Grammatical Evolution for Predicting Cervical Cancer Recurrence Risk

Cervical cancer is one of the most common tumors in women and has a high rate of recurrence after surgery. Early detection of symptoms leading to earlier treatment can significantly reduce the risk of patients. Artificial intelligence assists in diagnosing the recurrence of cervical cancer, which can reduce diagnostic costs and increase early identification of symptoms for timely treatment. In this study, grammatical evolution was used to develop a predictive model. The model included the corresponding grammar, evolutionary strategies and adaptations for common influences and important independent features. The grammatical evolution model was compared to the predictive model constructed using decision tree algorithms. The results show that the predictive model based on grammatical evolution performs well in predicting cervical cancer recurrence. Furthermore, due to the flexibility and adaptability of GE, the model can achieve desirable results by adjusting parameters. The model's efficiency improves when using important independent features obtained from the decision tree algorithm as the grammatical evolution inputs. Then, the importance of all features in the model was determined using the permutation importance. In this proposed model, the key features for predicting recurrence are pathologic T, pathologic stage, LNM, surgical margin involvement, and RT target summary in descending order of importance. These results will help clinicians to pay attention to patients with such risk factors before recurrence occurs.