Biomedical Engineering Reference
In-Depth Information
Fig. 3. Part of mass peak distribution with bins generated by PSB. The total number
of spectra is 382.
and is insensitive to various parameters in preprocessing steps. In the PSB
process, there is no precondition of bin location or bin width. The bins
are generated completely from the distribution of peaks in spectra. The
method introduces the window width function and the shifting unit. They
are the tools to characterize the spectrum data. The parameters used in
the functions of window width as well as the shifting are experimental. The
method only involves the presence of peaks, not the intensities of peaks
and the peak determination. Also, the PSB results are not aected much
by small variation in the data registration.
Remark 1. PSB can be used to generate a more reasonable mean spec-
trum compared to the one derived in [20]. The direct comparisons between
the projecting spectra between normal tissues and cancer tissues should be
very helpful in biomarkers discovery as well.
Remark 2. Similar to many other computer algorithms, one concern
of PSB is the stability of its performance when the size of the spectra is
exceeding to certain number, say 700. In this case, we suggest adding in
some statistical sub-sampling techniques when one applys PSB.
Search WWH ::
Custom Search