Recognition of honeycomb lung in CT images based on improved MobileNet model.

PURPOSE: The research is to improve the efficiency and accuracy of recognition of honeycomb lung in CT images. METHODS: Deep learning methods are used to achieve automatic recognition of honeycomb lung in CT images, however, are time consuming and less accurate due to the large amount of structural parameters. In this paper, a novel recognition method based on MobileNetV1 network, multiscale feature fusion method (MSFF), and dilated convolution is explored to deal with honeycomb lung in CT image classification. Firstly, the dilated convolution with different dilated rate is used to extract features to obtain receptive fields of different sizes, and then fuse the features of different scales at multiscale feature fusion block is used to solve the problem of feature loss and incomplete feature extraction. After that, by using linear activation functions (Sigmoid) instead of nonlinear activation functions (ReLu) in the improved deep separable convolution blocks to retain the feature information of each channel. Finally, by reducing the number of improved deep separable blocks to reduce the computation and resource consumption of the model. RESULTS: The experimental results show that improved MobileNet model has the best performance and the potential for recognition of honeycomb lung image datasets, which includes 6318 images. By comparing with 4 traditional models (SVM, RF, decision tree, and KNN) and 11 deep learning models (LeNet-5, AlexNet, VGG-16, GoogleNet, ResNet18, DenseNet121, SENet18, InceptionV3, InceptionV4, Xception, and MobileNetV1), our model achieved the performance with an accuracy of 99.52%, a sensitivity of 99.35%, and a specificity of 99.89%. CONCLUSION: Improved MobileNet model is designed for the automatic recognition and classification of honeycomb lung in CT images. Through experiments comparative analysis of other models of machine learning and deep learning, it is proved that the proposed improved MobileNet method has the best recognition accuracy with fewer the model parameters and less the calculation time.

Comparative analysis of computed tomography texture features between pulmonary inflammatory nodules and lung cancer / 中华肿瘤杂志

Objective@#To investigate the value of computed tomography (CT) texture analysis in differential diagnosis of inflammatory and malignant pulmonary nodules.@*Methods@#The image data of 54 patients with lung cancer and 36 patients with pulmonary inflammatory nodules were retrospectively collected in our hospital. All the patients received chest CT scan. CT texture analysis of entropy, correlation degree and contrast ratio were performed by the MaZda software. The receiver operating characteristic curve (ROC) was established and the area under the curve (AUC) was calculated to evaluate the value of CT texture analysis in differential diagnosis of inflammatory and malignant pulmonary nodules.@*Results@#In the lung cancer group, the value of entropy, correlation degree and contrast ratio were 1.58±0.07, 0.02±0.17 and 8.79±2.59, respectively. In the inflammatory nodules group, the value of entropy, correlation degree and contrast ratio were 1.51±0.04, 0.22±0.16 and 12.53±2.24, respectively. The differences were all statistically significant (P values were 0.008, 0.027, and 0.006, respectively) between two groups. There was not statistically significant difference (P>0.05) in the CT values between the lung cancer group and the inflammatory nodule group based on the non-contrast enhanced CT scan. Meanwhile, there was no statistically significant difference (P>0.05) in the value of entropy, correlation degree or contrast ratio between two groups based on arterial phase or venous phase of contrast enhanced CT. The ROC analysis showed that the AUC in differentiating the lung cancer and inflammatory nodules was 0.821, 0.778 and 0.875, respectively. The AUC of combination of three phases was 0.931, which was higher than the AUC of entropy, correlation degree and contrast ratio respectively (P<0.01). The sensitivity was 88.9%, and the specificity was 87.5%.@*Conclusion@#CT texture analysis is a high-potential image analysis method, which can provide more information for the differential diagnosis of benign and malignant pulmonary nodules.

Computed tomography quantitative analysis of components: a new method monitoring the growth of pulmonary nodule.

BACKGROUND: The estimation of the growth of solitary pulmonary nodules by using non-invasive methods is increasingly gaining clinical importance for performing the timely adequate treatment of these nodules. PURPOSE: To evaluate the application value of computed tomography (CT) quantitative analysis of components for dynamic assessment of the growth of solitary pulmonary nodules, and compare this approach with three-dimensional (3D) volumetric measurement of pulmonary nodules. MATERIAL AND METHODS: The imaging data of 21 patients who had undergone multiple follow-up CT scans for solitary pulmonary nodules were retrospectively analyzed, and the total volume of pulmonary nodules and the percentage change in the total volume of pulmonary nodules after multiple follow-up CT scans were measured using 3D volume measurement software. The volume of solid components in pulmonary nodules was measured using CT quantitative analysis; the percentage change in the volume of solid components across examinations was calculated; and the percentage change in the total volume of pulmonary nodules was compared and contrasted with the percentage change in the volume of solid components in the pulmonary nodules. RESULTS: All 21 cases were malignant adenocarcinomas. In the 21 cases of malignant nodules, the 3D volumes of the nodules and solid components were both increased, with the percentage change in the volume of the solid components (115.78-418.91%, 130.45 ± 119.48) significantly different from the percentage change in the total volume of the nodules (78.56-105.73% , 42.34 ± 32.17) (P = 0.02). CONCLUSION: By measuring volume changes in solid components in the nodules, CT quantitative analysis offers more sensitive and earlier evaluation of the dynamic growth of the nodules than measurement of volume changes in the nodules alone.

Volumetric measurement pulmonary ground-glass opacity nodules with multi-detector CT: effect of various tube current on measurement accuracy--a chest CT phantom study.

RATIONALE AND OBJECTIVES: The purpose of this study was to evaluate the effect of various tube currents on the accuracy of volumetric measurements of ground-glass opacity (GGO) nodules using a chest phantom. MATERIALS AND METHODS: A chest phantom containing 13 artificial GGO nodules with known volumes was scanned using a 64-slice computed tomographic scanner at different tube currents (30, 60, 90, 120, 150, 180, and 210 mA). Volumetric measurements were performed using software. The relative percentage error and the absolute percentage error between the volume measures on computed tomography and the reference-standard volumes were calculated. Correlations between the mean absolute percentage error and the mean attenuation of nodules and between the ratio of solid component and the mean attenuation of nodules were analyzed. RESULTS: The relative percentage errors showed that there was substantial underestimation of nodule volumes at 30, 60, and 90 mA and substantial overestimation of volumes at 120, 150, 180, and 210 mA, but there was no statistically significant difference in absolute percentage errors (P = .876). Pearson's correlation coefficient of the mean absolute percentage errors of nodules on volumetric measurement versus the mean attenuation value of nodules showed a negative correlation, and the ratio of solid component to whole nodule versus the mean attenuation of nodules showed a positive correlation. CONCLUSION: Volume measurement is a promising method for the quantification of GGO nodule volume. It is important to know that different tube currents can affect the accuracy of volumetric measurements.