Brain Tumor Detection and Classification by Hybrid CNN-DWA Model Using MR Images

Area of Science:

• Medical image processing
• Artificial intelligence in healthcare
• Neuro-oncology imaging

Background:

• Early and accurate detection of brain tumors is crucial for effective patient treatment.
• Existing methods for brain tumor analysis using MR images have limitations in performance and accuracy.
• Developing advanced computational models can significantly aid clinicians in diagnosing brain tumors.

Purpose of the Study:

• To propose a novel hybrid deep learning model for enhanced brain tumor detection and classification.
• To improve the accuracy and performance of automated brain tumor analysis using MR images.
• To assist medical professionals in the early diagnosis of brain tumors.

Main Methods:

• A hybrid deep convolution neural network and deep watershed auto-encoder (CNN-DWA) model was developed.
• The model incorporates six phases: image input, preprocessing, matrix representation, hybrid model application, classification/detection, and performance evaluation.
• Training and validation were conducted using five large-scale MR brain image databases (BRATS 2012-2015, ISLES-SISS 2015).

Main Results:

• The proposed hybrid CNN-DWA model demonstrated high performance with an accuracy of approximately 98%.
• The model achieved a low validation loss of 0.1, indicating robust learning.
• Comparative analysis showed the hybrid model outperformed standalone CNN, DNN, and DWA models in accuracy and detection capabilities.

Conclusions:

• The novel hybrid CNN-DWA model offers superior performance for brain tumor detection and classification.
• This advanced model can facilitate the development of computer-aided diagnostic systems for early brain tumor detection.
• The findings support the utility of the proposed model in improving diagnostic accuracy for clinicians.