This research aims to address the problem of discriminating benign cysts from malignant masses in breast ultrasound (BUS) images based on Convolutional Neural Networks (CNNs). The biopsy-proven benchmarking dataset was built from 1422 patient cases containing a total of 2058 breast ultrasound masses, comprising 1370 benign and 688 malignant lesions. Three transferred models, InceptionV3, ResNet50, and Xception, a CNN model with three convolutional layers (CNN3), and traditional machine learning-based model with hand-crafted features were developed for differentiating benign and malignant tumors from BUS data. Cross-validation results have demonstrated that the transfer learning method outperformed the traditional machine learning model and the CNN3 model, where the transferred InceptionV3 achieved the best performance with an accuracy of 85.13% and an AUC of 0.91. Moreover, classification models based on deep features extracted from the transferred models were also built, where the model with combined features extracted from all three transferred models achieved the best performance with an accuracy of 89.44% and an AUC of 0.93 on an independent test set.

This study reviews the technique of convolutional neural network (CNN) applied in a specific field of mammographic breast cancer diagnosis (MBCD). It aims to provide several clues on how to use CNN for related tasks. MBCD is a long-standing problem, and massive computer-aided diagnosis models have been proposed. The models of CNN-based MBCD can be broadly categorized into three groups. One is to design shallow or to modify existing models to decrease the time cost as well as the number of instances for training; another is to make the best use of a pretrained CNN by transfer learning and fine-tuning; the third is to take advantage of CNN models for feature extraction, and the differentiation of malignant lesions from benign ones is fulfilled by using machine learning classifiers. This study enrolls peer-reviewed journal publications and presents technical details and pros and cons of each model. Furthermore, the findings, challenges and limitations are summarized and some clues on the future work are also given. Conclusively, CNN-based MBCD is at its early stage, and there is still a long way ahead in achieving the ultimate goal of using deep learning tools to facilitate clinical practice. This review benefits scientific researchers, industrial engineers, and those who are devoted to intelligent cancer diagnosis.

Automatic classification of breast mass lesions in mammographic images remains an unsolved problem. This paper explored the technique of transfer learning to tackle this problem. It utilized the convolutional neural network (CNN) of GoogLeNet and AlexNet pre-trained on a large-scale visual database. The performance was evaluated a new dataset in terms of the area under the receiver operating characteristic curves (AUC). Results demonstrate that GoogLeNet (AUC=0.88) outperforms AlexNet (AUC=0.83) and other state-of-the-art traditional approaches in breast cancer diagnosis. The technique of transfer learning not only overcomes the unsatisfactory performance of traditional approaches, but also breaks the obstacle of limited samples for building deep CNNs.

Ring artifacts in cone beam computed tomography (CBCT) images are caused by pixel gain variations using flat-panel detectors, and may lead to structured non-uniformities and deterioration of image quality. The purpose of this study is to propose a method of general ring artifact removal in CBCT images. This method is based on the polar coordinate system, where the ring artifacts manifest as stripe artifacts. Using relative total variation, the CBCT images are first smoothed to generate template images with fewer image details and ring artifacts. By subtracting the template images from the CBCT images, residual images with image details and ring artifacts are generated. As the ring artifact manifests as a stripe artifact in a polar coordinate system, the artifact image can be extracted by mean value from the residual image; the image details are generated by subtracting the artifact image from the residual image. Finally, the image details are compensated to the template image to generate the corrected images. The proposed framework is iterated until the differences in the extracted ring artifacts are minimized. We use a 3D Shepp-Logan phantom, Catphan©504 phantom, uniform acrylic cylinder, and images from a head patient to evaluate the proposed method. In the experiments using simulated data, the spatial uniformity is increased by 1.68 times and the structural similarity index is increased from 87.12% to 95.50% using the proposed method. In the experiment using clinical data, our method shows high efficiency in ring artifact removal while preserving the image structure and detail. The iterative approach we propose for ring artifact removal in cone-beam CT is practical and attractive for CBCT guided radiation therapy.

Partly due to the use of exhaustive-annotated data, deep networks have achieved impressive performance on medical image segmentation. Medical imaging data paired with noisy annotation are, however, ubiquitous, but little is known about the effect of noisy annotation on deep learning based medical image segmentation. We studied the effect of noisy annotation in the context of mandible segmentation from CT images. First, 202 images of head and neck cancer patients were collected from our clinical database, where the organs-at-risk were annotated by one of twelve planning dosimetrists. The mandibles were roughly annotated as the planning avoiding structure. Then, mandible labels were checked and corrected by a head and neck specialist to get the reference standard. At last, by varying the ratios of noisy labels in the training set, deep networks were trained and tested for mandible segmentation. The trained models were further tested on other two public datasets. Experimental results indicated that the network trained with noisy labels had worse segmentation than that trained with reference standard, and in general, fewer noisy labels led to better performance. When using 20% or less noisy cases for training, no significant difference was found on the segmentation results between the models trained by noisy or reference annotation. Cross-dataset validation results verified that the models trained with noisy data achieved competitive performance to that trained with reference standard. This study suggests that the involved network is robust to noisy annotation to some extent in mandible segmentation from CT images. It also highlights the importance of labeling quality in deep learning. In the future work, extra attention should be paid to how to utilize a small number of reference standard samples to improve the performance of deep learning with noisy annotation.

Blind image quality assessment can be modeled as feature extraction followed by score prediction. It necessitates considerable expertise and efforts to handcraft features for optimal representation of perceptual image quality. This paper addresses blind image sharpness assessment by using a shallow convolutional neural network (CNN). The network takes single feature layer to unearth intrinsic features for image sharpness representation and utilizes multilayer perceptron (MLP) to rate image quality. Different from traditional methods, CNN integrates feature extraction and score prediction into an optimization procedure and retrieves features automatically from raw images. Moreover, its prediction performance can be enhanced by replacing MLP with general regression neural network (GRNN) and support vector regression (SVR). Experiments on Gaussian blur images from LIVE-II, CSIQ, TID2008 and TID2013 demonstrate that CNN features with SVR achieves the best overall performance, indicating high correlation with human subjective judgment.