An online decision support system for diagnosing hematologic malignancies by flow cytometry immunophenotyping

Immunophenotyping by flow cytometry (FCM) plays an important role in the diagnosis and subclassification of hematopoietic malignancies. Currently, almost all laboratories manually analyze FCM for interpretation and data entry. This time consuming and labor intensive process calls for a technology based streamlined automation. A decision support system to interpret flow results will be helpful for both hematopathologists and laboratory personnel in any busy laboratories and therefore reduce the healthcare cost. To meet this challenge, there have been several attempts made to potentially automate the diagnostic process of lymphoma and leukemias with only partial resolution of the current drawbacks of manual analysis, let alone the clinical acceptance. In this study, a knowledge-based decision support system to interpret online FCM results for hematologic malignancies has been developed as a complete Client-Server application. The listmode data files are loaded to the system where gating, dot plot, histogram and contour plot can be performed. Upon gating, the CD marker results are generated as a percentage with associated positive or negative designation. The knowledge base for a final diagnosis is based on the current World Health Organization (WHO) classification of hematologic malignancies which is depicted in a semantic network and further embedded in an eXtensible markup language (XML). Differential diagnosis is taken into consideration in our decision support system. The confidence level for a particular differential diagnosis is based on the sensitivity and specificity of a particular CD marker for making a clinical diagnosis, combined with clinical experience as well. Java programming is used to implement the inference engine where tree structure and search algorithm are employed. A set of 273 FCM listmode data files are fed into the system and diagnosis was correctly included in top three differential diagnoses in 94% of all cases tested. In conclusion, the website (http://www.flowcytometryonline.com) has been set up for online FCM analysis and decision support and data transportation. The system is expected to facilitate clinical diagnosis of hematologic malignancies and assist resident teaching as well.