TRANSFORMING STROKE PREVENTION IN CAMEROON: AN AI-BASED METHOD FOR DETECTING ATRIAL FIBRILLATION

Abstract Complex human diseases, particularly cardiovascular diseases, have a significant impact globally and in Cameroon. Among these diseases, stroke stands out due to its high mortality rate, exacerbated by the lack of medical infrastructure and appropriate technical resources for early diagnosis and effective management. In this context, prevention, especially through the identification of risk factors, is of crucial importance. Atrial fibrillation, an independent risk factor for stroke, increases the likelihood of being affected by this condition fivefold. However, its diagnosis currently relies on the manual classification of cardiac arrhythmias, a tedious, time-consuming method, prone to errors, and difficult to access in a country like Cameroon, where the doctor-to-population ratio is extremely low (1.1 doctors per 10,000 inhabitants). Faced with these challenges, this research proposes an innovative method for the automatic recognition of atrial fibrillation, and normal heart rhythm using wavelet transform and a convolutional neural network (CNN) based on transfer learning with the AlexNet architecture. Experimental results obtained by testing the approach demonstrates interesting accuracy across three international databases (the MIT-BIH, Chapman ECG Database, and The 2017 PhysioNet/CinC Challenge databases) while being computationally efficient for Cameroonian healthcare infrastructure. This automated solution could revolutionize the diagnosis of cardiac arrhythmias by reducing errors, speeding up the process, and addressing the lack of medical resources. In conclusion, this research offers a promising perspective for improving stroke prevention and management in Cameroon and other resource-limited regions. It also paves the way for future applications in other cardiovascular diseases, thereby strengthening healthcare systems in developing countries. Keywords: ECG, CNN, Stroke, atrial fibrillation, cardiac arrhythmias