellular carcinomas (HCC) and 58 metastases) were included in this study. Contrast curve, gray level histogram, and gray level co-occurrence matrix texture features were extracted from the DCE-MR and T2-weighted images. In addition, risk factors including the presence of steatosis, cirrhosis, and a known primary tumor were used as features. Fifty features with the highest ANOVA F-score were selected and fed to an extremely randomized trees classifier. The classifier evaluation was performed using the leave-one-out principle and receiver operating characteristic (ROC) curve analysis.

RESULTS

The overall accuracy for the classification of the five major focal liver lesion types is 0.77. The sensitivity/specificity is 0.80/0.78, 0.93/0.93, 0.84/0.82, 0.73/0.56, and 0.62/0.77 for adenoma, cyst, hemangioma, HCC, and metastasis, respectively.

CONCLUSION

The proposed classification system using features derived from clinical DCE-MR and T2-weighted images, with additional risk factors is able to differentiate five common types of lesions and is a step forward to a clinically useful aid for focal liver lesion diagnosis.

The association of intra-therapeutic heterogeneity of somatostatin receptor expression with morphological treatment response in patients undergoing PRRT with [177Lu]-DOTATATE

Aim

Purpose of this study was to evaluate the association of the spatial heterogeneity (asphericity, ASP) in intra-therapeutic SPECT/ CT imaging of somatostatin receptor (SSR) positive metastatic gastroenteropancreatic neuroendocrine neoplasms (GEP-NEN) for morphological treatment response to peptide receptor radionuclide therapy (PRRT). Secondly, we correlated ASP derived form a pre-therapeutic OctreoScan (ASP_[In]) and an intra-therapeutic [¹⁷⁷Lu]-SPECT/CT (ASP_[Lu]).

Materials and methods

Data from first therapy cycle [¹⁷⁷Lu-DOTA⁰-Tyr³]octreotate ([¹⁷⁷Lu]-DOTATATE)-PRRT was retrospectively analyzed in 33 patients (m = 20; w = 13; median age, 72 [46–88] years). The evaluation of response to PRRT was performed according to RECIST 1.1 in responding lesions [RL (SD, PR, CR), n = 104] and non-responding lesions [NRL (PD), n = 27]. The association of SSR tumor heterogeneity with morphological response was evaluated by Kruskal-Wallis test and receiver operating characteristic curve (ROC). The optimal threshold for separation (RL vs. NRL) was calculated using the Youden-index. Relationship between pre- and intra-therapeutic ASP was determined with Spearman’s rank correlation coefficient (ρ) and Bland-Altman plots.

Results

A total of 131 lesions (liver: n = 59, lymph nodes: n = 48, bone: n = 19, pancreas: n = 5) were analyzed. Lesions with higher ASP values showed a significantly poorer response to PRRT (PD, median: 11.3, IQR: 8.5–15.5; SD, median: 3.4, IQR: 2.1–4.5; PR, median 1.7, IQR: 0.9–2.8; CR, median: 0.5, IQR: 0.0–1.3); Kruskal-Wallis, p<0.001). ROC analyses revealed a significant separation between RL and NRL for ASP after 4 months (AUC 0.85, p<0.001) and after 12 months (AUC 0.94, p<0.001). The optimal threshold for ASP was >5.45% (sensitivity 96% and specificity 82%). The correlation coefficient of pre- and intra-therapeutic ASP revealed ρ = 0.72 (p <0.01). The mean absolute difference between ASP_[In] and ASP_[Lu] was -0.04 (95% Limits of Agreement, -6.1–6.0).

Conclusion

Pre- and intra-therapeutic ASP shows a strong correlation and might be an useful tool for therapy monitoring.

Gray-level invariant Haralick texture features

Haralick texture features are common texture descriptors in image analysis. To compute the Haralick features, the image gray-levels are reduced, a process called quantization. The resulting features depend heavily on the quantization step, so Haralick features are not reproducible unless the same quantization is performed. The aim of this work was to develop Haralick features that are invariant to the number of quantization gray-levels. By redefining the gray-level co-occurrence matrix (GLCM) as a discretized probability density function, it becomes asymptotically invariant to the quantization. The invariant and original features were compared using logistic regression classification to separate two classes based on the texture features. Classifiers trained on the invariant features showed higher accuracies, and had similar performance when training and test images had very different quantizations. In conclusion, using the invariant Haralick features, an image pattern will give the same texture feature values independent of image quantization.

Value of Texture Analysis on Gadoxetic Acid-Enhanced MRI for Differentiating Hepatocellular Adenoma From Focal Nodular Hyperplasia

OBJECTIVE

The objective of our study was to assess the diagnostic performance of texture analysis (TA) on gadoxetic acid-enhanced MR images for differentiation of hepatocellular adenoma (HCA) from focal nodular hyperplasia (FNH).

MATERIALS AND METHODS

This study included 40 patients (39 women and one man) with 51 HCAs and 28 patients (27 women and one man) with 32 FNH lesions. All lesions were histologically proven with preoperative MRI performed with gadoxetic acid. Two readers reviewed all the imaging sequences to assess the qualitative MRI characteristics. The T2-weighted fast spin-echo, hepatic arterial phase (HAP), and hepatobiliary phase (HBP) sequences were used for TA. Textural features were extracted using commercially available software (TexRAD). The differences in distributions of TA parameters of FNHs and HCAs were assessed using the Mann-Whitney U test. Area under the ROC curve (AUROC) values were calculated for statistically significant features. A logistic regression analysis was conducted to explore the added value of TA. A p value < 0.002 was considered statistically significant after Bonferroni correction for multiple comparisons.

RESULTS

Multiple TA parameters showed a statistically different distribution in HCA and FNH including skewness on T2-weighted imaging, skewness on HAP imaging, skewness on HBP imaging, and entropy on HBP imaging (p < 0.001). Skewness on HBP imaging showed the largest AUROC (0.869; 95% CI, 0.777-0.933). A skewness value on HBP imaging of greater than -0.06 had a sensitivity of 72.5% and a specificity of 90.6% for the diagnosis of HCA. Six of 51 (11.8%) HCAs lacked hypointensity on HBP imaging. A binary logistic regression analysis including hypointensity on HBP imaging and the statistically significant TA parameters yielded an AUROC of 0.979 for the diagnosis of HCA and correctly predicted 96.4% of the lesions.

CONCLUSION

TA may be of added value for the diagnosis of atypical HCA presenting without hypointensity on HBP imaging.

MRI texture analysis for differentiation of malignant and benign hepatocellular tumors in the non-cirrhotic liver

Purpose

To find potentially diagnostic texture analysis (TA) features and to evaluate the diagnostic accuracy of two-dimensional (2D) magnetic resonance (MR) TA for differentiation between hepatocellular carcinoma (HCC) and benign hepatocellular tumors in the non-cirrhotic liver in an exploratory MR-study.

Materials and methods

108 non-cirrhotic patients (62 female; 41.5 ± 18.3 years) undergoing preoperative contrast-enhanced MRI were retrospectively included in this multi-center-study. TA including gray-level histogram, co-occurrence and run-length matrix features (total 19 features) was performed by two independent readers. Native fat-saturated-T1w and T2w as well as arterial and portal-venous post contrast-enhanced 2D-image-slices were assessed. Conventional reading was performed by two separate independent readers. Differences in TA features between HCC and benign lesions were investigated using independent sample t-tests. Logistic regression analysis was performed to obtain the optimal number/combination of TA-features and diagnostic accuracy of TA analysis. Sensitivity and specificity of the better performing radiologist were compared to TA analysis.

Results

The highest number of significantly differing TA-features (n = 5) was found using the arterial-phase images including one gray-level histogram (skewness, p = 0.018) and four run-length matrix features (all, p < 0.02). The optimal binary logistic regression model for TA-features of the arterial-phase images contained 13 parameters with an accuracy of 84.5% (sensitivity 84.1%, specificity 84.9%) and area-under-the-curve of 0.92 (95%-confidence-interval 0.85–0.98) for diagnosis of HCC. Conventional reading yielded a significantly lower sensitivity (63.6%, p = 0.027) and no significant difference in specificity (94.6%, p = 0.289) at best.

Conclusion

2D-TA of MR images is a feasible objective method that may help to distinguish HCC from benign hepatocellular tumors in the non-cirrhotic liver. Most promising results were found in TA features in the arterial phase images.

Density estimation of grey-level co-occurrence matrices for image texture analysis

The Haralick texture features are common in the image analysis literature, partly because of their simplicity and because their values can be interpreted. It was recently observed that the Haralick texture features are very sensitive to the size of the GLCM that was used to compute them, which led to a new formulation that is invariant to the GLCM size. However, these new features still depend on the sample size used to compute the GLCM, i.e. the size of the input image region-of-interest (ROI). The purpose of this work was to investigate the performance of density estimation methods for approximating the GLCM and subsequently the corresponding invariant features. Three density estimation methods were evaluated, namely a piece-wise constant distribution, the Parzen-windows method, and the Gaussian mixture model. The methods were evaluated on 29 different image textures and 20 invariant Haralick texture features as well as a wide range of different ROI sizes. The results indicate that there are two types of features: those that have a clear minimum error for a particular GLCM size for each ROI size, and those whose error decreases monotonically with increased GLCM size. For the first type of features, the Gaussian mixture model gave the smallest errors, and in particular for small ROI sizes (less than about [Formula: see text]). In conclusion, the Gaussian mixture model is the preferred method for the first type of features (in particular for small ROIs). For the second type of features, simply using a large GLCM size is preferred.

Radiomics of liver MRI predict metastases in mice

Background

The purpose of this study was to investigate whether any texture features show a correlation with intrahepatic tumor growth before the metastasis is visible to the human eye.

Methods

Eight male C57BL6 mice (age 8–10 weeks) were injected intraportally with syngeneic MC-38 colon cancer cells and two mice were injected with phosphate-buffered saline (sham controls). Small animal magnetic resonance imaging (MRI) at 4.7 T was performed at baseline and days 4, 8, 12, 16, and 20 after injection applying a T2-weighted spin-echo sequence. Texture analysis was performed on the images yielding 32 texture features derived from histogram, gray-level co-occurrence matrix, gray-level run-length matrix, and gray-level size-zone matrix. The features were examined with a linear regression model/Pearson correlation test and hierarchical cluster analysis. From each cluster, the feature with the lowest variance was selected.

Results

Tumors were visible on MRI after 20 days. Eighteen features from histogram and the gray-level-matrices exhibited statistically significant correlations before day 20 in the experiment group, but not in the control animals. Cluster analysis revealed three distinct clusters of independent features. The features with the lowest variance were Energy, Short Run Emphasis, and Gray Level Non-Uniformity.

Conclusions

Texture features may quantitatively detect liver metastases before they become visually detectable by the radiologist.

Texture-based classification of different single liver lesion based on SPAIR T2W MRI images

Background

To assess the feasibility of texture analysis (TA) based on spectral attenuated inversion-recovery T2 weighted magnetic resonance imaging (SPAIR T2W-MRI) for the classification of hepatic hemangioma (HH), hepatic metastases (HM) and hepatocellular carcinoma (HCC).

Methods

The SPAIR T2W-MRI data of 162 patients with HH (n=55), HM (n=67) and HCC (n=40) were retrospectively analyzed. We used two independent cohorts for training (n = 112 patients) and validation (n = 50 patients). The TA was performed and textual parameters derived from the gray level co-occurrence matrix (GLCM), gray level gradient co-occurrence matrix (GLGCM), gray-level run-length matrix (GLRLM), Gabor wavelet transform (GWTF), intensity-size-zone matrix (ISZM), and histogram features were calculated. The capacity of each parameter to classify three types of single liver lesions was assessed using the Kruskal-Wallis test. Specificity and sensitivity for each of the studied parameters were derived using ROC curves. Four supervised classification algorithms were trained with the most influential textural features in the classification of tumor types. The test datasets validated the reliability of the models.

Results

The texture analyses showed that the HH versus HM, HM versus HCC, and HH versus HCC could be differentiated by 9, 16 and 10 feature parameters, respectively. The model’s misclassification rates were 11.7, 9.6 and 9.7% respectively. No texture feature was able to adequately distinguish among the three types of single liver lesions at the same time. The BP-ANN model had better predictive ability.

Conclusion

Texture features of SPAIR T2W-MRI can classify the three types of single liver lesions (HH, HM and HCC) and may serve as an adjunct tool for accurate diagnosis of these diseases.

Electronic supplementary material

The online version of this article (doi:10.1186/s12880-017-0212-x) contains supplementary material, which is available to authorized users.

Focal liver lesions segmentation and classification in nonenhanced T2-weighted MRI

PURPOSE

To automatically segment and classify focal liver lesions (FLLs) on nonenhanced T2-weighted magnetic resonance imaging (MRI) scans using a computer-aided diagnosis (CAD) algorithm.

METHODS

71 FLLs (30 benign lesions, 19 hepatocellular carcinomas, and 22 metastases) on T2-weighted MRI scans were delineated by the proposed CAD scheme. The FLL segmentation procedure involved wavelet multiscale analysis to extract accurate edge information and mean intensity values for consecutive edges computed using horizontal and vertical analysis that were fed into the subsequent fuzzy C-means algorithm for final FLL border extraction. Texture information for each extracted lesion was derived using 42 first- and second-order textural features from grayscale value histogram, co-occurrence, and run-length matrices. Twelve morphological features were also extracted to capture any shape differentiation between classes. Feature selection was performed with stepwise multilinear regression analysis that led to a reduced feature subset. A multiclass Probabilistic Neural Network (PNN) classifier was then designed and used for lesion classification. PNN model evaluation was performed using the leave-one-out (LOO) method and receiver operating characteristic (ROC) curve analysis.

RESULTS

The mean overlap between the automatically segmented FLLs and the manual segmentations performed by radiologists was 0.91 ± 0.12. The highest classification accuracies in the PNN model for the benign, hepatocellular carcinoma, and metastatic FLLs were 94.1%, 91.4%, and 94.1%, respectively, with sensitivity/specificity values of 90%/97.3%, 89.5%/92.2%, and 90.9%/95.6% respectively. The overall classification accuracy for the proposed system was 90.1%.

CONCLUSIONS

Our diagnostic system using sophisticated FLL segmentation and classification algorithms is a powerful tool for routine clinical MRI-based liver evaluation and can be a supplement to contrast-enhanced MRI to prevent unnecessary invasive procedures.

Effects of gadolinium-based contrast agent concentrations (0.5 M or 1.0 M) on the diagnostic performance of magnetic resonance imaging examinations: systematic review of the literature

Background To date there is no agreement as to what is the optimal concentration for gadolinium-based contrast agents (GBCAs). Purpose To assess whether diagnostic performance differences exist between 0.5 M and 1.0 M GBCAs used for magnetic resonance imaging (MRI). Material and Methods A PubMed literature search identified 21 clinical studies published between 2005 and 2013 which evaluated the diagnostic efficacy of both types of GBCAs. Study design, type of procedure, GBCA administration mode, imaging performances, impact on patient management, study limitations, and biases were analyzed. No statistical test was performed on pooled data. Results Sixteen comparative and five non-comparative studies were analyzed, involving 2183 patients who underwent MRI procedures for various indications. In 67% of the studies, 0.5 M and 1.0 M GBCAs were injected at equimolar gadolinium amounts per kg body weight. Only 33% applied the same molar flow rate for delivery of the GBCAs. No significant differences between GBCAs were reported for 23 out of 27 qualitative endpoints (mainly image quality, lesion, and vessel visualization) and 29 out of 40 quantitative endpoints. Three out of four studies with non-equimolar delivery rates showed better contrast-to-noise and signal-to-noise ratios for 1.0 M gadobutrol, without showing an impact on diagnostic performance. Methodological biases were identified in several studies impairing the interpretation of comparisons. Conclusion Imaging differences between 0.5 M and 1.0 M GBCAs were essentially observed under non-equimolar delivery rates. However, they did not result into greater diagnostic efficacy when performed under equimolar conditions.

Differentiation of Diffuse Large B-cell Lymphoma From Follicular Lymphoma Using Texture Analysis on Conventional MR Images at 3.0 Tesla

RATIONAL AND OBJECTIVES

Diffuse large B-cell lymphoma (DLBCL) represents the most common type of aggressive non-Hodgkin lymphoma (NHL); follicular lymphoma (FL) is the most frequent indolent NHL. The aim of this study was to investigate whether texture-based analysis of conventional magnetic resonance imaging (MRI) allows discrimination of DLBCL from FL, and further, to correlate the MRI texture features with diffusion-weighted imaging apparent diffusion coefficient (ADC) value and tumor tissue cellularity.

MATERIALS AND METHODS

Forty-one patients with histologically proven NHL (30 DLBCL and 11 FL) underwent conventional MRI and diffusion-weighted imaging examination before treatment. Based on regions of interest, texture analysis was performed on T1-weighted images pre- and postcontrast enhancement and on T2-weighted images with and without fat suppression, and features derived from the run-length matrix- and co-occurrence matrix-based methods were analyzed. Receiver operating characteristic curves were performed for the three most discriminative texture features for the differentiation of the two most common types of lymphoma. The analyzed MRI texture features were correlated with the ADC value and the tumor tissue cellularity.

RESULTS

We found that on T1-weighted images postcontrast enhancement, run-length matrix-based texture analysis for lesion classification differentiated DLBCL from FL, with specificity and sensitivity of 76.6% and 76.5%, respectively. There was no correlation between the texture features and the ADC value or tumor tissue cellularity.

CONCLUSIONS

DLBCL and FL can be differentiated by means of texture analysis on T1-weighted MRI postcontrast enhancement. These results could serve as a basis for the use of the texture features on conventional MRI as adjunct to clinical examination to distinguish DLBCL from FL.