In recent years, the real-time diagnosis in the E-health is a widely used practice. Employing distributed computing systems, it is possible to obtain excellent results, avoiding long delays and invasive processes. However, the data processing stage, generally assigned on standard computational CPU environments, is a critical aspect, especially when the computational complexity of the numerical method used for the analysis is very high. In this paper, we consider as case of study the analysis of electrocardiogram (ECG) signals. In order to obtain a diagnosis as quickly as possible, we propose to exploit the computational power of Graphics Processing Unit (GPU) environment. Using GPUs on High Performance Computing (HPC), the signal processing step can be accelerated by speeding the whole diagnosis procedure. More in detail, we designed and implemented a GPUparallel algorithm, for ECG signals denoising based on the Non Local Means (NLM) method. This method is well suited for parallelization and multithreading implementation, especially for GPU architectures. The results show a significant improvement, in terms of execution time, of the entire healthcare practice procedure, with a percentage gain of 96% with respect to the sequential version on standard CPU environment.