In the past half century, the incidence and mortality of lung cancer have increased significantly, it is one of the malignant tumors that seriously threaten people's lives and health. However, due to its complex clinical manifestations, it is difficult to detect in the early stage, and timely treatment and intervention are required. The early diagnosis of lung cancer has a significant impact on the clinical treatment effect.
Our CT Image analysis model uses the neural network model of deep learning algorithm to perform feature learning on a large number of sample images. It realizes medical image recognition, automatically identifies possible lesions, and credibility analysis, which can assist doctors in diagnosis and treatment, and improve doctors' diagnosis efficiency and accuracy.
The specific functions of the automatic detection algorithm for lung nodules in our CT Image analysis model are as follows: Automatically analyze the patient CT image files in DICOM format without the involvement of doctors, and perform preliminary screening of lung nodules through the U-Net neural network model (Figure 1). Check, generate a candidate set of lung nodules, and make full use of the three-dimensional information of CT images through the 3DCNN neural network model (Figure 2) to reduce false positives of the candidate set of lung nodules to improve the accuracy of automatic screening. Finally, the screened lung nodule information is returned, including the center point coordinates and radius of the nodule, reliability, and the slice number of the nodule (Figure 3,4).
Our model uses the gold standard and Dice loss function for performance evaluation：the Dice Score is 0.90, the sensitivity is 0.96, and the accuracy is 0.98.
Figure 1 the U-Net neural network model
Figure 2 the 3DCNN neural network model
Figure 3 the workflow of CT Image analysis model
Figure 4 the output of CT Image analysis model
Accurate
Our CT Image analysis are trained from over 407 patients with tuberculosis. A large amount of training data and robust algorithms make our model's accuracy rate as high as 98%, which is significantly higher than manual recognition accuracy.
Efficient
Supported by high-performance servers and robust algorithms, the CT Image analysis can diagnose a large number of X-ray image in a short period of time and keep working 24/7, which can greatly reduce the workload of doctors and improve their diagnosis effectiveness.
Applicable
Our CT Image analysis model can accurately and quickly complete image screening, and make decision assistance based on AI screening results. According to the test results, our CT Image analysis model can help junior radiologists to improve the quality of diagnosis, so that their diagnostic level can be improved to be equivalent to that of senior radiologists in a short time.
Services