Virtual Microscopy Large Slide Automated Acquisition: Error Analysis and Validation

The aim of this work is to assess and analyze the discrepancies introduced in the reconstruction of an entire tumoral bone slice from multiple field acquisitions of a large microscopy slide. The reconstruction tends to preserve the original structural information and its error is estimated by comparing the reconstructed images of eight samples against single pictures of these samples. This comparison is held using the Structural Similarity index. The measurements show that smaller samples yield better results. The detected errors are introduced by the insufficiently corrected optical distortion caused by the camera lens, which tends to accumulate along the sample. Nevertheless, the maximum error encountered does not exceed 0.39 mm, which is smaller than the maximum tolerable error for the intended application, stated in 1 mm.