Prediction Of EGFR Mutations In Non-small Cell Lung Cancer Based On Radiomic Features

According to the statistical reports of the World Health Organization in recent years,it can be found that,cancer as a tumour disease,its morbidity and mortality are gradually increasing,which has caused a great impact on human health.With the continuous development of machine learning in the medical field,many methods of machine learning have been able to more accurately predict the occurrence and metastasis of cancer,seize the best opportunity to treat the disease,and then effectively control the mortality of cancer.In view of this situation,this article collected 397 radiology characteristics of 100 patients in a large hospital in Changchun from 2016 to 2018,and based on statistical methods to establish mathematical models of the collected radiology characteristics,and then predict Whether non-small cell lung cancer patients have EGFR mutations.There are many influencing factors for EGFR mutation.For a more effective and predictive model is obtained.In this paper,the 397 collected radiology features are selected by the LASSO algorithm to obtain the 15 radiology features that have the greatest impact on EGFR mutations,and then to reduce the dimensionality of the radiology features and eliminate redundant feature.Next,three predictive models based on machine learning algorithms are established for the obtained radionomics features,namely Gaussian process,naive Bayes model,and LightGBM algorithm.After a simple comparative analysis,it can be found that LightGBM algorithm has more obvious advantages.According to this prediction model,it can effectively judge whether EGFR mutations occur in patients with non-small cell lung cancer,and can provide new ideas for the subsequent research,and at the same time provide more valuable diagnostic information for clinicians struggling to the front line.