Safety in medical signal analysis

A number of issues are raised about the possibility of erroneous results obtained from signal analysis techniques that could lead to an incorrect diagnosis or an inadequate decision. An example is the use of a filter to attenuate interference when the (unidimensional) signal of interest contains wavelets. The output wavelets may be distorted by the filter in such a way that their interpretation leads to an incorrect diagnosis. In medical imaging, peripheral hot spots and sunburst streaking in computed tomography, blurring of tomographic reconstruction of cardiac images (as in single-photon-emission computed tomography, or SPECT) due to organ movement, and artifact due to image enhancement are among the difficulties introduced by signal analysis methodology. Some general recommendations are put forward as to how to deal with these kinds of problems.<>