Lightweight-CancerNet: a deep learning approach for brain tumor detection.

Abstract

Detecting brain tumors in medical imaging is challenging, requiring precise and rapid diagnosis. Deep learning techniques have shown encouraging results in this field. However, current models require significant computer resources and are computationally demanding. To overcome these constraints, we suggested a new deep learning architecture named Lightweight-CancerNet, designed to detect brain tumors efficiently and accurately. The proposed framework utilizes MobileNet architecture as the backbone and NanoDet as the primary detection component, resulting in a notable mean average precision (mAP) of 93.8% and an accuracy of 98%. In addition, we implemented enhancements to minimize computing time without compromising accuracy, rendering our model appropriate for real-time object detection applications. The framework's ability to detect brain tumors with different image distortions has been demonstrated through extensive tests combining two magnetic resonance imaging (MRI) datasets. This research has shown that our framework is both resilient and reliable. The proposed model can improve patient outcomes and facilitate decision-making in brain surgery while contributing to the development of deep learning in medical imaging.