For images compressed with CRs of 10 K and 11 K, model performance decreased (average 0.79 in AUROC and 0.49 in AUPRC). Models trained on images with maximum CRs of 5 K had an average area under the receiver operating characteristic curve (AUROC) of 0.87 and area under the precision-recall curve (AUPRC) of 0.75 across the five folds and compression ratios. Convolutional neural networks (CNNs) with three convolutional layers and three fully-connected layers were trained using these images to classify a mammogram as malignant or not malignant across a range of CRs using five-fold cross-validation. Image compression was applied to mammograms with compression ratios (CRs) ranging from 15 to 11 K. In this retrospective study, 9111 unique mammograms–5672 normal, 1686 benign, and 1754 malignant cases were collected from the National Cancer Center in the Republic of Korea. This study aimed to analyze the impact of image compression on the performance of deep learning-based models in classifying mammograms as “malignant”-cases that lead to a cancer diagnosis and treatment-or “normal” and “benign,” non-malignant cases that do not require immediate medical intervention. These methods reduce file size by using a compact representation of the original image. More information can be obtained from the JPEG XS website.Image compression is used in several clinical organizations to help address the overhead associated with medical imaging. UCL-PiLAB and its spin-off intoPIX have been leading this standardization project since its beginning and the new International Standard is expected by the end of 2018. Hence, very efficient implementations can be made on all usual platforms such as CPU, GPU, ASIC, and FPGAs. However, it has been specifically designed to meet the requirements of live production, broadcast and digital cinema workflows, Pro-AV markets and Virtual Reality (VR) gaming.Ĭompared to other image compression standardization efforts, the trade-off between complexity and compression efficiency has driven all technological choices during the standard development.

Basically, everywhere uncompressed video is used today can be a potential use case for JPEG XS. In this context, the JPEG XS standardization project has been initiated to bring an interoperable, lightweight (compression ranging from 2:1 to 6:1), low complexity, low latency and visually lossless solution. Facing this reality, the use of a lightweight compression scheme is very attractive, to smooth the everlasting transition between successive generations of infrastructures and protocols. Capacity of video links and communication channels are growing too but slower and, more importantly, in a much less finer way: extending or replacing a complete infrastructure usually brings a big improvement in capacity but involves a huge investment that needs to be amortized on several years.Īs a consequence, uncompressed storage and live video transmission becomes unaffordable and unmanageable within current systems and infrastructures, while next generation channels are still being tested and not yet available or affordable. Video Bandwidth requirements are growing fast, as video resolutions, frame rates and numbers of streams to manage are constantly increasing. UCL-PiLAB has been deeply involved in this process, by providing the test framework and all objective quality assessment results having lead to the technology selection and improvement. JPEG XS will have applications in areas such as video transport over IP, virtual reality, augmented reality, space imagery, DSLR cameras, and self-driving cars.

With this new format, the image-compression process uses less energy, and higher-quality images can be sent with low latency over broadband networks like 5G.

The Joint Photographic Experts Group ( JPEG), an international committee developing image compression standards within ISO, has just unveiled JPEG XS. A new JPEG lightweight compression standard for video transport and storage

Image compression standards