Page 95 - The-5th-MCAIT2021-eProceeding

Image Compression in Digital Pathology

a*
c
b
Goh Jee Yuan , Afzan Adam , Zaid Alyasseri
a,b,c Center for Artificial Intelligence Technology, Faculty of Information Science and Technology, UKM 43600 Bangi,
Selangor, Malaysia
* Email:P102192@siswa.ukm.edu.my

Abstract
The rise of digital pathology has brought about great conv enience to both the public and medical practitioners. The
process of diagnosing a patient has now become much more efficient with the aid of digital pathology, which enable the
managing and diagnosing of pathology slides digitally. At the same time, this has also led to a surge in the amount and size
of image data generated, as pathology slide images are high in density and zoomable. As the amount of image data increases
exponentially, it leads to the problem of insufficient storage space and low transfer efficiency across devices, due to the
image complexity and large image size. This research aims to review the latest image compression algorithm, in an attempt
to reduce the size of pathology slide images. These microscopic images can get very large as the images can have very large
dimensions or requires very fine attention to small details. The objective of this research is to investigate the performance
of some common image compression algorithms when tested on high-density pathological image datasets.
Keywords: Digital Pathology; Image Compression; High-density Images

1. Introduction

The digitization of images has brought upon great convenience to our daily lives as images can now be
digitized and stored in devices as files that can be easily and readily accessed. These digital images can also be
easily shared to other devices over the internet to remote places in a matter of minutes. But with the digitization
process comes the need for storage. All files, whether it be images, videos, texts, etc. takes up space to store, in
the form of data bytes (Buchholz 1962). As digital imaging advances in quality, the resolution, colour depth and
detail of such images greatly improves (Tyagi 2018). As such, more bytes are required to preserve the data of
each individual image, which causes an exponential increase in image size. This results in the problem of
insufficient storage on devices and slow transfer speeds across devices, especially for large images due to the
large volume of data to be transferred (Wu 2006).
The aim of this research is to look into image compression as a means to reduce image size for ease of storage
and transfer, while at the same time maintaining a similar visual perception in the resulting image when
compared to the original image. Image compression is a form of data compression that is applied on digital
images to reduce the size of images for storage and transfer. Compression is achieved through the application
of an algorithm on the digital image data that modifies the data in a way that the data retains a similar visual
perception but is reorganized in a more storage cost effective solution hence reducing the size of the digital
image (Rahman 2019).

2. Methodology

2.1. Data acquisition

The focus of this research is to investigate the performance of some common image compression algorithms
when tested on high-density pathological image dataset. The dataset is open source and obtained ethically under

E- Proceedings of The 5th International Multi-Conference on Artificial Intelligence Technology (MCAIT 2021) [82]
Artificial Intelligence in the 4th Industrial Revolution