.SLE-HYMIXNET: HYBRID CONVOLUTIONAL-TRANSFORMER ARCHITECTURE WITH CORAL ORDINAL LEARNING FOR SYSTEMIC LUPUS ERYTHЕМАТОSUS SEVERITY PREDICTION

Systemic lupus erythematosus (SLE) is a chronic autoimmune disorder that may range from mild to severe, affecting several organs and causing extensive inflammation. Depending on the patients disease activity level, immunosuppressive regimen selection, monitoring frequency, and treatment intensity are all dictated by an accurate severity classification. Manually interpreting various clinical and laboratory signs is a time-consuming, subjective, and potentially variable approach that is used in conventional SLE Disease Activity Index (SLEDAI) severity assessments. Plus, most of the current computational approaches use basic machine learning classifiers that dont take into account the natural development of severity in SLE and instead consider the levels as separate groups. This highlights the need developing a deep learning system that is automated, objective, and ordinal-aware in order to reliably predict the severity of diseases using common clinical biomarkers. We developeda new hybrid deep learning architecture called SLE-HyMixNet to estimate the ordinal severity of SLE from numerical clinical data in order to overcome these difficulties. A combination of 1D Convolutional Neural Networks (1D-CNNs) and Transformer encoders is what the suggested framework, SLE-HyMixNet, is all about. If we compare our suggested strategy to other current ones, we find that it outperforms them all with an overall accuracy of 99.25 percent.