Fully Convolutional Model for Variable Bit Length and Lossy High Density Compression of Mammograms

TL;DR

This paper introduces a fully convolutional autoencoder for high-density lossy compression of mammograms, achieving competitive results with traditional methods at very low bit rates while preserving diagnostic features.

Contribution

It presents a novel deep learning-based autoencoder combined with arithmetic coding for variable bit length mammogram compression, outperforming existing methods at high compression ratios.

Findings

Achieves >300x compression (~0.04 bpp) with comparable quality to JPEG and JPEG2000.

Rivals traditional codecs in radiologist visual assessments at high compression.

Demonstrates domain adaptability across different mammography datasets.

Abstract

Early works on medical image compression date to the 1980's with the impetus on deployment of teleradiology systems for high-resolution digital X-ray detectors. Commercially deployed systems during the period could compress 4,096 x 4,096 sized images at 12 bpp to 2 bpp using lossless arithmetic coding, and over the years JPEG and JPEG2000 were imbibed reaching upto 0.1 bpp. Inspired by the reprise of deep learning based compression for natural images over the last two years, we propose a fully convolutional autoencoder for diagnostically relevant feature preserving lossy compression. This is followed by leveraging arithmetic coding for encapsulating high redundancy of features for further high-density code packing leading to variable bit length. We demonstrate performance on two different publicly available digital mammography datasets using peak signal-to-noise ratio (pSNR), structural similarity (SSIM) index and domain adaptability tests between datasets. At high density compression factors of >300x (~0.04 bpp), our approach rivals JPEG and JPEG2000 as evaluated through a Radiologist's visual Turing test.