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.

This paper presents a modular approach to the high-frequency simulation of large networks. By utilizing the latency insertion method (LIM) and by studying the stability criteria of partitions of different latencies in the circuit, a robust method is formulated that is able to perform transient simulations significantly faster than the conventional LIM method. The LIM method is also extended to handle dependent sources, and a general stability criterion for selecting the time step, independent of the topology of the circuit, is proposed. The new method is verified with existing commercial tools for circuit simulations, and the improvement in run time is also depicted.

This paper investigates the possibility of using visual imagery tasks, which are mental imagery tasks that involve imagining the images of objects perceptually without seeing them, as a control paradigm that can increase the control’s dimensionality of electroencephalography (EEG)-based braincomputer interfaces. Specifically, we propose an EEG-based approach for decoding visually imagined objects by using the Choi-Williams time-frequency distribution to analyze the EEG signals in the joint timefrequency domain and extract a set of twelve time-frequency features (TFFs). The extracted TFFs are used to construct a multi-class support vector machine classifier to decode various visually imagined objects. To validate the performance of our proposed approach, we have recorded an EEG dataset for 16 healthy subjects while imagining objects that belong to four different categories, namely nature (fruits and animals), decimal digits, English alphabet (capital letters), and arrow shapes (arrows with different colors and orientations). Moreover, we have designed two performance evaluation analyses, namely the channel-based analysis and feature-based analysis, to quantify the impact of utilizing different groups of EEG electrodes that cover various regions on the scalp and the effect of reducing the dimensionality of the extracted TFFs on the performance of our proposed approach in decoding the imagined objects within each of the four categories. The experimental results demonstrate the efficacy of our proposed approach in decoding visually imagined objects. Particularly, the average decoding accuracies obtained for each of the four categories were as high as 96.67%, 93.64%, 88.95%, and 92.68%.