Enhanced Watershed Segmentation Algorithm-Based Modified ResNet50 Model for Brain Tumor Detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Rising incidence of brain tumors necessitates advanced detection and classification methods.
• Traditional machine learning approaches often lack the required accuracy and trustworthiness for clinical application.
• There is a critical need for effective deep feature extraction techniques in neuro-oncology.

Purpose of the Study:

• To develop and validate a novel technique for brain tumor detection and classification.
• To integrate enhanced watershed segmentation with a modified ResNet50 architecture for improved diagnostic performance.
• To extract deep features for effective brain tumor tissue diagnosis.

Main Methods:

• A modified ResNet50 architecture with five convolutional and three fully connected layers was employed.
• Enhanced Watershed Segmentation (EWS) algorithm was integrated with the modified ResNet50 model.
• Stochastic approaches were utilized for developing the enhanced watershed modeling.
• Hybrid deep features were extracted from the ResNet50 model for classification.

Main Results:

• The proposed hybrid deep feature-based modified ResNet50 model achieved 92% classification accuracy.
• The EWS-based modified ResNet50 model demonstrated 90% classification accuracy.
• The method effectively extracts diverse deep features for accurate brain tumor diagnosis.
• Optimal computational efficiency was maintained with high-dimensional deep features.

Conclusions:

• The integrated approach of modified ResNet50 and EWS offers a highly accurate method for brain tumor classification.
• The novel technique provides an effective solution for brain tumor tissue detection and deep feature extraction.
• This research contributes a trustworthy and efficient tool for neuro-oncology diagnostics.