ISE-YOLO: Improved Squeeze-and-Excitation Attention Module based YOLO for Blood Cells Detection

Accurate detection of human peripheral blood cells is of great significance to assist doctors to diagnose blood-related diseases. Traditional clinical detection of peripheral blood cells is usually identified by manual microscopy. However, such artificial analysis and processing methods are easily affected by subjective factors, which will cause certain errors. In recent years, the convolutional neural network has been applied to various medical image processing tasks and has shown satisfactory performance. However, traditional CNNs are limited by the lack of feature expression ability. This work introduces the idea of visual attention mechanism into the deep learning detection model and design an improved Squeeze-and-Excitation based YOLO-v3 detection model (ISE-YOLO). This model adds our improved SE module into the different structural blocks of the YOLO to strengthen the network information discrimination of input features and improve the detection performance. Experimental results demonstrate that the proposed ISE-YOLO improves the performance over 96.5% on WBCs, 92.7% on RBCs, and 89.6% on platelets, and outperforms other advanced classification methods.

Automated summary

The accurate detection of human peripheral blood cells is crucial for diagnosing blood-related diseases. Traditional clinical detection methods are easily influenced by subjective factors, leading to errors. The introduction of visual attention mechanism into the deep learning detection model has significantly improved the performance, outperforming other advanced methods.