رسائل ماجيستير
English
2021
Machine learning for identifying severity clusters in Omani patients with systemic lupus erythematosus
Sumaya Khalifa Aamir AL-Sawafiyah , Aliya AL-Ansari , Hamza Zidoum, Imran Khan
كلية العلوم-جامعة السلطان قابوس · عمان
Systemic lupus erythematosus (SLE) is an autoimmune disease characterized by the presence of autoantibodies targeting self-antigens, immune complex formation, and immune dysregulation, resulting in damage to multiple organs. The disease can affect the kidneys, skin, blood cells, and nervous system. The natural course of SLE is unpredictable; patients may experience many years of mild symptoms or present with acute, life-threatening conditions. Although advancements in diagnostic techniques and treatment strategies have improved prognosis, there remains a need for deeper understanding and more targeted therapeutic options.
Aims:
This study aims to:
(a) Identify severity clusters among Omani patients with SLE,
(b) Detect features associated with disease severity, and
(c) Examine the correlation between the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) and Physician Global Assessment (PGA) within each subgroup.
Methods:
Data were collected from Sultan Qaboos University Hospital (SQUH) and included demographic, clinical, laboratory, and treatment information. The dataset underwent several stages: data cleaning, feature extraction, and exploratory data analysis to understand data types and distributions. Three clustering methods—hierarchical agglomerative clustering, K-Means clustering, and spectral clustering—were applied to group patients. Clustering results were evaluated by examining correlations with SLEDAI and PGA scores.
Results:
Exploratory data analysis revealed that joint pain was the most common symptom among Omani SLE patients, followed by positive anti-dsDNA antibodies, low complement levels (C3, C4), acute cutaneous lupus (ACL), renal disorders, and hemolytic anemia.
Clustering analysis identified two distinct patient groups: a mild and a severe cluster. Patients in the severe cluster had a higher prevalence of renal disorders, hemolytic anemia, positive anti-dsDNA antibodies, and low complement levels. These patients also exhibited cumulative clinical manifestations such as malar rash and proteinuria and required more aggressive treatment, including cyclophosphamide, mycophenolate mofetil, and azathioprine. The mild disease activity cluster was associated mainly with joint pain, low complement levels, and positive anti-dsDNA antibodies, but less severe organ involvement.