A nonlinear recurrent encoders for early detection of strep throat infection to prevent acute rheumatic fever

Rheumatic Fever or Acute Rheumatic Fever (RF) and Rheumatic Heart Disease (RHD) are commonly suffered by people of all ages around the world. RF usually begins with strep throat infections, skin problems, joint pain etc. Clinical tests and medical informatics are useful to save patients from Streptococcus Group-A bacteria using medical treatments and detection mechanism. Clinical data based statistical analysis approaches assist medical diagnosis model in detecting diseases based on standard decision. Standard decision making frameworks are effective when using these approaches. Nevertheless, existing mechanisms cannot detect or predict the incidence of strep throat infection at an earlier stage based on unclear datasets or complex datasets. The recent computational healthcare informatics systems use Machine Learning (ML) and Deep Learning (DL) techniques, but they are not sufficient in terms of their field integration. The proposed work implements a nonlinear recurrent auto encoder using complex data evaluation procedures to accurately detect and predict the strep throat infections accurately. In this regard, the proposed model builds on the technical features of Denoised Variational Stacked Auto Encoders (DVSE), nonlinear regression computations and Long Shot Term Memory (LSTM) to perform recurrent data analysis practices. Compared with the existing techniques such as Blood Serum Content Analysis model (BSCA), Smartphone based Strep Throat Detection (SSTD) approach and Logistic Regression Based RF Detection approach (LRRF), the multi-level medical data evaluation (throat images and blood samples) provides 10–15% more accuracy for strep throat detection. As shown in the implementation section, the proposed nonlinear recurrent system has the potential to detect strep throat infections early.