The classification of human brain tumors with machine learning

Abstract

Abstract This paper aims to report human brain tumor classification based on machine learning. Brain tumors have long been regarded as one of the deadliest tumors. Conventionally, without the assistance of automation tools, brain tumor diagnosis is a skill-demanding work for the doctors and a time-consuming burden on the medic system. Applying machine learning to clinical medicine and classifying brain tumors enables physicians to increase their diagnosis correctness and allows the health system to treat more patients. This paper covers the methods used in the research: the dataset, network models, algorithm, and implementation. The experiment results and discussion section follow this. Focusing on comparing and discussing convolutional neural networks’ performance and resource consumption under different network settings and network models. In the experiment, by feeding the MRI images to the network, the network automates the results indicating whether a tumor is detected in the image and classifies the underlying tumor. The experiment results indicate Machine Learning’s great accuracy of 93.38 percent in classifying brain tumors and the significant correlations between the model’s performance and resource cost, and network settings as well as model complexity.