Radiology: Convolution-based neural network software can improve the detection rate of malignant tumors in chest

Background
the additional benefits of detecting malignant lung tumor nodules in chest syllomas using deep convolutional neural network (DCNN) software are yet to be verified by multicenter studiespurpose
study was designed to evaluate the difference between the performance of radiologists in detecting malignant lung nodules in chest syllables with the aid of deep learning-based DCNN software and the performance of simple radiologists or DCNN softwarematerials and methods
researchers retrospectively identified 600 cases of chest and 200 normal chest tablets with lung malignanciesThe chest slices of each lung tumor were tested by CT or pathology to contain 1 lung malignant noduleThe chest tablets were independently analyzed and marked by 12 radiologists from 4 medical centersUse deep learning computer-aided diagnostics to make money to train, test and verify 1930 chest tablets to detect suspicious lung nodulesThe radiologist was reviewed with the help of DCNN softwareUsing logistics regression and poisson regression analysis, DCNN software, radiologists and radiologists were evaluated to detect the sensitivity and false positive number of lung nodules in each image with the aid of DCNN softwareresultsthe average sensitivity of radiologists when they re-examine the chest tablets with the help of DCNN software (from 65.1% (1375 / 2112; 95% CI: 62.0%, 68.1%) to 70.3% (1484 / 2112); 95% CI: 67.2%, 73.1%, P .001), decreased in number of false positive sms (from 0.2 (488 / 2400; 95% CI: 0.18, 0.22) to 0.18 (422 / 2400; 95% CI: 0.16, 0.2), P 001) Of the 12 radiologists in this study, a total of 104 / 2400 chest tablets were correctly diagnosed (from false negative to positive or true to negative); conclusion
s
this study shows that radiologists have performed better in the diagnosis of malignant pulmonary nodules in chest syllables, aided by deep learning-based neural convolution network software