A Comparative Analysis of Brain Tumor Classification and Prediction techniques by applying MRI Images encompassing SVM and CNN with Transfer Learning Method.

Main Article Content

Jitin Sharma, Prashant C. Chavan, T. B. Patil, Supriya C Sawant, Milind Gaykawad


Brain tumor is a serious disease leading to a life-threatening condition when worsen resulting in brain cancer. With a rapid fall in 5-year survival rate, brain cancer is the root cause for a large number of deaths worldwide. Diagnosis of brain tumor and detection of nature of cancer with higher accuracy at earliest possible stage is very important for proper treatment that will help patients have a better chance of surviving the disease. As per our study in this paper we conducted comparative research on various machine learning methods for detection of brain tumor at an early stage using Magnetic Resonance Images (MRI). MRIs is one of the most preferable diagnostic techniques for Brain Tumor image diagnosing as it gives more detailed images of the brain structures and blood vessels without using ionized radiations and any physical movement of the patient. This paper majorly focuses on a comparative analysis of various approaches for classification and detection of brain tumor and at last analysing the recent best technique. The process of detection followed in research till date follows Image Pre-Processing, Image Segmentation, Feature Extraction, and Support Vector Machine Classification. Various sources of datasets have been used that includes limited amount of MRI Images. Our analysis is also focused on the limitations and shortcomings of current algorithms. Thus, by comparing various researches, we summarized that when support vector machine (SVM) was used to classify brain tumor images, accuracy was trustworthy for patients but usage of limited amount of dataset reduced the essence of the prediction as using larger dataset can surely bring worthy conversion. We sum up that CNN with Transfer Learning model when used for brain tumor detection increase the reliability of prediction as this model when used with Paediatric Pneumonia detection and Covid-19 detection gave impressive results with increased dataset volume, resulting in early detection of disease and rapid treatment. Thus, for better results, classification and detection of Brain Tumor using CNN with Transfer Learning will be a good option.

Article Details