Document

Machine learning for identifying severity clusters in Omani patients with systemic lupus erythematosus.

Publisher
Sultan Qaboos University.
Gregorian
2021
Language
English
English abstract
Background: Systemic lupus erythematosus (SLE) is an autoimmune disease characterized by autoantibodies directed against self-antigens, immune complex formation, and immune dysregulation, resulting in damage to any organ. The disease can affect body organs, such as kidneys, skin, blood cells, and the nervous system. The natural history of SLE is unpredictable; patients may present with many years of symptoms or with an acute, life-threatening disease. Although the prognosis of SLE has improved with the advent of better detection methods and enhanced treatment strategies, the need remains for a better understanding of the disease and more targeted treatment options. Aims: To: (a) identifying severity clusters in Omani patients with Systemic Lupus Erythematosus, (b) detecting features related to diseases severity, and (c) examining the correlation between disease activity index (SLEDAI) and physician global assessment (PGA) with each subgroup. Methods: Our method consists of collecting a broad spectrum of data from Sultan Qaboos University Hospital (SQUH). The data include demographic, clinical, laboratory, and therapy. After the collection, the data goes through several stages. The first stage is data cleaning and feature extraction. The next stage is to exploring data analysis to identify the data types and exploring the data distribution to get a full view and understand the dataset. Then, three clustering methods are used which are hierarchical agglomerative clustering, K-Means clustering, and spectral clustering to cluster our dataset. Next, the clustering results are evaluated using correlation with SLEDAI and PGA. iv Results: The exploratory data analysis shows that joint pain is the most common symptom in Omani SLE patients, followed by positive anti-dsDNA antibody, low complement (C3, C4), acute cutaneous lupus (ACL), renal disorder, and hemolytic anemia. The clustering analysis results show two separate patients clusters which are mild cluster and severe. Patients in the severe cluster have a higher prevalence of the renal disorder, hemolytic anemia, anti-dsDNA antibody, and low complements (C3, C4). As a result of analyzing cumulative manifestations and treatment, the severe cluster patients suffer from malar rash and proteinuria with higher use of cyclophosphamide, mycophenolate mofetil, and azathioprine. The second cluster is mild disease activity, and it is associated with joint pain, low complements (C3, C4), and a positive anti-dsDNA antibody.
Category
Theses and Dissertations