Prediction of Treatment Option by Human Chorionic Gonadotropin (hCG) Levels in Ectopic Pregnancy using Machine Learning

Abstract

Yoko Nagayasu, Rei Mitsuhashi, Hikaru Murakami, Ruri Nishie, Natsuko Morita, Atsushi Daimon, Misa Nunode, Hiroshi Maruoka, Satoe Fujiwara, Satoshi Tsunetoh, Daisuke Fujita and Masahide Ohmichi

Aim: The objective of this study is to construct a model using a random forest to predict the treatment option of ectopic pregnancy based on hCG levels, as well as to confirm the model’s accuracy.

Methods: We selected 17 variables related to ectopic pregnancy and extracted data from our records for cases of possible ectopic pregnancy. We then divided the cases into two groups: 1) laparoscopic surgery and 2) MTX or conservative treatment. We created a model for predicting the prognosis of ectopic pregnancy. Afterward, we confirmed the model’s accuracy using the test data. Additionally, we compared the model’s accuracy with that of two specialized obstetrician- gynecologists (OB/GYNs) specialists who judged the same data. This study was approved by our ethics committee.

Results: One hundred and twenty-eight patients were eligible for this research, of whom 52.3% (67) underwent laparoscopic surgery and 7.0% (9) had emergent laparoscopic surgery. MTX and conservative treatment, including normal pregnancies or miscarriages, were 25.0% (32) and 25.0% (32), respectively. The model’s accuracy using a random forest was 87.3%, and the area under curve (AUC) was 0.784. The two OB/GYNs judged the same data with respective accuracies of 77.3% and 79.7%.

Conclusion: In conclusion, this model using a random forest was superior to the judgment of specialists. Moreover, this research is new in the fact that it has presented a numerical model involving multiple risks that, until now, have been judged empirically by humans. In the future, it may help develop and elucidate a more extensive prediction system for ectopic pregnancy.

PDF