A method for detecting leukocytes in bone marrow slices based on a regional recommendation convolutional neural network

Abstract

The invention discloses a method for detecting leukocytes in bone marrow slices based on regional recommendation convolution, which comprises the following steps: performing preprocessing, leukocyte frame selection and leukocyte category labeling, and increasing the number of data sets as a training data set; constructing a regional recommendation convolutional neural network model, taking a ResNet 50 network and an FPN network as a feature extraction network, connecting the output of the FPN network to an RPN network, and allowing the RPN to output a recommendation region to a Faster-RCNN network, and allowing the Faster-RCNN network to output the leukocyte region, and taking the leukocyte category and the category probability corresponding to the leukocyte region as prediction results; training the constructed regional recommendation convolutional neural network model, adjusting parameters until convergence, and obtaining a trained regional recommendation convolutional neural networkmodel; And taking the image of the bone marrow cell slice as input, and outputting a prediction result by the trained regional recommendation convolutional neural network model. The detection methodprovided by the invention is high in leukocyte detection accuracy and precision, and the indexes of the algorithm are measured by the accuracy rate and can reach more than 90%.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to a method for detecting white blood cells in bone marrow cell slices based on a regional convolutional neural network. Background technique [0002] Leukocytes in traditional bone marrow slices are detected manually: under a microscope, professional pathologists scan the entire slice with the naked eye through the movement of the slice, and count the number of abnormal white blood cells to diagnose leukemia. This kind of work is heavy and time-consuming. With the increase of film time, the error rate also increases. [0003] With the continuous development of technology, the detection of white blood cells in bone marrow slices can be initially screened with the help of computers. Object detection is an important research direction of computer vision, and its task is to complete the frame selection and classification of target areas through computer algorithms. ...

AI Technical Summary

Problems solved by technology

[0005] However, due to the similar shape and size of white blood cells in bone marrow cell slices, the characteristics are not obvious, and it is difficult to distinguish them. The accuracy of the current computer white blood cell detection algorithm is not high and the precision is poor.

[0006] The target detection problem is to use the model to frame the target that needs to be identified in the picture and predict the category. The accuracy of the position and size of the frame box and the accuracy of the classification are important indicators for judging the target detection, and at the same time make the target detection problem quite difficult.

With the rapid development of deep learning in recent years, the problem of target detection has made some breakthroughs, but in the field of complex medical images, target detection is still a relatively difficult problem.

Images