MedZip: 3D Medical Images Lossless Compressor Using Recurrent Neural Network (LSTM)

Nagoor, Omniah H.; Whittle, Joss; Deng, Jingjing; Mora, Benjamin; Jones, Mark W.

doi:10.1109/icpr48806.2021.9413341

Cited by 7 publications

(8 citation statements)

References 16 publications

(14 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Would the use of a compact learning-based sequence prediction model practically lead to better generalisability and higher compression performance? Addressed in Chapter 5, Publication Outcome: [44].…”

Section: Practically E Cient and E Ective"mentioning

confidence: 99%

“…Recognition (ICPR) 2020, [44], by the thesis author alongside Dr J. Whittle, Dr J. Deng, Prof. B. Mora, and Prof. M. W. Jones.…”

Section: This Work Was Originally Published In the 25th International...mentioning

confidence: 99%

“…Generally, there is a trade-o between the sequence size and the computational cost of a model. For 3D data, various block coverage around the target voxel can be applied [43,44]. This chapter examines further options for the causal neighbouring sequence and determines the compression trade-o s between size and speed.…”

Section: Lstm Predictive Modelmentioning

confidence: 99%

“…The level of compression that can be achieved depends on the type of patterns and assumption that can be made about the target data such as spatial, coding and spectral (psycho-visual) redundancy [133]. According to current compression research trends, reduction methodologies can be classified into four main types: predictionbased [9,10,13,43,44,[134][135][136][137][138][139][140][141], transform-based [7,142], quantisation-based [143,144], and end-to-end compression frameworks [147]. This chapter will highlight the current research trends of the prediction-based paradigm, including classical (non-learned) lossless compression literature and deep learning compression literature.…”

Section: Related Workmentioning

confidence: 99%

“…Due to hardware or framework limitations, these methods perform encoding and decoding on a single CPU thread to guarantee deterministic computation. In [43,44] only the decoder needs to run in a deterministic environment due to the sequential nature of voxels' dependencies (generating the neighbouring voxels before processing the next target), as shown in Fig. 6.5 (a).…”

Section: Benchmark1: Optimal Input Sequence (Shape and Size)mentioning

confidence: 99%

See 4 more Smart Citations

3D Medical Image Lossless Compressor Using Deep Learning Approaches

Nagoor¹

View full text Add to dashboard Cite

The ever-increasing importance of accelerated information processing, communica-tion, and storing are major requirements within the big-data era revolution. With the extensive rise in data availability, handy information acquisition, and growing data rate, a critical challenge emerges in eﬃcient handling. Even with advanced technical hardware developments and multiple Graphics Processing Units (GPUs) availability, this demand is still highly promoted to utilise these technologies eﬀectively. Health-care systems are one of the domains yielding explosive data growth. Especially when considering their modern scanners abilities, which annually produce higher-resolution and more densely sampled medical images, with increasing requirements for massive storage capacity. The bottleneck in data transmission and storage would essentially be handled with an eﬀective compression method. Since medical information is critical and imposes an inﬂuential role in diagnosis accuracy, it is strongly encouraged to guarantee exact reconstruction with no loss in quality, which is the main objective of any lossless compression algorithm. Given the revolutionary impact of Deep Learning (DL) methods in solving many tasks while achieving the state of the art results, includ-ing data compression, this opens tremendous opportunities for contributions. While considerable eﬀorts have been made to address lossy performance using learning-based approaches, less attention was paid to address lossless compression. This PhD thesis investigates and proposes novel learning-based approaches for compressing 3D medical images losslessly.Firstly, we formulate the lossless compression task as a supervised sequential prediction problem, whereby a model learns a projection function to predict a target voxel given sequence of samples from its spatially surrounding voxels. Using such 3D local sampling information eﬃciently exploits spatial similarities and redundancies in a volumetric medical context by utilising such a prediction paradigm. The proposed NN-based data predictor is trained to minimise the diﬀerences with the original data values while the residual errors are encoded using arithmetic coding to allow lossless reconstruction.Following this, we explore the eﬀectiveness of Recurrent Neural Networks (RNNs) as a 3D predictor for learning the mapping function from the spatial medical domain (16 bit-depths). We analyse Long Short-Term Memory (LSTM) models’ generalisabil-ity and robustness in capturing the 3D spatial dependencies of a voxel’s neighbourhood while utilising samples taken from various scanning settings. We evaluate our proposed MedZip models in compressing unseen Computerized Tomography (CT) and Magnetic Resonance Imaging (MRI) modalities losslessly, compared to other state-of-the-art lossless compression standards.This work investigates input conﬁgurations and sampling schemes for a many-to-one sequence prediction model, speciﬁcally for compressing 3D medical images (16 bit-depths) losslessly. The main objective is to determine the optimal practice for enabling the proposed LSTM model to achieve a high compression ratio and fast encoding-decoding performance. A solution for a non-deterministic environments problem was also proposed, allowing models to run in parallel form without much compression performance drop. Compared to well-known lossless codecs, experimental evaluations were carried out on datasets acquired by diﬀerent hospitals, representing diﬀerent body segments, and have distinct scanning modalities (i.e. CT and MRI).To conclude, we present a novel data-driven sampling scheme utilising weighted gradient scores for training LSTM prediction-based models. The objective is to determine whether some training samples are signiﬁcantly more informative than others, speciﬁcally in medical domains where samples are available on a scale of billions. The eﬀectiveness of models trained on the presented importance sampling scheme was evaluated compared to alternative strategies such as uniform, Gaussian, and sliced-based sampling.

show abstract

Section: Practically E Cient and E Ective"mentioning

confidence: 99%

“…Recognition (ICPR) 2020, [44], by the thesis author alongside Dr J. Whittle, Dr J. Deng, Prof. B. Mora, and Prof. M. W. Jones.…”

Section: This Work Was Originally Published In the 25th International...mentioning

confidence: 99%

Section: Lstm Predictive Modelmentioning

confidence: 99%