Automation and Updation of Screening Process of Leukemia Considering Features of CBC Reports

Abstract

Data analytics together with Machine Learning (ML) and Artificial Intelligence (AI) techniques bring a new era of automated diagnostics to the healthcare domains. Despite the availability of diagnostic tests, the mortality rate of leukemia is increasing, especially in developing countries. Therefore, there is a need to improve efficiency in the screening processes by supporting healthcare professionals through modern computing resources. This research illustrates a data-driven procedure using ML algorithms for screening leukemia considering significant features of Complete Blood Count (CBC) reports. A data set of 302 CBC reports labeled by health care professionals of eight different hospitals/labs of Islamabad/Rawalpindi has been used along with the 1287 hybrid synthetic data generated by using Burr distribution, to make more generalizable and standard models that might be more robust and reliable in its functioning. Machine Learning methods namely, Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Gradient Boosting (GB) have been used with different combinations of significant features of CBC reports. According to evaluation metrics, the random forest algorithm's performance with 13 features—namely, hemoglobin, hematocrit, red blood cell count, monocyte percentage, platelets count, neutrophil percentage, white blood cell percentage, lymphocyte percentage, mean corpuscular volume, basophil percentage, and lymphocyte count—performs best in comparison to the other methods, with accuracy, precision, recall, specificity, and F1 score of 93% & 97%, 96% & 98% & 96%, 83% & 96%, and 96% & 97%, for local indigenous data and hybrid synthetic data respectively. By using this ML model, we aim to develop a clinical decision support system (CDSS), that aids doctors, particularly hematologists as E-opinion in making future diagnostic and treatment decisions. This smart solution as a Clinical Decision Support System named "Smart Screening Leukemia" aims to be userfriendly, time-saving, easy to adopt, and cost-effective, with the potential to deliver significant benefits in healthcare domains. This study is the one of the first that infers that the ML predictive models based on the significant features of complete blood count (CBC) report alone might be successfully applied to screen leukemia through a web-based application. The Smart Screening Leukemia (SSL) could open unprecedented possibilities for the future of screening and diagnosis of various types of leukemia and other hematological malignancies.