<p>Detection of brain tumors should be both accurate and timely so that patient outcomes improve, and clinical intervention becomes feasible. This study develops a trustworthy machine learning pipeline for brain tumor identification using magnetic resonance imaging data. The pipeline's primary phases include preprocessing, segmentation, feature extraction, and classification. During the preprocessing phase, intensity standardization and noise removal are done via data normalization and cleaning to enhance image quality. Subsequently, an improved gray-level co-occurrence matrix and Fourier transform are done for more stable feature extraction, and an improved secretary sand cat optimization algorithm is done for more efficient feature selection and computation. These are then distinguished as optimized features by an ensemble model of deep belief networks, recurrent neural networks, and convolutional neural networks to identify tumor and non-tumor regions. Experimental results demonstrate the better accuracy, sensitivity, and specificity of the proposed method, and thus it is a trustworthy tool for automated brain tumor detection and diagnosis. Experimental validation confirms the putative pipeline's ability to achieve high accuracy, sensitivity, and specificity in the early detection and diagnosis of brain tumors.</p>