Chagas disease infects over 7 million people worldwide with mortality of 10,000 deaths annually. Currently various studies confine the modulation of ECG signals indicating progression and severity of Chagas disease. Hence this work emphasis on developing fast convergence kernels based ordinary differential equations (fck-ODEs) based 1D-Convolutional neural network (1D-CNN) that learns the variation and abnormalities of both physical parameters of P,Q,R,S,T and data driven dynamics of ECG signals to predict the disease, its progression and severity. At first, a set of basis equations for the fast convergence kernel based ODE to be written by using the variation and abnormalities and data driven dynamics of ECG signals. Let the ECG signals of patients infected with Chagas disease (f) be represented as the state of the continuous system based on the ODE model by approximating the dynamics of f, by means of 1D-CNN as, du/dt≈conv1D(t,u(t),θ). During forward pass, a stack of Linear Convolutional Network Layers 〖(L〗i),∀i∈{0..n} as, Cov1D(x)=Ln 〖L〗(n-1) 〖..L〗0 (x) are cascaded to analyze each of the ECG channel sampler units. During error back propagation, the weights with respect to their previous layer are learnt by using derivatives of the fck-ODEs. After extracting the feature maps, the transformation of output mapping is to be done with Gated Recurrent Network (GRU) units. GRU deals with the temporal dynamics of regions such as, PR segment, ST segment, PR interval, ST interval, QRS complex, R-peak. The memory cells of GRU units are helpful to remember the important temporal relationship between the input ECG signals over a period of time. The final output layer of GRU unit derives the relationship dynamics between two consecutive R-peaks (R-R segment) that captures the complete systolic and diastolic phases of the cardiac cycle, and finally generates the desired label to mark the severity levels / stages of Chagas disease.