A study of data compression using neural networks and principal component analysis [of pulmonary scintigrams]

PCA analysis is similar to neural networks in data compression of segments that have been `seen´, but is superior in compressing `unseen´ images. The difference between `seen´ and `unseen´ images with respect to tsse (total sum square error) is more pronounced in 32 by 16 pixel segments than 8 by 8 segments in PCA compression. This suggests that the dimensional reduction is more consistent in the smaller segments. However the tsse is lower for `seen´ segments in the larger segment PCs. Artificial neural networks also seem to generalise less readily on larger segments. Since the time taken for neural network compression is about an order of magnitude higher than PCA, and PCA is more repeatable in terms of the error magnitude, and produces lower error for `unseen´ segments, it would seem preferable to use PCA analysis than neural network methods to produce the reduced dimensional input to a diagnostic network