Prediction of postoperative complications of pediatric cataract patients using data mining

• DOI: 10.1186/s12967-018-1758-2
• Publication Date: 2019-01-02T19:32:09Z
• Authors: Kai Zhang; Xiyang Liu; Jiewei Jiang; Wangting Li; Shuai Wang; Lin Liu; Xiaojing Zhou; Liming Wang
• Keywords: Male; 0301 basic medicine; Clinical Decision-Making; Cataract Extraction; 03 medical and health sciences; Postoperative Complications; Naïve Bayesian; Data Mining; Humans; Child; Research; R; Infant; Bayes Theorem; Prognosis; 3. Good health; Medical decision making system; Genetic feature selection; ROC Curve; Area Under Curve; Child, Preschool; Association rules mining; Medicine; Female; Algorithms; Random forest

The common treatment for pediatric cataracts is to replace the cloudy lens with an artificial one. However, patients may suffer complications (severe proliferation into visual axis and abnormal high intraocular pressure; SLPVA AHIP) within 1 year after surgery factors causing these are unknown.Apriori algorithm employed find association rules related complications. We use random forest (RF) Naïve Bayesian (NB) predict datasets preprocessed by SMOTE (synthetic minority oversampling technique). Genetic feature selection exploited real features complications.Average classification accuracies in three binary problems over 75%. Second, relationship between performance number of tree studied. Results show except gender age at (AS); other attributes Except secondary IOL placement, operation mode, AS area cataracts; SLPVA. gender, laterality; AHIP. Next, mined out. Then additional 50 data were used test RF NB, both then obtained 65% problems. Finally, we developed a webserver assist doctors.The postoperative can be predicted. found. that about provide reference doctors.