Multimodal Convolutional Neural Network Models Allow for the Accurate Classification and Grading of Preoperative Meningioma Brain Tumors: Artificial Intelligence and Neural Radiology

Artificial intelligence and neural radiology

Abstract

Magnetic resonance imaging (MRI) and computed tomography (CT) scans are vital for diagnosing meningioma brain tumors. However, human error, image subtleties, cyst growth, and nuances in World Health Organization (WHO) grading significantly impede accuracy. Invasive biopsies remain the only definitive method for meningioma diagnosis. Convolutional Neural Networks (CNNs), machine learning models used in image classification, offer a promising solution. By fine-tuning the pre-trained CNN EfficientNetB0 on various preoperative brain tumors and meningioma subtypes, safer image-based diagnosis can become more robust and accurate. In this study, one CNN model classified multimodal CT and MRI images, while the other performed grading. The first dataset included several tumor types (meningioma, glioma, pituitary, cysts, or none), and the second consisted of meningioma tumors assigned a WHO grade (one to three). The images, from accurately annotated and diverse open-source databases, were normalized, augmented, and skull-stripped. In the training and validation stages, class-average and Focal Tversky loss functions assessed and reduced incorrect outputs. After testing, both CNNs achieved accuracy and precision over 98% with recall and f1 scores over 95%. Additionally, receiver operating characteristic (ROC) area under the curve (AUC) scores above 0.978 indicated strong class discrimination. Lastly, an included attention study demonstrated the model focusing primarily on the tumor mass, rather than on extraneous variables. These findings demonstrate how multimodal CNNs, particularly lightwork models like EfficientNetB0, can serve as more reliable and cost-effective alternatives to invasive biopsies and human evaluation. Their capability to handle complex meningioma cases suggests promising avenues for other tumor types and diagnostic modalities.