Differential diagnostic value of benign and malignant solid lung nodules based on deep learning

There are numerous imaging methods for clinical screening and diagnosis of solid lung nodules, but they all have certain limitations. This paper selected patients with lung nodules in the People’s Hospital of Province X as an example. C.T. scans were performed on the patients with lung nodules to obtain their imaging histologic features. Then, based on the generative adversarial network in deep learning and using self-supervised learning to optimize the generative negative discriminator, a semi-supervised GAN model was established for the identification and predictive diagnosis of benign and malignant solid lung nodules. A regression analysis model was constructed and data analysis was performed to identify the independent risk factors related to the malignancy of solid lung nodules. The results showed that the patient’s tumor diameter became the most significant independent risk factor for the benign-malignant nature of lung nodules, with an OR of 3.421, which showed a significant difference at the 1% level. The I.C.C. value of each feature of solid lung nodules was more critical than 0.85 in the impactomics feature screening, and the A.U.C. value of benign and malignant prediction diagnosis of solid lung nodules using semi-supervised GAN model reached 0.98. Combining CT impactomics with deep learning can improve the differential prediction of benign and malignant diagnosis of solid lung nodules, which can provide high value for the clinical workers to treat solid lung nodules. The combination of C.T. impactomics and deep learning can improve the differential diagnosis of benign and malignant solid lung nodules and provide high value for clinical workers to treat solid lung nodules.