目标检测在肺癌病理诊断中的应用

Application of object detection algorithm in pathology diagnosis of lung cancer

基于深度学习的目标检测已经在交通、军事等多个领域得到了广泛研究,并取得了显著成果.为进一步研究其在医学图像诊断中的可用性,提出将目标检测算法的2类经典模型SSD和Faster RCNN应用于肺癌病理图像的病灶检测.实验对比发现,Faster RCNN平均处理1张图像所消耗的时间大概是SSD时长的4.6倍,但其识别准确度、精度均比SSD更高,模型性能更优.研究结果表明,目标检测算法能够实现肺癌病理图像的智能诊断,提高肺癌诊断率.

Object detection based on deep learning has been widely studied in many fields such as transportation and military,and has achieved remarkable results.In order to further study its usability in medical image diagnosis,two classical models of object detection algorithms,SSD and Faster RCNN,are proposed to appliy to the lesion detection of lung cancer pathological images.The experimental comparison shows that the average processing time of Faster RCNN is about 4.6 times longer than that of SSD,but the recognition accuracy and precision are higher than that of SSD,and the model performance of Faster RCNN is much better.The research results show that the object detection algorithm can realize the intelligent diagnosis of lung cancer pathological images and improve the diagnosis rate of lung cancer.