Brain Tumor Detection with Vision Transformers and Faster R-CNN

Please enable javascript in your browser.

Momen Tayefeh, Mehrshad | 2024

0 Viewed

Type of Document: M.Sc. Thesis
Language: English
Document No: 57083 (52)
University: Sharif University of Technology, International Campus, Kish Island
Department: Science and Engineering
Advisor(s): Hemmatyar, Ali Mohammad Afshin; Ghahramani Ghahramani, Amir Ali
Abstract:
In the field of cancer diagnosis, particularly brain tumors, the priority is to achieve highly accurate tumor detection. Deep learning has shown remarkable potential in object detection tasks, making it a valuable tool for identifying brain tumors. We have proposed a new method for combining the strength of Faster R-CNN in detecting objects and the ability of Vision Transformer’s (Faster-VIT) ability to extract essential features. The proposed method significantly improves the accuracy and efficiency of brain tumor detection in MRI images. We have called the proposed combination Faster-VIT. To assess the effectiveness of the proposed method, we have utilized the Br35H dataset, comprising 801MRI images for training, validation, and testing phases. The proposed method improves Mean Average Precision (MAP) and Mean Average Recall (MAR) metrics in comparison to earlier methods. It should be noted that 0.9% improvement in MAP50 score is achieved while the state-of-the-art methods have over 80 million parameters. The proposed method has a substantially smaller parameter count of 19 million
Keywords:
Deep Learning ; Vision Transformer ; Brain Tumor ; Object Detection ; Faster R-CNN Algorithm ; Cancer Diagnosis