Telepathology is defined as the practice of pathology at a distance using video imaging and telecommunications. In this thesis we address the two main technology challenges in implementing telepathology, viz. compression and transmission of digital pathology images. One of the barriers to telepathology is the availability and the affordability of high bandwidth communication resources. High bandwidth links are required because of the large size of the uncompressed digital pathology images. For efficient utilization of available bandwidth, these images need to be compressed. However aggressive image compression may introduce objectionable artifacts and result in an inaccurate diagnosis. This discussion helps us to identify two main design challenges in implementing telepathology, 1. Compression: There is a need to develop or select an appropriate image compression algorithm and an image quality criterion to ensure maximum possible image compression, while ensuring that diagnostic accuracy is not compromised. 2. Transmission: There is a need to develop or select a smart image transmission scheme which can facilitate the transmission of the compressed image to the remote pathologist without violating the specified bandwidth and delay constraints. We addressed the image compression problem by conducting subjective tests to determine the maximum compression that can be tolerated before the pathology images lose their diagnostic value. We concluded that the diagnostically lossless compression ratio is at least around 5 to 10 times higher than the mathematically lossless compression ratio, which is only about 2:1. We also set up subjective tests to compare the performance of the JPEG and the JPEG 2000 compression algorithms which are commonly used for compression of medical images. We concluded that JPEG 2000 outperforms JPEG at lower bitrates (bits/pixel), but both the algorithms perform equally well at higher bitrates. We also addressed the issue of image transmission for telepathology by proposing a two-stage transmission scheme, where coarse image information compressed at diagnostically lossless level is sent to the clients at the first stage, and the Region of Interest is transmitted at mathematically lossless compression levels at the second stage, thereby reducing the total image transmission delay.

Semantic segmentation as an approach to recognizing and localizing objects within an image is a major research area in computer vision. Now that convolutional neural networks are being increasingly used for such tasks, there have been many improve- ments in grand challenge results, and many new research opportunities in previously untennable areas. Using fully convolutional networks, we have developed a semantic segmentation pipeline for the identification of melanocytic tumor regions, epidermis, and dermis lay- ers in whole slide microscopy images of cutaneous melanoma or cutaneous metastatic melanoma. This pipeline includes processes for annotating and preparing a dataset from the output of a tissue slide scanner to the patch-based training and inference by an artificial neural network. We have curated a large dataset of 50 whole slide images containing cutaneous melanoma or cutaneous metastatic melanoma that are fully annotated at 40× ob- jective resolution by an expert pathologist. We will publish the source images of this dataset online. We also present two new FCN architectures that fuse multiple deconvolutional strides, combining coarse and fine predictions to improve accuracy over similar networks without multi-stride information. Our results show that the system performs better than our comparators. We include inference results on thousands of patches from four whole slide images, reassembling them into whole slide segmentation masks to demonstrate how our system generalizes on novel cases.