AN AI-BASED DECISION SUPPORT SYSTEM FOR THE DIAGNOSIS OF HEMATOLOGICAL DISORDERS: THE CASE OF ACUTE MYELOID LEUKEMIA (AML) IN QATAR

Abstract

Mapping a cancer patient pathway consists of analyzing the steps followed by the patient from symptoms to diagnosis and treatment. It intends to gain a general overview of the whole patient experience and allows an understanding of the different weaknesses and challenges to be addressed in the patient journey so as to ensure quality cancer care. For hematological cancer patients in Qatar, the pathway from symptoms to diagnosis is quite complex due to the lack of clarity in blood cancer symptomatology and the patient’s ignorance of the malignancy’s warning signs. This can negatively affect the patient’s chances of recovery and usually results in huge delays that could have been avoided if the patient diagnosis took place earlier in the process. Fortunately, it has been proven by several studies that hematologic malignancies are perfectly curable at early stages, allowing for more treatment options. Therefore, this thesis work is designed to support early diagnosis of hematological cancer patients in Qatar, namely Acute Myeloid Leukemia (AML), at the primary healthcare level using artificial intelligence (AI). We propose here a framework that specifically relies on the patient’s complete blood count (CBC) test, considering it is the first diagnostic routine blood test performed among all the stages, which can help assess the patients’ risk of having blood cancer and ensure that they’re timely referred to professionals, when needed. Using linear and non-linear Support Vector Machine algorithms, K Nearest Neighbor, Random Forest Algorithm and 3 feature selection techniques, CBC was successfully used to classify AML and non-AML patients and the model performance evaluation exhibited very high accuracies.

Date of Award	2022
Original language	American English
Awarding Institution	HBKU College of Science and Engineering