Feature selection techniques play a vital role in the real-time medical databases. Since, most of the medical databases contain high dimensionality and large data size, it is difficult to find an essential key feature using traditional feature sub-set selection approaches. Also, conventional medical data filtering techniques fail to find the essential outliers due to large data size and feature space. In this work, a hybrid outlier detection and data transformation approaches are implemented to remove the noise in the medical databases. Proposed data filtering module is applicable to high dimensional data size and feature space for classification problem. Experimental results are simulated on different medical datasets such as tonsil and trauma databases with different feature space size and data size. Simulation results proved that the proposed outlier detection approach has better noise detection rate than the conventional approaches.