Python-based Automatic Detection and Diagnosis for Pathological Changes of Cells in Sputum

As a consequence of poor air quality, the incidence of lung cancer is increasing year by year. To detect and diagnose cell pathological changes in a timely manner and reduce the impact of delayed diagnosis on people’s health, a technology for automatic diagnosis from sputum pathological cell images is desirable. At present, medical institutions mainly depend on manual work to diagnose and identify cell pathological changes, which has the problems of low recognition efficiency, low accuracy, and the possibility of misjudgment due to fatigue. In this study, advanced digital image processing technology and methods are used to diagnose pathological changes of cells in sputum images automatically. A training object detection model based on Python language is built by installing and configuring TensorFlow and using an object detection application programming interface. The data sets are labeled by identification labeling software, and the parameters are reconfigured using the Faster RCNN pre-training model. The training model is used to detect and diagnose sputum pathological cell images automatically. The effects of the sample size and the number of training steps on the successful recognition of a single image are studied.