Computer-aided diagnosis of breast masses using quantified BI-RADS findings

Multi-Task/Single-Task Joint Learning of Ultrasound BI-RADS Features

Breast cancer is one of the most common cancers among women in the world. The breast imaging reporting and data system (BI-RADS) features effectively improve the accuracy and sensitivity of breast tumors. Based on the description of signs in BI-RADS, a quantitative scoring scheme is proposed based on ultrasound (US) data. This scheme includes feature extraction of high-level semantic, that is, an intermediate step interpreting the subsequent diagnosis. However, the scheme requires doctors to score the features of breast data, which is labor-intensive. To reduce the burden of doctors, we design a multi-task learning (MTL) framework, which can directly output the scores of different BI-RADS features from the raw US images. The MTL framework consists of a shared network that learns global features and K soft attention networks for different BI-RADS features. Thus, it enables the network not only to learn the potential correlation among different BI-RADS features, but also learn the unique specificity of each feature, which can assist each other and jointly improve the scoring accuracy. In addition, we group different BI-RADS features according to the correlation among tasks and build a multi-task/single-task joint framework. Experimental results on the US breast tumor dataset collected from 1859 patients with 4458 US images show that the proposed BI-RADS feature scoring framework achieves an average scoring accuracy of 84.91% for 11 BI-RADS features on the test dataset, which is helpful for the subsequent diagnosis of breast tumors.

A bimodal BI-RADS-guided GoogLeNet-based CAD system for solid breast masses discrimination using transfer learning

Numerous solid breast masses require sophisticated analysis to establish a differential diagnosis. Consequently, complementary modalities such as ultrasound imaging are frequently required to evaluate mammographically further detected masses. Radiologists mentally integrate complementary information from images acquired of the same patient to make a more conclusive and effective diagnosis. However, it has always been a challenging task. This paper details a novel bimodal GoogLeNet-based CAD system that addresses the challenges associated with combining information from mammographic and sonographic images for solid breast mass classification. Each modality is initially trained using two distinct monomodal models in the proposed framework. Then, using the high-level feature maps extracted from both modalities, a bimodal model is trained. In order to fully exploit the BI-RADS descriptors, different image content representations of each mass are obtained and used as input images. In addition, using an ImageNet pre-trained GoogLeNet model, two publicly available databases, and our collected dataset, a two-step transfer learning strategy has been proposed. Our bimodal model achieves the best recognition results in terms of sensitivity, specificity, F1-score, Matthews Correlation Coefficient, area under the receiver operating characteristic curve, and accuracy metrics of 90.91%, 89.87%, 90.32%, 80.78%, 95.82%, and 90.38%, respectively. The promising results indicate that the proposed CAD system can facilitate bimodal suspicious mass analysis and thus contribute significantly to improving breast cancer diagnostic performance.

Methods for the segmentation and classification of breast ultrasound images: a review

PURPOSE

Breast ultrasound (BUS) is one of the imaging modalities for the diagnosis and treatment of breast cancer. However, the segmentation and classification of BUS images is a challenging task. In recent years, several methods for segmenting and classifying BUS images have been studied. These methods use BUS datasets for evaluation. In addition, semantic segmentation algorithms have gained prominence for segmenting medical images.

METHODS

In this paper, we examined different methods for segmenting and classifying BUS images. Popular datasets used to evaluate BUS images and semantic segmentation algorithms were examined. Several segmentation and classification papers were selected for analysis and review. Both conventional and semantic methods for BUS segmentation were reviewed.

RESULTS

Commonly used methods for BUS segmentation were depicted in a graphical representation, while other conventional methods for segmentation were equally elucidated.

CONCLUSIONS

We presented a review of the segmentation and classification methods for tumours detected in BUS images. This review paper selected old and recent studies on segmenting and classifying tumours in BUS images.

Quantifying Axial Spine Images Using Object-Specific Bi-Path Network

Automatic estimation of indices from medical images is the main goal of computer-aided quantification (CADq), which speeds up diagnosis and lightens the workload of radiologists. Deep learning technique is a good choice for implementing CADq. Usually, to acquire high-accuracy quantification, specific network architecture needs to be designed for a given CADq task. In this study, considering that the target organs are the intervertebral disc and the dural sac, we propose an object-specific bi-path network (OSBP-Net) for axial spine image quantification. Each path of the OSBP-Net comprises a shallow feature extraction layer (SFE) and a deep feature extraction sub-network (DFE). The SFEs use different convolution strides because the two target organs have different anatomical sizes. The DFEs use average pooling for downsampling based on the observation that the target organs have lower intensity than the background. In addition, an inter-path dissimilarity constraint is proposed and applied to the output of the SFEs, taking into account that the activated regions in the feature maps of two paths should be different theoretically. An inter-index correlation regularization is introduced and applied to the output of the DFEs based on the observation that the diameter and area of the same object express an approximately linear relation. The prediction results of OSBP-Net are compared to several state-of-the-art machine learning-based CADq methods. The comparison reveals that the proposed methods precede other competing methods extensively, indicating its great potential for spine CADq.

Breast Tumor Classification in Ultrasound Images Using Combined Deep and Handcrafted Features

This study aims to enable effective breast ultrasound image classification by combining deep features with conventional handcrafted features to classify the tumors. In particular, the deep features are extracted from a pre-trained convolutional neural network model, namely the VGG19 model, at six different extraction levels. The deep features extracted at each level are analyzed using a features selection algorithm to identify the deep feature combination that achieves the highest classification performance. Furthermore, the extracted deep features are combined with handcrafted texture and morphological features and processed using features selection to investigate the possibility of improving the classification performance. The cross-validation analysis, which is performed using 380 breast ultrasound images, shows that the best combination of deep features is obtained using a feature set, denoted by CONV features that include convolution features extracted from all convolution blocks of the VGG19 model. In particular, the CONV features achieved mean accuracy, sensitivity, and specificity values of 94.2%, 93.3%, and 94.9%, respectively. The analysis also shows that the performance of the CONV features degrades substantially when the features selection algorithm is not applied. The classification performance of the CONV features is improved by combining these features with handcrafted morphological features to achieve mean accuracy, sensitivity, and specificity values of 96.1%, 95.7%, and 96.3%, respectively. Furthermore, the cross-validation analysis demonstrates that the CONV features and the combined CONV and morphological features outperform the handcrafted texture and morphological features as well as the fine-tuned VGG19 model. The generalization performance of the CONV features and the combined CONV and morphological features is demonstrated by performing the training using the 380 breast ultrasound images and the testing using another dataset that includes 163 images. The results suggest that the combined CONV and morphological features can achieve effective breast ultrasound image classifications that increase the capability of detecting malignant tumors and reduce the potential of misclassifying benign tumors.

Intelligent Glioma Grading Based on Deep Transfer Learning of MRI Radiomic Features

According to a classification of central nervous system tumors by the World Health Organization, diffuse gliomas are classified into grade 2, 3, and 4 gliomas in accordance with their aggressiveness. To quantitatively evaluate a tumor’s malignancy from brain magnetic resonance imaging, this study proposed a computer-aided diagnosis (CAD) system based on a deep convolutional neural network (DCNN). Gliomas from a multi-center database (The Cancer Imaging Archive) composed of a total of 30 grade 2, 43 grade 3, and 57 grade 4 gliomas were used for the training and evaluation of the proposed CAD. Using transfer learning to fine-tune AlexNet, a DCNN, its internal layers, and parameters trained from a million images were transferred to learn how to differentiate the acquired gliomas. Data augmentation was also implemented to increase possible spatial and geometric variations for a better training model. The transferred DCNN achieved an accuracy of 97.9% with a standard deviation of ±1% and an area under the receiver operation characteristics curve (Az) of 0.9991 ± 0, which were superior to handcrafted image features, the DCNN without pretrained features, which only achieved a mean accuracy of 61.42% with a standard deviation of ±7% and a mean Az of 0.8222 ± 0.07, and the DCNN without data augmentation, which was the worst with a mean accuracy of 59.85% with a standard deviation ±16% and a mean Az of 0.7896 ± 0.18. The DCNN with pretrained features and data augmentation can accurately and efficiently classify grade 2, 3, and 4 gliomas. The high accuracy is promising in providing diagnostic suggestions to radiologists in the clinic.

Performance of machine learning software to classify breast lesions using BI-RADS radiomic features on ultrasound images

BACKGROUND

The purpose of this work was to evaluate computable Breast Imaging Reporting and Data System (BI-RADS) radiomic features to classify breast masses on ultrasound B-mode images.

METHODS

The database consisted of 206 consecutive lesions (144 benign and 62 malignant) proved by percutaneous biopsy in a prospective study approved by the local ethical committee. A radiologist manually delineated the contour of the lesions on greyscale images. We extracted the main ten radiomic features based on the BI-RADS lexicon and classified the lesions as benign or malignant using a bottom-up approach for five machine learning (ML) methods: multilayer perceptron (MLP), decision tree (DT), linear discriminant analysis (LDA), random forest (RF), and support vector machine (SVM). We performed a 10-fold cross validation for training and testing of all classifiers. Receiver operating characteristic (ROC) analysis was used for providing the area under the curve with 95% confidence intervals (CI).

RESULTS

The classifier with the highest AUC at ROC analysis was SVM (AUC = 0.840, 95% CI 0.6667-0.9762), with 71.4% sensitivity (95% CI 0.6479-0.8616) and 76.9% specificity (95% CI 0.6148-0.8228). The best AUC for each method was 0.744 (95% CI 0.677-0.774) for DT, 0.818 (95% CI 0.6667-0.9444) for LDA, 0.811 (95% CI 0.710-0.892) for RF, and 0.806 (95% CI 0.677-0.839) for MLP. Lesion margin and orientation were the optimal features for all the machine learning methods.

CONCLUSIONS

ML can aid the distinction between benign and malignant breast lesion on ultrasound images using quantified BI-RADS descriptors. SVM provided the highest ROC-AUC (0.840).

Two-stage CNNs for computerized BI-RADS categorization in breast ultrasound images

BACKGROUND

Quantizing the Breast Imaging Reporting and Data System (BI-RADS) criteria into different categories with the single ultrasound modality has always been a challenge. To achieve this, we proposed a two-stage grading system to automatically evaluate breast tumors from ultrasound images into five categories based on convolutional neural networks (CNNs).

METHODS

This new developed automatic grading system was consisted of two stages, including the tumor identification and the tumor grading. The constructed network for tumor identification, denoted as ROI-CNN, can identify the region contained the tumor from the original breast ultrasound images. The following tumor categorization network, denoted as G-CNN, can generate effective features for differentiating the identified regions of interest (ROIs) into five categories: Category "3", Category "4A", Category "4B", Category "4C", and Category "5". Particularly, to promote the predictions identified by the ROI-CNN better tailor to the tumor, refinement procedure based on Level-set was leveraged as a joint between the stage and grading stage.

RESULTS

We tested the proposed two-stage grading system against 2238 cases with breast tumors in ultrasound images. With the accuracy as an indicator, our automatic computerized evaluation for grading breast tumors exhibited a performance comparable to that of subjective categories determined by physicians. Experimental results show that our two-stage framework can achieve the accuracy of 0.998 on Category "3", 0.940 on Category "4A", 0.734 on Category "4B", 0.922 on Category "4C", and 0.876 on Category "5".

CONCLUSION

The proposed scheme can extract effective features from the breast ultrasound images for the final classification of breast tumors by decoupling the identification features and classification features with different CNNs. Besides, the proposed scheme can extend the diagnosing of breast tumors in ultrasound images to five sub-categories according to BI-RADS rather than merely distinguishing the breast tumor malignant from benign.

Breast tumor classification using different features of quantitative ultrasound parametric images

RATIONALE AND OBJECTIVES

The ultrasound B-mode-based morphological and texture analysis and Nakagami parametric imaging have been proposed to characterize breast tumors. Since these three feature categories of ultrasonic tissue characterization supply information on different physical characteristics of breast tumors, by combining the above methods is expected to provide more clues for classifying breast tumors.

MATERIALS AND METHODS

To verify the validity of the concept, raw data were obtained from 160 clinical cases. Six different types of morphological-feature parameters, four texture features, and the Nakagami parameter of benignancy and malignancy were extracted for evaluation. The Pearson's correlation matrix was used to calculate the correlation between different feature parameters. The fuzzy c-means clustering and stepwise regression techniques were utilized to determine the optimal feature set, respectively. The logistic regression, receiver operating characteristic curve, and support vector machine were used to estimate the diagnostic ability.

RESULTS

The best performance was obtained by combining morphological-feature parameter (e.g., standard deviation of the shortest distance), texture feature (e.g., variance), and the Nakagami parameter, with an accuracy of 89.4%, a specificity of 86.3%, a sensitivity of 92.5%, and an area under receiver operating characteristic curve of 0.96. There was no significant difference between using fuzzy c-means clustering, logistic regression, and support vector machine based on the optimal feature set for breast tumors classification.

CONCLUSION

Therefore, we verified that different physical ultrasonic features are functionally complementary and thus improve the performance in diagnosing breast tumors. Moreover, the optimal feature set had the maximum discriminating performance should be irrelative to the power of classifiers.

A Review of Denoising Medical Images Using Machine Learning Approaches

BACKGROUND

This paper attempts to identify suitable Machine Learning (ML) approach for image denoising of radiology based medical application. The Identification of ML approach is based on (i) Review of ML approach for denoising (ii) Review of suitable Medical Denoising approach.

DISCUSSION

The review focuses on six application of radiology: Medical Ultrasound (US) for fetus development, US Computer Aided Diagnosis (CAD) and detection for breast, skin lesions, brain tumor MRI diagnosis, X-Ray for chest analysis, Breast cancer using MRI imaging. This survey identifies the ML approach with better accuracy for medical diagnosis by radiologists. The image denoising approaches further includes basic filtering techniques, wavelet medical denoising, curvelet and optimization techniques. In most of the applications, the machine learning performance is better than the conventional image denoising techniques. For fast and computational results the radiologists are using the machine learning methods on MRI, US, X-Ray and Skin lesion images. The characteristics and contributions of different ML approaches are considered in this paper.

CONCLUSION

The problem faced by the researchers during image denoising techniques and machine learning applications for clinical settings have also been discussed.

Computer-aided prediction model for axillary lymph node metastasis in breast cancer using tumor morphological and textural features on ultrasound

BACKGROUND AND OBJECTIVES

Axillary lymph node (ALN) status is a key indicator in assessing and determining the treatment strategy for patients with newly diagnosed breast cancer. Previous studies suggest that sonographic features of a primary tumor have the potential to predict ALN status in the preoperative staging of breast cancer. In this study, a computer-aided prediction (CAP) model as well as the tumor features for ALN metastasis in breast cancers were developed using breast ultrasound (US) images.

METHODS

A total of 249 malignant tumors were acquired from 247 female patients (ages 20-84 years; mean 55 ± 11 years) to test the differences between the non-metastatic (130) and metastatic (119) groups based on various features. After applying semi-automatic tumor segmentation, 69 quantitative features were extracted. The features included morphology and texture of tumors inside a ROI of breast US image. By the backward feature selection and linear logistic regression, the prediction model was constructed and established to estimate the likelihood of ALN metastasis for each sample collected.

RESULTS

In the experiments, the texture features showed higher performance for predicting ALN metastasis compared to morphology (Az, 0.730 vs 0.667). The difference, however, was not statistically significant (p-values > 0.05). Combining the textural and morphological features, the accuracy, sensitivity, specificity, and Az value achieved 75.1% (187/249), 79.0% (94/119), 71.5% (93/130), and 0.757, respectively.

CONCLUSIONS

The proposed CAP model, which combines textural and morphological features of primary tumor, may be a useful method to determine the ALN status in patients with breast cancer.

Automated classification of focal breast lesions according to S-detect: validation and role as a clinical and teaching tool

PURPOSE

To assess the diagnostic performance and the potential as a teaching tool of S-detect in the assessment of focal breast lesions.

METHODS

61 patients (age 21-84 years) with benign breast lesions in follow-up or candidate to pathological sampling or with suspicious lesions candidate to biopsy were enrolled. The study was based on a prospective and on a retrospective phase. In the prospective phase, after completion of baseline US by an experienced breast radiologist and S-detect assessment, 5 operators with different experience and dedication to breast radiology performed elastographic exams. In the retrospective phase, the 5 operators performed a retrospective assessment and categorized lesions with BI-RADS 2013 lexicon. Integration of S-detect to in-training operators evaluations was performed by giving priority to S-detect analysis in case of disagreement. 2 × 2 contingency tables and ROC analysis were used to assess the diagnostic performances; inter-rater agreement was measured with Cohen's k; Bonferroni's test was used to compare performances. A significance threshold of p = 0.05 was adopted.

RESULTS

All operators showed sensitivity > 90% and varying specificity (50-75%); S-detect showed sensitivity > 90 and 70.8% specificity, with inter-rater agreement ranging from moderate to good. Lower specificities were improved by the addition of S-detect. The addition of elastography did not lead to any improvement of the diagnostic performance.

CONCLUSIONS

S-detect is a feasible tool for the characterization of breast lesions; it has a potential as a teaching tool for the less experienced operators.

A computer-aided diagnosis system for breast ultrasound based on weighted BI-RADS classes

BACKGROUND AND OBJECTIVE

Conventional computer-aided diagnosis (CAD) systems for breast ultrasound (BUS) are trained to classify pathological classes, that is, benign and malignant. However, from a clinical perspective, this kind of classification does not agree totally with radiologists' diagnoses. Usually, the tumors are assessed by using a BI-RADS (Breast Imaging-Reporting and Data System) category and, accordingly, a recommendation is emitted: annual study for category 2 (benign), six-month follow-up study for category 3 (probably benign), and biopsy for categories 4 and 5 (suspicious of malignancy). Hence, in this paper, a CAD system based on BI-RADS categories weighted by pathological information is presented. The goal is to increase the classification performance by reducing the common class imbalance found in pathological classes as well as to provide outcomes quite similar to radiologists' recommendations.

METHODS

The BUS dataset considers 781 benign lesions and 347 malignant tumors proven by biopsy. Moreover, every lesion is associated to one BI-RADS category in the set {2, 3, 4, 5}. Thus, the dataset is split into three weighted classes: benign, BI-RADS 2 in benign lesions; probably benign, BI-RADS 3 and 4 in benign lesions; and malignant, BI-RADS 4 and 5 in malignant lesions. Thereafter, a random forest (RF) classifier, denoted by RF_w, is trained to predict the weighted BI-RADS classes. In addition, for comparison purposes, a RF classifier is trained to predict pathological classes, denoted as RF_p.

RESULTS

The ability of the classifiers to predict the pathological classes is measured by the area under the ROC curve (AUC), sensitivity (SEN), and specificity (SPE). The RF_w classifier obtained AUC=0.872,SEN=0.826, and SPE=0.919, whereas the RF_p classifier reached AUC=0.868,SEN=0.808, and SPE=0.929. According to a one-way analysis of variance test, the RF_w classifier statistically outperforms (p < 0.001) the RF_p classifier in terms of the AUC and SEN. Moreover, the classification performance of RF_w to predict weighted BI-RADS classes is given by the Matthews correlation coefficient that obtained 0.614.

CONCLUSIONS

The division of the classification problem into three classes reduces the imbalance between benign and malignant classes; thus, the sensitivity is increased without degrading the specificity. Therefore, the CAD based on weighted BI-RADS classes improves the classification performance of the conventional CAD systems. Additionally, the proposed approach has the advantage of being capable of providing a multiclass outcome related to radiologists' recommendations.

Computer-aided diagnosis program for classifying the risk of hepatocellular carcinoma on MR images following liver imaging reporting and data system (LI-RADS)

PURPOSE

To develop and evaluate a computer-aided diagnosis (CAD) program for liver lesions on magnetic resonance (MR) images for classification of the risk of hepatocellular carcinoma (HCC) following the liver imaging reporting and data system (LI-RADS).

MATERIALS AND METHODS

Liver MR images from 41 patients with hyperenhancing liver lesions categorized as LR 3, 4, and 5 were evaluated by two radiologists. The major LI-RADS features of each index liver lesion were recorded, including size (maximum transverse diameter), presence of hyperenhancement, washout appearance, and capsule appearance. A CAD program was implemented to register MR images at different contrast-enhancement phases, segment liver lesions, extract lesion features, and classify lesions according to LI-RADS. The LI-RADS features quantified by CAD were compared with those assessed by radiologists using the intraclass correlation coefficient (ICC) and receiver operator curve (ROC) analyses. The LI-RADS categorization between CAD and radiologists was evaluated using the weighted Cohen's kappa coefficient.

RESULTS

The mean and standard deviation of the lesion diameters were 21 ± 11 mm (range, 7-70 mm) by radiologists and 22 ± 11 mm (range, 8-72 mm) by CAD (ICC, 0.96-0.97). The area under the curve (AUC) for the washout assessment by CAD was 0.79-0.93 with sensitivity 0.69-0.82 and specificity 0.79-1. The AUC for the capsule assessment by CAD was 0.79-0.9 with sensitivity 0.75-0.9 and specificity 0.82-0.96. The classifications by the radiologists and CAD coincided in 76-83% lesions (k = 0.57-0.71), while the agreements between radiologists were in 78% lesions (k = 0.59).

CONCLUSION

We developed a CAD program for liver lesions on MR images and showed a substantial agreement in the LI-RADS-based classification of the risk of HCCs between the CAD and radiologists.

LEVEL OF EVIDENCE

1 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2018;47:710-722.

Quantitative glioma grading using transformed gray-scale invariant textures of MRI

BACKGROUND

A computer-aided diagnosis (CAD) system based on intensity-invariant magnetic resonance (MR) imaging features was proposed to grade gliomas for general application to various scanning systems and settings.

METHOD

In total, 34 glioblastomas and 73 lower-grade gliomas comprised the image database to evaluate the proposed CAD system. For each case, the local texture on MR images was transformed into a local binary pattern (LBP) which was intensity-invariant. From the LBP, quantitative image features, including the histogram moment and textures, were extracted and combined in a logistic regression classifier to establish a malignancy prediction model. The performance was compared to conventional texture features to demonstrate the improvement.

RESULTS

The performance of the CAD system based on LBP features achieved an accuracy of 93% (100/107), a sensitivity of 97% (33/34), a negative predictive value of 99% (67/68), and an area under the receiver operating characteristic curve (Az) of 0.94, which were significantly better than the conventional texture features: an accuracy of 84% (90/107), a sensitivity of 76% (26/34), a negative predictive value of 89% (64/72), and an Az of 0.89 with respective p values of 0.0303, 0.0122, 0.0201, and 0.0334.

CONCLUSIONS

More-robust texture features were extracted from MR images and combined into a significantly better CAD system for distinguishing glioblastomas from lower-grade gliomas. The proposed CAD system would be more practical in clinical use with various imaging systems and settings.

Effect of a computer-aided diagnosis system on radiologists' performance in grading gliomas with MRI

The effects of a computer-aided diagnosis (CAD) system based on quantitative intensity features with magnetic resonance (MR) imaging (MRI) were evaluated by examining radiologists' performance in grading gliomas. The acquired MRI database included 71 lower-grade gliomas and 34 glioblastomas. Quantitative image features were extracted from the tumor area and combined in a CAD system to generate a prediction model. The effect of the CAD system was evaluated in a two-stage procedure. First, a radiologist performed a conventional reading. A sequential second reading was determined with a malignancy estimation by the CAD system. Each MR image was regularly read by one radiologist out of a group of three radiologists. The CAD system achieved an accuracy of 87% (91/105), a sensitivity of 79% (27/34), a specificity of 90% (64/71), and an area under the receiver operating characteristic curve (Az) of 0.89. In the evaluation, the radiologists' Az values significantly improved from 0.81, 0.87, and 0.84 to 0.90, 0.90, and 0.88 with p = 0.0011, 0.0076, and 0.0167, respectively. Based on the MR image features, the proposed CAD system not only performed well in distinguishing glioblastomas from lower-grade gliomas but also provided suggestions about glioma grading to reinforce radiologists' confidence rating.

Computer-aided grading of gliomas based on local and global MRI features

BACKGROUND AND OBJECTIVES

A computer-aided diagnosis (CAD) system based on quantitative magnetic resonance imaging (MRI) features was developed to evaluate the malignancy of diffuse gliomas, which are central nervous system tumors.

METHODS

The acquired image database for the CAD performance evaluation was composed of 34 glioblastomas and 73 diffuse lower-grade gliomas. In each case, tissues enclosed in a delineated tumor area were analyzed according to their gray-scale intensities on MRI scans. Four histogram moment features describing the global gray-scale distributions of gliomas tissues and 14 textural features were used to interpret local correlations between adjacent pixel values. With a logistic regression model, the individual feature set and a combination of both feature sets were used to establish the malignancy prediction model.

RESULTS

Performances of the CAD system using global, local, and the combination of both image feature sets achieved accuracies of 76%, 83%, and 88%, respectively. Compared to global features, the combined features had significantly better accuracy (p = 0.0213). With respect to the pathology results, the CAD classification obtained substantial agreement κ = 0.698, p < 0.001.

CONCLUSIONS

Numerous proposed image features were significant in distinguishing glioblastomas from lower-grade gliomas. Combining them further into a malignancy prediction model would be promising in providing diagnostic suggestions for clinical use.

A contrast enhancement method for improving the segmentation of breast lesions on ultrasonography

PURPOSE

This paper presents an adaptive contrast enhancement method based on sigmoidal mapping function (SACE) used for improving the computerized segmentation of breast lesions on ultrasound.

METHODS

First, from the original ultrasound image an intensity variation map is obtained, which is used to generate local sigmoidal mapping functions related to distinct contextual regions. Then, a bilinear interpolation scheme is used to transform every original pixel to a new gray level value. Also, four contrast enhancement techniques widely used in breast ultrasound enhancement are implemented: histogram equalization (HEQ), contrast limited adaptive histogram equalization (CLAHE), fuzzy enhancement (FEN), and sigmoid based enhancement (SEN). In addition, these contrast enhancement techniques are considered in a computerized lesion segmentation scheme based on watershed transformation. The performance comparison among techniques is assessed in terms of both the quality of contrast enhancement and the segmentation accuracy. The former is quantified by the measure, where the greater the value, the better the contrast enhancement, whereas the latter is calculated by the Jaccard index, which should tend towards unity to indicate adequate segmentation.

RESULTS

The experiments consider a data set with 500 breast ultrasound images. The results show that SACE outperforms its counterparts, where the median values for the measure are: SACE: 139.4, SEN: 68.2, HEQ: 64.1, CLAHE: 62.8, and FEN: 7.9. Considering the segmentation performance results, the SACE method presents the largest accuracy, where the median values for the Jaccard index are: SACE: 0.81, FEN: 0.80, CLAHE: 0.79, HEQ: 77, and SEN: 0.63.

CONCLUSION

The SACE method performs well due to the combination of three elements: (1) the intensity variation map reduces intensity variations that could distort the real response of the mapping function, (2) the sigmoidal mapping function enhances the gray level range where the transition between lesion and background is found, and (3) the adaptive enhancing scheme for coping with local contrasts. Hence, the SACE approach is appropriate for enhancing contrast before computerized lesion segmentation.

A Fusion-Based Approach for Breast Ultrasound Image Classification Using Multiple-ROI Texture and Morphological Analyses

Ultrasound imaging is commonly used for breast cancer diagnosis, but accurate interpretation of breast ultrasound (BUS) images is often challenging and operator-dependent. Computer-aided diagnosis (CAD) systems can be employed to provide the radiologists with a second opinion to improve the diagnosis accuracy. In this study, a new CAD system is developed to enable accurate BUS image classification. In particular, an improved texture analysis is introduced, in which the tumor is divided into a set of nonoverlapping regions of interest (ROIs). Each ROI is analyzed using gray-level cooccurrence matrix features and a support vector machine classifier to estimate its tumor class indicator. The tumor class indicators of all ROIs are combined using a voting mechanism to estimate the tumor class. In addition, morphological analysis is employed to classify the tumor. A probabilistic approach is used to fuse the classification results of the multiple-ROI texture analysis and morphological analysis. The proposed approach is applied to classify 110 BUS images that include 64 benign and 46 malignant tumors. The accuracy, specificity, and sensitivity obtained using the proposed approach are 98.2%, 98.4%, and 97.8%, respectively. These results demonstrate that the proposed approach can effectively be used to differentiate benign and malignant tumors.

Feasibility Testing: Three-dimensional Tumor Mapping in Different Orientations of Automated Breast Ultrasound

A tumor-mapping algorithm was proposed to identify the same regions in different passes of automated breast ultrasound (ABUS). A total of 53 abnormal passes with 41 biopsy-proven tumors and 13 normal passes were collected. After computer-aided tumor detection, a mapping pair was composed of a detected region in one pass and another region in another pass. Location criteria, including the radial position as on a clock, the relative distance and the distance to the nipple, were used to extract mapping pairs with close regions. Quantitative intensity, morphology, texture and location features were then combined in a classifier for further classification. The performance of the classifier achieved a mapping rate of 80.39% (41/51), with an error rate of 5.97% (4/67). The trade-offs between the mapping and error rates were evaluated, and Az = 0.9094 was obtained. The proposed tumor-mapping algorithm was capable of automatically providing location correspondence information that would be helpful in reviews of ABUS examinations.

Computer-Aided Diagnosis for Breast Ultrasound Using Computerized BI-RADS Features and Machine Learning Methods

This work identifies effective computable features from the Breast Imaging Reporting and Data System (BI-RADS), to develop a computer-aided diagnosis (CAD) system for breast ultrasound. Computerized features corresponding to ultrasound BI-RADs categories were designed and tested using a database of 283 pathology-proven benign and malignant lesions. Features were selected based on classification performance using a "bottom-up" approach for different machine learning methods, including decision tree, artificial neural network, random forest and support vector machine. Using 10-fold cross-validation on the database of 283 cases, the highest area under the receiver operating characteristic (ROC) curve (AUC) was 0.84 from a support vector machine with 77.7% overall accuracy; the highest overall accuracy, 78.5%, was from a random forest with the AUC 0.83. Lesion margin and orientation were optimum features common to all of the different machine learning methods. These features can be used in CAD systems to help distinguish benign from worrisome lesions.

Quantitative breast lesion classification based on multichannel distributions in shear-wave imaging

BACKGROUND AND OBJECTIVES

A computer-aided diagnosis (CAD) system based on the quantified color distributions in shear-wave elastography (SWE) was developed to evaluate the malignancies of breast tumors.

METHODS

For 57 benign and 31 malignant tumors, 18 SWE features were extracted from regions of interest (ROI), including the tumor and peritumoral areas. In the ROI, a histogram in each color channel was described using moments such as the mean, variance, skewness, and kurtosis. Moreover, three color channels were combined as a vector to evaluate tissue elasticity. The SWE features were then combined in a logistic regression classifier for breast tumor classification.

RESULTS

The performance of the CAD system achieved an accuracy of 81%. Combining the CAD system with a BI-RADS assessment obtained an Az improvement from 0.77 to 0.89 (p-value <0.05).

CONCLUSIONS

The combination of the proposed CAD system based on SWE features and the BI-RADS assessment would provide a promising diagnostic suggestion.

Intensity-Invariant Texture Analysis for Classification of BI-RADS Category 3 Breast Masses

Radiologists likely incorrectly classify benign masses as Breast Imaging Reporting and Data System (BI-RADS) category 3. A computer-aided diagnosis (CAD) system was developed in this study as a second viewer to avoid misclassification of carcinomas. Sixty-nine biopsy-proven BI-RADS category 3 masses, including 21 malignant and 48 benign masses, were used to evaluate the CAD system. To improve the texture features, gray-scale variations between images were reduced by transforming pixels into intensity-invariant ranklet coefficients. The textures of the tumor and speckle pixels were extracted from the transformed ranklet images to provide more robust features than in conventional CAD systems. As a result, tumor texture and speckle texture with ranklet transformation achieved significantly better areas under the receiver operating characteristic curve (Az) compared with those without ranklet transformation (Az = 0.83 vs. 0.58 and Az = 0.80 vs. 0.56, p value < 0.05). The improved CAD system can be a second reader to confirm the classification of BI-RADS category 3 masses.

Classification of Benign and Malignant Breast Tumors in Ultrasound Images with Posterior Acoustic Shadowing Using Half-Contour Features

Posterior acoustic shadowing (PAS) can bias breast tumor segmentation and classification in ultrasound images. In this paper, half-contour features are proposed to classify benign and malignant breast tumors with PAS, considering the fact that the upper half of the tumor contour is less affected by PAS. Adaptive thresholding and disk expansion are employed to detect tumor contours. Based on the detected full contour, the upper half contour is extracted. For breast tumor classification, six quantitative feature parameters are analyzed for both full contours and half contours, including standard deviation of degree (SDD), which is proposed to describe tumor irregularity. Fifty clinical cases (40 with PAS and 10 without PAS) were used. Tumor circularity (TC) and SDD were both effective full- and half-contour parameters in classifying images without PAS. Half-contour TC [74 % accuracy, 72 % sensitivity, 76 % specificity, 0.78 area under the receiver operating characteristic curve (AUC), p > 0.05] significantly improved the classification of breast tumors with PAS compared to that with full-contour TC (54 % accuracy, 56 % sensitivity, 52 % specificity, 0.52 AUC, p > 0.05). Half-contour SDD (72 % accuracy, 76 % sensitivity, 68 % specificity, 0.81 AUC, p < 0.05) improved the classification of breast tumors with PAS compared to that with full-contour SDD (62 % accuracy, 80 % sensitivity, 44 % specificity, 0.61 AUC, p > 0.05). The proposed half-contour TC and SDD may be useful in classifying benign and malignant breast tumors in ultrasound images affected by PAS.

Computer-aided diagnosis system based on fuzzy logic for breast cancer categorization

BACKGROUND

Fuzzy logic can help reduce the difficulties faced by computational systems to represent and simulate the reasoning and the style adopted by radiologists in the process of medical image analysis. The study described in this paper consists of a new method that applies fuzzy logic concepts to improve the representation of features related to image description in order to make it semantically more consistent. Specifically, we have developed a computer-aided diagnosis tool for automatic BI-RADS categorization of breast lesions. The user provides parameters such as contour, shape and density and the system gives a suggestion about the BI-RADS classification.

METHODS

Initially, values of malignancy were defined for each image descriptor, according to the BI-RADS standard. When analyzing contour, for example, our method considers the matching of features and linguistic variables. Next, we created the fuzzy inference system. The generation of membership functions was carried out by the Fuzzy Omega algorithm, which is based on the statistical analysis of the dataset. This algorithm maps the distribution of different classes in a set.

RESULTS

Images were analyzed by a group of physicians and the resulting evaluations were submitted to the Fuzzy Omega algorithm. The results were compared, achieving an accuracy of 76.67% for nodules and 83.34% for calcifications.

CONCLUSIONS

The fit of definitions and linguistic rules to numerical models provided by our method can lead to a tighter connection between the specialist and the computer system, yielding more effective and reliable results.

Computer-Aided Strain Evaluation for Acoustic Radiation Force Impulse Imaging of Breast Masses

Acoustic radiation force impulse (ARFI) is a newly developed elastography technique that uses acoustic radiation force to provide additional stiffness information to conventional sonography. A computer-aided diagnosis (CAD) system was proposed to automatically specify the tumor boundaries in ARFI images and quantify the statistical stiffness information to reduce user dependence. The level-set segmentation was used to delineate tumor boundaries in B-mode images, and the segmented boundaries were then mapped to the corresponding area in ARFI images for a gray-scale calculation. A total of 61 benign and 51 malignant tumors were evaluated in the experiment. The CAD system based on the proposed ARFI features achieved an accuracy of 80% (90/112), a sensitivity of 80% (41/51), and a specificity of 80% (49/61), which is significantly better than that of the quantitative B-mode features (p < 0.05). The ARFI features were further combined with the B-mode features, including shape and texture features, to further improve performance (area under the curve [AUC], 0.90 vs. 0.86). In conclusion, the CAD system based on the proposed ARFI features is a promising and efficient diagnostic method.