Data preprocessing has become a crucial step in data mining, particularly in socially impactful domains such as healthcare, where the presence of irrelevant or faulty information contained in data can lead to severe, potentially dangerous consequences for patients. To address this challenge, we developed a novel dimensionality reduction algorithm based on fuzzy-rough set theory (DRFRS), which integrates feature selection and instance selection into a unified framework. The fundamental contribution of this approach is the use of the Fuzzy-Rough Feature Selection (FRFS) algorithm to select relevant features, followed by a novel criterion derived from fuzzy boundary region to identify and remove faulty instances. Resulting in an improvement in the accuracy of learning models, ultimately contributing to improved patient outcomes and more equitable access to high-quality care. The technique was evaluated on medical datasets of various sizes to assess the performance of the proposed method.