AN EFFICIENT LIGHTWEIGHT MODEL FOR BRAIN TUMOR CLASSIFICATION

Brain tumors are fatal illnesses that affects brain, the fundamental organ to live as normal human in this world.They are one of the causes of death among other cancers and their early detection is crucial to avoid it.Deep learning approaches demonstrate their effectiveness on brain tumor classification to avoid subjective and timeconsuming manual diagnosis.The models are heavy in general, but, as Madagascar is a developing country, it faces lots of challenges in the health sectors: lack of infrastructures, unequal distribution of health resources, household poverty and low purchasing power This situation prevents diagnosis of brain tumor in its early stage.The present work deals with this problem by introducing a robust lightweight model that classifies effectively brain tumor, and could run to a low-resource environment like a smartphone.Evaluated with two different datasets, it achieves an accuracy and validation accuracy of 100%, showing that the proposed lightweight model is really effective on brain tumor classification.