Year: 2024
Type: article
Abstract: Diagnosis of brain tumours quickly and accurately is crucial for effective treatment and better patient outcome.Although the traditional diagnosis procedures like MRI, CT scan and biopsies are useful, they pose significant human inconsistencies.This research explores five of the most popular transfer learning techniques in CNN to find an optimal model for classification of brain tumours.Pre-trained models-VGG16, ResNet50, DenseNet121, InceptionResNetV2 and InceptionV3 have been used to find the optimal model for this task.The used dataset includes 7023 MRI images divided into four categories: glioma, meningioma, pituitary tumour, and no tumour.Experimental results highlight DenseNet121's superior performance, achieving a validation accuracy of 94%, precision, recall, and F1-score of 0.94, outperforming other models.This study shows that deep learning and transfer learning can significantly improve the accuracy and efficiency of medical image analysis, leading to better healthcare outcomes.
Authors: Anil Poudyal, Niruta Devkota, Gayatri Budha
Institution: London Metropolitan University
Language: English
Subfield: Neurology
Field: Neuroscience
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4403083427