Correcting spatial distortion in histological images.

We described an interactive method for correcting spatial distortion in histology samples, applied them to a large set of image data, and quantitatively evaluated the quality of the corrections. We demonstrated registration of histology samples to photographs of macroscopic tissue samples and to MR images. We first described methods for obtaining corresponding fiducial and anatomical points, including a new technique for determining boundary correspondence points. We then describe experimental methods for tissue preparation, including a technique for adding color-coded internal and boundary ink marks that are used to validate the method by measuring the registration error. We applied four different transformations with internal and boundary correspondence points, and measured the distance error between other internal ink fiducials. A large number of boundary points, typically 20-30, and at least two internal points were required for accurate warping registration. Interior errors with the transformation methods were ordered: thin plate spline (TPS) approximately non-warping<<triangle warping<polynomial warping. Although non-warping surprisingly gave the lowest interior distance error (0.5+/-0.3mm), TPS was more robust, gave an insignificantly greater error (0.6+/-0.3mm) and much better results near boundaries where distortion was more evident, and allowed us to correct torn histology samples, a common problem. Using the method to evaluate RF thermal ablation, we found good zonal correlation between MR images and corrected histology samples. The method can be practically applied to this and other emerging applications such as in vivo molecular imaging.