A lightweight histopathologic dataset for image segmentation

Despite the limited quantity of gigapixel WSIs available, each contains numerous lesion areas of interest for segmentation. This paradox highlights the rich, yet under-exploited, potential of existing WSI resources for detailed pathological analysis. The method of standardised sliding window simplifies the otherwise cumbersome preprocessing required for each study, significantly reducing the barrier to entry for high-resolution medical image analysis. After a rigorous downsampling and threshold-based region selection operation followed by chromosomal regularisation, our proposed lightweight dataset addresses the problems of tedious preprocessing and optimisation of annotation consistency as well as colour space uniformity. The CAMELYON 16, 17 and ICIAR 2018 datasets provided by the competition organizers are considered as the benchmark, and the Automated Slide Analysis Platform (ASAP) is used for ultra-high resolution image display and manual annotation.