Feature detection and peak detection are one of the first steps of mass spectrometry data processing. This data comes in large volumes; thus, the processing needs to be optimized, not overloaded. State-of-the-art clustering algorithms can not perform feature detection for several reasons. First issue is the volume of the data, second is the disparity of the sampling frequency in the MZ and RT axis. Here we show the data transformation to utilize the clustering algorithms without the need to redefine its kernel. Data are first pre-clustered to obtain regions that can be processed independently. Then we transform the data so that the numerical differences between consecutive points should be the same in both space axes. We applied a set of clustering algorithms for each region to find the features, and we compared the result with the Gridmass peak detector. These findings may facilitate better utilization of the 2D clustering method as feature detectors for mass spectra.