Predicting pre-collection umbilical cord blood clotting using advanced machine learning algorithms

Umbilical cord blood is a valuable source of stem cells used in transplants to treat various diseases including leukemia, lymphoma and genetic disorders. However, cord blood clotting during the collection process can reduce sample quality and quantity and impact its efficacy in cord blood banking. This article aims to predict pre-collection cord blood clotting in donors using advanced machine learning techniques.

In this retrospective study, data was gathered using 928127 samples available in the fetal cord blood bank, and with using supervised machine learning classification algorithms, including decision tree, naïve Bayes, K-Nearest Neighbors, Support vector machine, Random forest, Majority voting and Multilayer perceptron, prediction of cord blood clotting was performed on the Royan cord blood bank database and their performance was compared using evaluation metrics such as Accuracy, Precision, Recall, and F1 Score.

In this study, the algorithm accuracy of Decision Tree was 0.80, Naive Bayes was 0.63, K-Nearest Neighbors was 0.83, Support Vector Machine was 0.65, Random Forest was 0.84, Majority Voting Classifier was 0.81, and Multilayer Perceptron was 0.74.

Conclusions  :

In this study, the performance of Random Forest and K-Nearest Neighbors algorithms demonstrated the best accuracy showing that machine learning algorithms can predict prenatal cord blood clotting with high accuracy which can help prevent sampling of clotted specimens in order to reduce costs and storage problems.