With increasing amount of data coming from functional observations we are able to develop and test methods which enable us to analyze them cross-sectionally and longitudinally. In all such analysis observations with missing data are often removed or not considered in the analysis, as such decreasing its power. Thus being able to impute missing data with error can enable to enhance the power of our analysis. The dataset we are interested in comes from diffusion tensor imaging. The variability in the observations could be decomposed into subject specific, subject-visit specific and temporal components using longitudinal functional principal component analysis. Using the decomposed information were were able to reconstruct entire and parts of the missing tracts. Such reconstruction is especially useful since we can use the reconstructed tracts in our further analysis.