Objectives The precise diagnosis of brain tumors using magnetic resonance imaging (MRI) presents several complex challenges. Traditional methods have primarily focused on grayscale anatomical data, often neglecting vital physiological indicators such as tissue temperature, which reflects metabolic activity. The main objective of this study is to develop an integrated AI-driven MATLAB framework that enhances tumor diagnosis accuracy by combining MRI features with thermal and textural biomarkers. Methods To address this gap, a comprehensive MATLAB pipeline was developed that integrates deep learning segmentation, morphological analysis, thermal estimation, texture quantification, and malignancy prediction, utilizing datasets from Kaggle and Figshare. The first step involved creating a specialized model to identify tumor regions and evaluate their size and shape. A compact three-layer convolutional neural network (CNN) was then employed to classify images into categories, including glioma, meningioma, pituitary tumor, and healthy tissue. Results It was found that gliomas had the most significant areas, ranging from 72.75 to 6365 mm 2 , and displayed the most irregular shapes. The CNN model achieved high accuracy, with near-perfect detection of healthy cases and an F1 score of 99.2%. However, the recall for pituitary tumors was 29%, and the precision for meningiomas was 48.6%, indicating areas for improvement. For temperature estimation, a formula was derived: T = 37.0 + 0.7 × log (1 + Area). This formula suggests that malignant lesions could reach temperatures as high as 42.2 °C, while benign tumors remained at or below 38.5 °C. Despite hardware limitations that limited training to eight of the planned 30 epochs, the results demonstrate the potential of combining thermal and textural biomarkers with MRI. Conclusion These findings demonstrate significant innovation potential and highlight the need to transition toward graphical processing unit-accelerated training to refine temperature baselines. By integrating multimodal features, major advances in clinical applications can be achieved, ultimately enhancing patient outcomes.
Almomany et al. (Sun,) studied this question.