An application of machine learning to haematological diagnosis

Abstract

Quick and accurate medical diagnoses are crucial for the successful treatment of diseases. Using machine learning algorithms and based on laboratory blood test results, we have built two models to predict a haematologic disease. One predictive model used all the available blood test parameters and the other used only a reduced set that is usually measured upon patient admittance. Both models produced good results, obtaining prediction accuracies of 0.88 and 0.86 when considering the list of five most likely diseases and 0.59 and 0.57 when considering only the most likely disease. The models did not differ significantly, which indicates that a reduced set of parameters can represent a relevant “fingerprint” of a disease. This knowledge expands the model’s utility for use by general practitioners and indicates that blood test results contain more information than physicians generally recognize. A clinical test showed that the accuracy of our predictive models was on par with that of haematology specialists. Our study is the first to show that a machine learning predictive model based on blood tests alone can be successfully applied to predict haematologic diseases. This result and could open up unprecedented possibilities for medical diagnosis.

Artificial Intelligence based Models for Screening of Hematologic Malignancies using Cell Population Data

“…A recent study using CPD showed Random Forest algorithm as the best model with two www.nature.com/scientificreports www.nature.com/scientificreports/ practices, using all parameters and reduced parameters. It showed the accuracy of 59% for 181 parameters and accuracy of 57% for 61 parameters 22 . Another study took CPD data with 103 parameters for prediction of relapse in childhood with Acute Lymphoblastic Leukemia 33 .…”

“…Though there is substantial literature of AI and ML in healthcare research, most of the research focuses in the fields of Cancer, Neurology and Cardiology 11,[13][14][15][16][17][18][19][20][21] . In addition, the literature lacks successful applications of ML that deal with complex medical diagnostic fields like Hematology 22 . Blood tests are the most common measure to diagnose the hematological diseases in the laboratories and clinicians need the hematological parameters to analyze the numerical patterns, deviations and relations; and that's where ML algorithms can come into action by performing intelligent handling, detection and utilization of these parameters, and developing models to predict the future diagnosis and outcomes 22 .…”

Automated prediction of mastitis infection patterns in dairy herds using machine learning

“…The application of machine learning in the identification of disease has often focussed on image recognition, for example the accurate classification of skin cancer 15 , and retinal disease 16 . The use of machine learning techniques with diagnostic data, such as haematological results, has also been described, and were able to achieve an accuracy of 88% when compared with specialist haematologists; outperforming internal medicine specialists in achieving a correct diagnosis 17 . Machine learning has been used in the diagnosis of diabetes using features such as sex, age and blood pressure 18 , and the diagnosis of cardiac arrhythmia 19 , with random forest algorithms receiving particular attention in their ability to outperform many other machine learning algorithms in classification exercises 20 .…”

Using Machine Learning to Identify Health Outcomes from Electronic Health Record Data

“…Given this complex diagnostic decision-making process, machine learning could be useful for systematically detecting disease-specific patterns from blood test results to improve the detection of hematologic diseases, especially diseases that lack widely accepted diagnostic criteria, such as basophilic leukemias (30). Gunčar et al (28) built random forests to perform differential diagnosis of hematologic diseases based on a wide array of structured blood test results (181 attributes from 179 different blood tests). Among 43 possible hematological disease categories, their final random forest correctly identified the true disease category – as recorded among patients’ admission and discharge diagnoses in the hospital system – with an overall accuracy (ie, percent correct classification) of 0.60, which was on par with the average accuracy of six hematology specialists (0.62) and better than the average accuracy of eight non-hematology internal medicine specialists (0.26).…”