Medical image classification plays an important role in disease diagnosis since it can provide important reference information for doctors. The supervised convolutional neural networks (CNNs) such as DenseNet provide the versatile and effective method for medical image classification tasks, but they require large amounts of data with labels and involve complex and time-consuming training process. The unsupervised CNNs such as principal component analysis network (PCANet) need no labels for training but cannot provide desirable classification accuracy. To realize the accurate medical image classification in the case of a small training dataset, we have proposed a light-weighted hybrid neural network which consists of a modified PCANet cascaded with a simplified DenseNet. The modified PCANet has two stages, in which the network produces the effective feature maps at each stage by convoluting inputs with various learned kernels. The following simplified DenseNet with a small number of weights will take all feature maps produced by the PCANet as inputs and employ the dense shortcut connections to realize accurate medical image classification. To appreciate the performance of the proposed method, some experiments have been done on mammography and osteosarcoma histology images. Experimental results show that the proposed hybrid neural network is easy to train and it outperforms such popular CNN models as PCANet, ResNet and DenseNet in terms of classification accuracy, sensitivity and specificity.

Purpose: Quantification of carotid plaques has been shown to be important for assessing as well as monitoring the progression and regression of carotid atherosclerosis. Various metrics have been proposed and methods of measurements ranging from manual tracing to automated segmentations have also been investigated. Of those metrics, quantification of carotid plaques by measuring vessel‐wall‐volume (VWV) using the segmented media‐adventitia (MAB) and lumen‐intima (LIB) boundaries has been shown to be sensitive to temporal changes in carotid plaque burden. Thus, semi‐automatic MAB and LIB segmentation methods are required to help generate VWV measurements with high accuracy and less user interaction.Methods: In this paper, we propose a semiautomatic segmentation method based on deep learning to segment the MAB and LIB from carotid three‐dimensional ultrasound (3DUS) images. For the MAB segmentation, we convert the segmentation problem to a pixel‐by‐pixel classification problem. A dynamic convolutional neural network (Dynamic CNN) is proposed to classify the patches generated by sliding a window along the norm line of the initial contour where the CNN model is fine‐tuned dynamically in each test task. The LIB is segmented by applying a region‐of‐interest of carotid images to a U‐Net model, which allows the network to be trained end‐to‐end for pixel‐wise classification.Results: A total of 144 3DUS images were used in this development, and a threefold cross‐validation technique was used for evaluation of the proposed algorithm. The proposed algorithm‐generated accuracy was significantly higher than the previous methods but with less user interactions. Comparing the algorithm segmentation results with manual segmentations by an expert showed that the average Dice similarity coefficients (DSC) were 96.46 ± 2.22% and 92.84 ± 4.46% for the MAB and LIB, respectively, while only an average of 34 s (vs 1.13, 2.8 and 4.4 min in previous methods) was required to segment a 3DUS image. The interobserver experiment indicated that the DSC was 96.14 ± 1.87% between algorithm‐generated MAB contours of two observers' initialization.Conclusions: Our results showed that the proposed carotid plaque segmentation method obtains high accuracy and repeatability with less user interactions, suggesting that the method could be used in clinical practice to measure VWV and monitor the progression and regression of carotid plaques.

Measurement of total-plaque-area (TPA) is important for determining long term risk for stroke and monitoring carotid plaque progression. Since delineation of carotid plaques is required, a deep learning method can provide automatic plaque segmentations and TPA measurements; however, it requires large datasets and manual annotations for training with unknown performance on new datasets. A UNet++ ensemble algorithm was proposed to segment plaques from 2D carotid ultrasound images, trained on three small datasets (n = 33, 33, 34 subjects) and tested on 44 subjects from the SPARC dataset (n = 144, London, Canada). The ensemble was also trained on the entire SPARC dataset and tested with a different dataset (n = 497, Zhongnan Hospital, China). Algorithm and manual segmentations were compared using Dice-similarity-coefficient (DSC), and TPAs were compared using the difference (ΔTPA), Pearson correlation coefficient (r) and Bland-Altman analyses. Segmentation variability was determined using the intra-class correlation coefficient (ICC) and coefficient-of-variation (CoV). For 44 SPARC subjects, algorithm DSC was 83.3-85.7%, and algorithm TPAs were strongly correlated (r = 0.985-0.988; p <; 0.001) with manual results with marginal biases (0.73-6.75) mm$^2$ using the three training datasets. Algorithm ICC for TPAs (ICC = 0.996) was similar to intra- and inter-observer manual results (ICC = 0.977, 0.995). Algorithm CoV = 6.98% for plaque areas was smaller than the inter-observer manual CoV (7.54%). For the Zhongnan dataset, DSC was 88.6% algorithm and manual TPAs were strongly correlated (r = 0.972, p <; 0.001) with ΔTPA = -0.44±4.05 mm$^2$ and ICC = 0.985. The proposed algorithm trained on small datasets and segmented a different dataset without retraining with accuracy and precision that may be useful clinically and for research.

Vessel-wall-volume (VWV) is an important three-dimensional ultrasound (3DUS) metric used in the assessment of carotid plaque burden and monitoring changes in carotid atherosclerosis in response to medical treatment. To generate the VWV measurement, we proposed an approach that combined a voxel-based fully convolution network (Voxel-FCN) and a continuous max-flow module to automatically segment the carotid media-adventitia (MAB) and lumen-intima boundaries (LIB) from 3DUS images. Voxel-FCN includes an encoder consisting of a general 3D CNN and a 3D pyramid pooling module to extract spatial and contextual information, and a decoder using a concatenating module with an attention mechanism to fuse multi-level features extracted by the encoder. A continuous max-flow algorithm is used to improve the coarse segmentation provided by the Voxel-FCN. Using 1007 3DUS images, our approach yielded a Dice-similarity-coefficient (DSC) of 93.2±3.0% for the MAB in the common carotid artery (CCA), and 91.9±5.0% in the bifurcation by comparing algorithm and expert manual segmentations. We achieved a DSC of 89.5±6.7% and 89.3±6.8% for the LIB in the CCA and the bifurcation respectively. The mean errors between the algorithm- and manually-generated VWVs were 0.2±51.2 mm3 for the CCA and -4.0±98.2 mm3 for the bifurcation. The algorithm segmentation accuracy was comparable to intra-observer manual segmentation but our approach required less than 1s, which will not alter the clinical work-flow as 10s is required to image one side of the neck. Therefore, we believe that the proposed method could be used clinically for generating VWV to monitor progression and regression of carotid plaques.

In the field of disease diagnosis where only a small dataset of medical images may be accessible, the light-weight convolutional neural network (CNN) has become popular because it can help to avoid the over-fitting problem and improve computational efficiency. However, the feature extraction capability of the light-weight CNN is inferior to that of the heavy-weight counterpart. Although the attention mechanism provides a feasible solution to this problem, the existing attention modules, such as the squeeze and excitation module and the convolutional block attention module, have insufficient non-linearity, thereby influencing the ability of the light-weight CNN to discover the key features. To address this issue, we have proposed a spiking cortical model based global and local (SCM-GL) attention module. The SCM-GL module analyzes the input feature maps in parallel and decomposes each map into several components according to the relation between pixels and their neighbors. The components are weighted summed to obtain a local mask. Besides, a global mask is produced by discovering the correlation between the distant pixels in the feature map. The final attention mask is generated by combining the local and global masks, and it is multiplied by the original map so that the important components can be highlighted to facilitate accurate disease diagnosis. To appreciate the performance of the SCM-GL module, this module and some mainstream attention modules have been embedded into the popular light-weight CNN models for comparison. Experiments on the classification of brain MR, chest X-ray, and osteosarcoma image datasets demonstrate that the SCM-GL module can significantly improve the classification performance of the evaluated light-weight CNN models by enhancing the ability of discovering the suspected lesions and it is generally superior to state-of-the-art attention modules in terms of accuracy, recall, specificity and F1 score.