-
Uncertainty Driven Bottleneck Attention U-net for Organ at Risk Segmentation
Authors:
Abdullah Nazib,
Riad Hassan,
Zahidul Islam,
Clinton Fookes
Abstract:
Organ at risk (OAR) segmentation in computed tomography (CT) imagery is a difficult task for automated segmentation methods and can be crucial for downstream radiation treatment planning. U-net has become a de-facto standard for medical image segmentation and is frequently used as a common baseline in medical image segmentation tasks. In this paper, we propose a multiple decoder U-net architecture…
▽ More
Organ at risk (OAR) segmentation in computed tomography (CT) imagery is a difficult task for automated segmentation methods and can be crucial for downstream radiation treatment planning. U-net has become a de-facto standard for medical image segmentation and is frequently used as a common baseline in medical image segmentation tasks. In this paper, we propose a multiple decoder U-net architecture and use the segmentation disagreement between the decoders as attention to the bottleneck of the network for segmentation refinement. While feature correlation is considered as attention in most cases, in our case it is the uncertainty from the network used as attention. For accurate segmentation, we also proposed a CT intensity integrated regularization loss. Proposed regularisation helps model understand the intensity distribution of low contrast tissues. We tested our model on two publicly available OAR challenge datasets. We also conducted the ablation on each datasets with the proposed attention module and regularization loss. Experimental results demonstrate a clear accuracy improvement on both datasets.
△ Less
Submitted 26 February, 2024; v1 submitted 19 March, 2023;
originally announced March 2023.
-
A Multiple Decoder CNN for Inverse Consistent 3D Image Registration
Authors:
Abdullah Nazib,
Clinton Fookes,
Olivier Salvado,
Dimitri Perrin
Abstract:
The recent application of deep learning technologies in medical image registration has exponentially decreased the registration time and gradually increased registration accuracy when compared to their traditional counterparts. Most of the learning-based registration approaches considers this task as a one directional problem. As a result, only correspondence from the moving image to the target im…
▽ More
The recent application of deep learning technologies in medical image registration has exponentially decreased the registration time and gradually increased registration accuracy when compared to their traditional counterparts. Most of the learning-based registration approaches considers this task as a one directional problem. As a result, only correspondence from the moving image to the target image is considered. However, in some medical procedures bidirectional registration is required to be performed. Unlike other learning-based registration, we propose a registration framework with inverse consistency. The proposed method simultaneously learns forward transformation and backward transformation in an unsupervised manner. We perform training and testing of the method on the publicly available LPBA40 MRI dataset and demonstrate strong performance than baseline registration methods.
△ Less
Submitted 15 February, 2020;
originally announced February 2020.
-
Dense Deformation Network for High Resolution Tissue Cleared Image Registration
Authors:
Abdullah Nazib,
Clinton Fookes,
Dimitri Perrin
Abstract:
The recent application of deep learning in various areas of medical image analysis has brought excellent performance gains. In particular, technologies based on deep learning in medical image registration can outperform traditional optimisation-based registration algorithms both in registration time and accuracy. However, the U-net based architectures used in most of the image registration framewo…
▽ More
The recent application of deep learning in various areas of medical image analysis has brought excellent performance gains. In particular, technologies based on deep learning in medical image registration can outperform traditional optimisation-based registration algorithms both in registration time and accuracy. However, the U-net based architectures used in most of the image registration frameworks downscale the data, which removes global information and affects the deformation. In this paper, we present a densely connected convolutional architecture for deformable image registration. Our proposed dense network downsizes data only in one stage and have dense connections instead of the skip connections in U-net architecture. The training of the network is unsupervised and does not require ground-truth deformation or any synthetic deformation as a label. The proposed architecture is trained and tested on two different versions of tissue-cleared data, at 10\% and 25\% resolution of the original single-cell-resolution dataset. We demonstrate comparable registration performance to state-of-the-art registration methods and superior performance to the deep-learning based VoxelMorph method in terms of accuracy and increased resolution handling ability. In both resolutions, the proposed DenseDeformation network outperforms VoxelMorph in registration accuracy. Importantly, it can register brains in one minute where conventional methods can take hours at 25\% resolution.
△ Less
Submitted 5 September, 2019; v1 submitted 13 June, 2019;
originally announced June 2019.
-
A Comparative Analysis of Registration Tools: Traditional vs Deep Learning Approach on High Resolution Tissue Cleared Data
Authors:
Abdullah Nazib,
Clinton Fookes,
Dimitri Perrin
Abstract:
Image registration plays an important role in comparing images. It is particularly important in analyzing medical images like CT, MRI, PET, etc. to quantify different biological samples, to monitor disease progression and to fuse different modalities to support better diagnosis. The recent emergence of tissue clearing protocols enable us to take images at cellular level resolution. Image registrat…
▽ More
Image registration plays an important role in comparing images. It is particularly important in analyzing medical images like CT, MRI, PET, etc. to quantify different biological samples, to monitor disease progression and to fuse different modalities to support better diagnosis. The recent emergence of tissue clearing protocols enable us to take images at cellular level resolution. Image registration tools developed for other modalities are currently unable to manage images of such extreme high resolution. The recent popularity of deep learning based methods in the computer vision community justifies a rigorous investigation of deep-learning based methods on tissue cleared images along with their traditional counterparts. In this paper, we investigate and compare the performance of a deep learning based registration method with traditional optimization based methods on samples from tissue-clearing methods. From the comparative results it is found that a deep-learning based method outperforms all traditional registration tools in terms of registration time and has achieved promising registration accuracy.
△ Less
Submitted 18 October, 2018;
originally announced October 2018.
-
Performance of Image Registration Tools on High-Resolution 3D Brain Images
Authors:
Abdullah Nazib,
James Galloway,
Clinton Fookes,
Dimitri Perrin
Abstract:
Recent progress in tissue clearing has allowed for the imaging of entire organs at single-cell resolution. These methods produce very large 3D images (several gigabytes for a whole mouse brain). A necessary step in analysing these images is registration across samples. Existing methods of registration were developed for lower resolution image modalities (e.g. MRI) and it is unclear whether their p…
▽ More
Recent progress in tissue clearing has allowed for the imaging of entire organs at single-cell resolution. These methods produce very large 3D images (several gigabytes for a whole mouse brain). A necessary step in analysing these images is registration across samples. Existing methods of registration were developed for lower resolution image modalities (e.g. MRI) and it is unclear whether their performance and accuracy is satisfactory at this larger scale. In this study, we used data from different mouse brains cleared with the CUBIC protocol to evaluate five freely available image registration tools. We used several performance metrics to assess accuracy, and completion time as a measure of efficiency. The results of this evaluation suggest that the ANTS registration tool provides the best registration accuracy while Elastix has the highest computational efficiency among the methods with an acceptable accuracy. The results also highlight the need to develop new registration methods optimised for these high-resolution 3D images.
△ Less
Submitted 13 July, 2018;
originally announced July 2018.
