Medical image classification has recently attracted increased attention. Effective feature extraction and learning are key means to improve classification performance. However, in the current study, handcrafted feature are mostly designed with intuitive mode, while the deep feature depends on a large amount of training samples and has weak interpretability. To capture more discriminative features for medical image, a novel feature fusion approach, termed multi-layer visual feature fusion (MLVSF), has been proposed on the basis of low-level, mid-level and deep features. More specifically, by fusing the handcrafted and deep features generated by local binary pattern variant, bag-of-visual-words, convolutional neural network, respectively, MLVSF can effectively enhance the discriminating power of features for medical image recognition. Experimental results on two medical image datasets show that MLVSF can improve convolutional neural networks, and achieve a better classification accuracy in comparison with some state-of-the-art methods.

The existing density clustering algorithms have high error rates on processing data sets with mixed density clusters. For overcoming shortcomings of these algorithms, a double-density clustering method based on Nearest-to-First-in strategy, DDNFC, is proposed, which calculates two densities for each point by using its reverse k nearest neighborhood and local spatial position deviation, respectively. Points whose densities are both greater than respective average densities of all points are core. By searching the strongly connected subgraph in the graph constructed by the core objects, the data set is clustered initially. Then each non-core object is classified to its nearest cluster by using a strategy dubbed as ‘Nearest-to-First-in’: the distance of each unclassified point to its nearest cluster calculated firstly; only the points with the minimum distance are placed to their nearest cluster; this procedure is repeated until all unclassified points are clustered or the minimum distance is infinite. To test the proposed method, experiments on several artificial and real-world data sets are carried out. The results show that DDNFC is superior to the state-of-art methods like DBSCAN, DPC, RNN-DBSCAN, and so on.

The efficient and automatic detection of chest abnormalities is vital for the auxiliary diagnosis of medical images. Many studies utilize computer vision and deep learning approaches involving symmetry and asymmetry concepts to detect chest abnormalities, and achieve promising findings. However, an accurate instance-level and multi-label detection of abnormalities in chest X-rays remains a significant challenge. Here, a novel anomaly detection method for symmetric chest X-rays using dual-attention and multi-scale feature fusion is proposed. Three aspects of our method should be noted in comparison with the previous approaches. We improved the deep neural network with channel-dimensional and spatial-dimensional attention to capture the abundant contextual features. We then used an optimized multi-scale learning framework for feature fusion to adapt to the scale variation in the abnormalities. Considering the influence of the data imbalance and other factors, we introduced a seesaw loss function to flexibly adjust the sample weights and enhance the model learning efficiency. The rigorous experimental evaluation of a public chest X-ray dataset with fourteen different types of abnormalities demonstrates that our model has a mean average precision of 0.362 and outperforms existing methods.

The optimization of proportional-integral-derivative (PID) controller parameters poses a critical challenge in automatic control. Considering the complexity and uncertainty in actual systems, it is difficult to achieve desired dynamic performance, robustness, and stability at the same time. To overcome the inefficiency and low accuracy of conventional tuning methods, an improved particle swarm optimization (PSO) algorithm, logistic-tent Lévy PSO (LTLPSO), is proposed. This algorithm introduces logistic-tent chaotic mapping to initialize the population, enhance the uniformity and diversity of initial particles, and improve global exploration performance. In addition, the position update process integrates a Lévy flight strategy to help particles overcome local optima and improve the speed of convergence. The comparative evaluations with basic PSO, line-decreasing weight PSO, and chaotic mapping PSO confirm that LTLPSO has high convergence speed, accuracy, and stability. MATLAB/Simulink simulations are carried out on a delayed temperature system and a CNC-feed servo system to validate the performance of this algorithm. The results show that, compared with genetic algorithms and seeker optimization algorithms, the LTLPSO-PID controller can reduce the overshoot by up to 68% and the settling time by 30%, significantly improving dynamic and steady-state responses. This method can provide a robust and efficient solution for PID parameter tuning.