Emil Popa1,
Enrico Capobianco2, Jan Willem van der Veen3, Ronald de
Beer4, Dirk van Ormondt5, Danielle Graveron-Demilly1
1Universit Lyon 1, Villeurbanne, France;
2CRS4 Bioinformatics Laboratory, Pula (Cagliari), Italy; 3NIH,
NIMH, Bethesda, United States; 4Delft University of Technology; 5Applied
Physics, Delft University of Technology, Delft, Zuid Holland, Netherlands
This
work concerns a new way of dealing with in vivo spectral lineshapes for the
case that a reference line is not available. It is based on dual-criterion
non-linear least-squares fitting. All data-points are used simultaneously, in
conjunction with the general a priori knowledge that a lineshape can be
confined to a certain frequency region. The experimental lineshape at hand
can be arbitrary, including asymmetric shapes. Modelling with analytical
mathematical functions like splines, wavelets, or decaying sinusoids is circumvented.
As a result, setting of hyper-parameters by a user is avoided. This favours
automation.