HistoStitcher©: An Interactive Software Package for Reconstructing Digitized Whole Histological Sections from Fragmented Slices
Honorable Mention - Imaging InformaticsJonathan C Chappelow ; Rutgers University;
Content:
We present a software program to accurately reconstruct whole histological sections from digital slides of smaller adjacent tissue slices, such as quadrants of a whole prostate histological section. The program allows interactive identification of anatomical landmarks on adjacent sections, automatic generation of the optimal transform with optional constraints, followed by automatic spatial transformation and combination of the individual fragments. We demonstrate the program in combining two adjacent histological sections of a prostate gland, which was required to be sectioned into quadrants due to limitations in slide imaging hardware.
Technology:
Digital images of quadrant histology sections obtained by digitally scanning tissue slides are first aligned and then reconstructed using a set of image processing routines implemented in the MATLAB© software package.
Design:
The developed graphical user interface (GUI) utilizes an interactive anatomical landmark-based approach to facilitate simple and accurate reconstruction of histology sections from smaller fragments. The user first identifies pairs of corresponding anatomical landmarks that exist along the incision separating two adjacent sections. The program then automatically solves for the optimal linear transform required to bring the landmarks into closest possible alignment. Finally, the images are conjoined via the determined spatial transformation and the resulting larger image is returned. The user may again execute the program with the previous result and another adjacent fragment to obtain a new larger semi section. The program also offers options to constrain the transformation to use isotropic image scaling, no scaling, or no reflection.
Results:
A screen shot of the program being used to combine two histological quadrants sections of the prostate is shown in the figure below. Control points are first interactively selected by clicking directly on the image (red arrows). The reconstructed image is then generated on demand when the user is satisfied with the identified landmarks. An example of a reconstructed image is also shown in the figure below.
Conclusion:
We have presented an image reconstruction program which offers a powerful image alignment engine with flexible spatial transformation options. The program was used for reconstruction of fragments of prostate histological sections into corresponding whole sections and required only the specification of visible landmarks for alignment via mouse clicks from the program operator. The GUI is friendly and intuitive enough to be used by anyone.
