To What Extent Does Downsampling, Compression, and Data Scarcity Impact Renal Image Analysis?

Abstract

The condition of the Glomeruli, or filter sacks, in renal Direct Immunofluorescence (DIF) specimens is a critical indicator for diagnosing kidney diseases. A digital pathology system which digitizes a glass histology slide into a Whole Slide Image (WSI) and then automatically detects and zooms in on the glomeruli with a higher magnification objective will be extremely helpful for pathologists. In this paper, using glomerulus detection as the study case, we provide analysis and observations on several important issues to help with the development of Computer Aided Diagnostic (CAD) systems to process WSIs. Large image resolution, large file size, and data scarcity are always challenging to deal with. To this end, we first examine image downsampling rates in terms of their effect on detection accuracy. Second, we examine the impact of image compression. Third, we examine the relationship between the size of the training set and detection accuracy. To understand the above issues, experiments are performed on the state-of-the-art detectors: Faster R-CNN, R-FCN, Mask R-CNN and SSD. Critical findings are observed: (1) The best balance between detection accuracy, detection speed and file size is achieved at 8 times downsampling captured with a 40× objective; (2) compression which reduces the file size dramatically, does not necessarily have an adverse effect on overall accuracy;(3) reducing the amount of training data to some extents causes a drop in precision but has a negligible impact on the recall; (4) in most cases, Faster R-CNN achieves the best accuracy in the glomerulus detection task. We show that the image file size of 40× WSI images can be reduced by a factor of over 6000 with negligible loss of glomerulus detection accuracy.

Evaluating and Mitigating Neutrons Effects on COTS EdgeAI Accelerators

“…By down sampling the MNIST images with a mean-box-filtering technique [14], the images were reduced to 256 bytes, making them compatible with the device. Since this approach is used to reduce computational complexity even in powerful ANN devices [15], we claim that this is still an effective test case scenario. The accuracy of the network with the down sampled dataset was approximately 80%.…”

Instance segmentation for whole slide imaging: end-to-end or detect-then-segment

“…In recent years, there has been a paradigm A. Jha, R. Deng, Y. Huo were with the Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN, 37215, USA, email: yuankai.huo@vanderbilt.edu H. Yang, M. Kapp, A. Fogo were with the Department of Pathology, Vanderbilt University Medical Center, Nashville, TN, 37215, USA shift towards automatic glomerular instance segmentation, which aims to provide instance-level, pixel-wise annotation for each glomerulus driven by Convolutional Neural Networks (CNNs) [4], [5]. The de facto standard method of instance segmentation of glomeruli, and more broadly the kidney, is Mask-RCNN [6], [7], an end-to-end pipeline which performs detection and instance segmentation simultaneously [8]. Since the end-to-end architecture of Mask-RCNN is designed for natural images (e.g., ≈ 1000×1000 pixels), both downsampling and tiling are utilized in order to leverage processing speeds and fit modern GPU memory when Mask-RCNN is applied to high resolution WSI (e.g., > 10000×10000 pixels on 40x).…”

“…However, a loss of information is often associated with the downsampling process that is inherent to the endto-end framework of Mask-RCNN. In particular, a single glomerulus from a WSI can be more than 1,000×1,000 pixels in image resolution, which yields significant information loss when the corresponding features maps are downsampled to the 28×28 resolution via the end-to-end Mask-RCNN segmentation head [6], as demonstrated in Fig. 1.…”