Segnet with Unet3+ and EfficientNet: a novel framework of brain tumour segmentation and classification model by multiscale attention-based deep learning techniques with hybrid heuristic improvement using 3D MRI brain images

ABSTRACT

An adaptive deep learning is recommended to segment and classify the brain tumor using 3D MRI images. Initially, the original 3D MRI images are gathered and fed into pre-processing, which is accomplished using the 'Contrast Limited Adaptive Histogram Equalization (CLAHE)' technique. The pre-processed image is given as input to 'Multiscale Self-guided Attention based Hybrid Adaptive Networks (MSA-HAN)', where it encompasses Segnet with Unet3+ for segmentation. In this model, certain parameters are optimized using the Position Updating of Black Widow and Shark Smell (PUBWSS). Finally, the segmented region is subjected into the Adaptive Atrous Spatial Pyramid Pooling (ASPP) based EfficientNet (AA-ENet) that considers MobileNet and DenseNet. Thus, the model's performance is evaluated, and results are carried out with diverse measures. The performance of the offered method shows 95.74%, and 95.80% in terms of accuracy and F1-score. Hence, the results declare that the developed model achieves higher segmentation accuracy than other approaches.

KEYWORDS:

• Brain tumour
• 3D magnetic resonance images
• contrast limited adaptive histogram equalization
• multiscale self-guided attention-based hybrid adaptive networks
• Segnet with Unet3+
• position updating of black widow and shark smell
• MobileNet and DenseNet
• adaptive atrous spatial pyramid pooling-based EfficientNet

Additional information

Notes on contributors

D Ramya

Ramya D completed her BTech in Information Technology from Priyadarshini Engineering College, Vellore, affiliated to Anna University, India, in the year 2006. She earned her PG degree, MTech in Information Technology, in Vel Tech Dr. RR and Dr. SR Technical University, in 2012. Currently, she is pursuing PhD in SRM Institute of Science and Technology, Chennai. Her Research areas are Machine Learning, Deep Learning and Image Processing. She has eight years of teaching experience. She is working as an Assistant Professor in Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai. She has published seven papers in international journals and conferences.

C Lakshmi

Dr. Lakshmi C holds Ph.D in Computer Science and Engineering from SRM University and Professor in the department of Computational Intelligence, School of Computing, SRM Institute of Science and Technology. Her main area of research interest includes pattern Recognition, image processing, machine learning and web services. She is a Life time member of the Indian Society for Technical Education (ISTE), International Association of Computer Science and Information Technology, Singapore and International Association of Engineers-IAENG, Hong Kong. She has published several papers in well-known peer-reviewed journals.