Medical images of coronary artery plaque are always accompanied by the situation of extreme class imbalance. The traditional two-step methods locate the region of interest (ROI) in the sample firstly, and then segment the sample within the ROI. On the other hand, the traditional resampling methods use resampling strategies to increase the number of minority class samples to mitigate the effects of class imbalance. These two types of methods either make the network structure more complex or decrease training efficiency and performance of the model due to the increase of samples. This paper proposes a method including a novel focal weighted accuracy loss function and improved metrics evaluation algorithms to address the issues in the segmentation of coronary artery calcification plaque mentioned above. Experimental results on the selected dataset show the proposed method increased the training speed and improved the segmentation performance of the model without performing resampling on the dataset. Specifically, the F1-score was 0.873 5, the precision was 0.929 6, and the recall was 0.823 8. The F1-score was largely improved compared with the method using focal loss function. Furthermore, compared with methods with multiple models and methods via resampling the minority class samples, research results demonstrate that the proposed method improved the accuracy and efficiency in coronary artery plaque segmentation while has a shorter training time, which lays the foundation for improving the efficiency and scientific nature of diagnosing related diseases in the future.
Aiming at the problem that the unbalanced distribution of data in sleep electroencephalogram(EEG) signals and poor comfort in the process of polysomnography information collection will reduce the model's classification ability, this paper proposed a sleep state recognition method using single-channel EEG signals (WKCNN-LSTM) based on one-dimensional width kernel convolutional neural networks(WKCNN) and long-short-term memory networks (LSTM). Firstly, the wavelet denoising and synthetic minority over-sampling technique-Tomek link (SMOTE-Tomek) algorithm were used to preprocess the original sleep EEG signals. Secondly, one-dimensional sleep EEG signals were used as the input of the model, and WKCNN was used to extract frequency-domain features and suppress high-frequency noise. Then, the LSTM layer was used to learn the time-domain features. Finally, normalized exponential function was used on the full connection layer to realize sleep state. The experimental results showed that the classification accuracy of the one-dimensional WKCNN-LSTM model was 91.80% in this paper, which was better than that of similar studies in recent years, and the model had good generalization ability. This study improved classification accuracy of single-channel sleep EEG signals that can be easily utilized in portable sleep monitoring devices.
