Efficient framework for brain tumor detection using different deep learning techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Brain tumors are a critical malignancy resulting from uncontrolled cell proliferation.
• Current tumor classification relies on post-surgical biopsies, delaying diagnosis.
• Deep learning shows promise in enhancing medical image analysis for diagnosis.

Purpose of the Study:

• To introduce and evaluate deep learning models for classifying brain magnetic resonance images (MRIs) as either normal or tumorous.
• To compare a novel Convolutional Neural Network (CNN), BRAIN-TUMOR-net, against established transfer-learning models.

Main Methods:

• Developed BRAIN-TUMOR-net, a CNN trained from scratch for brain tumor classification.
• Compared BRAIN-TUMOR-net with pre-trained InceptionResNetv2, Inceptionv3, and ResNet50 models.
• Validated model performance on three public MRI datasets using k-fold cross-validation and unsupervised clustering for segmentation.

Main Results:

• BRAIN-TUMOR-net demonstrated superior performance across all three MRI datasets.
• Achieved accuracy levels of 100%, 97%, and 84.78% on the tested datasets.
• K-fold cross-validation confirmed robust classification capabilities.

Conclusions:

• BRAIN-TUMOR-net is a highly effective deep learning model for brain tumor classification from MRI data.
• The proposed model offers a promising non-invasive approach to aid in the early diagnosis of brain tumors.
• Deep learning, particularly CNNs, significantly advances medical imaging analysis in oncology.