•Different types of uncertainties for deep-learning based medical image segmentation were analysed.•We propose a general aleatoric uncertainty estimation method based on test-time augmentation.•A ...theoretical formulation of test-time augmentation was proposed.•The proposed method was validated with 2D fetal brain segmentation and 3D brain tumor segmentation tasks.
Despite the state-of-the-art performance for medical image segmentation, deep convolutional neural networks (CNNs) have rarely provided uncertainty estimations regarding their segmentation outputs, e.g., model (epistemic) and image-based (aleatoric) uncertainties. In this work, we analyze these different types of uncertainties for CNN-based 2D and 3D medical image segmentation tasks at both pixel level and structure level. We additionally propose a test-time augmentation-based aleatoric uncertainty to analyze the effect of different transformations of the input image on the segmentation output. Test-time augmentation has been previously used to improve segmentation accuracy, yet not been formulated in a consistent mathematical framework. Hence, we also propose a theoretical formulation of test-time augmentation, where a distribution of the prediction is estimated by Monte Carlo simulation with prior distributions of parameters in an image acquisition model that involves image transformations and noise. We compare and combine our proposed aleatoric uncertainty with model uncertainty. Experiments with segmentation of fetal brains and brain tumors from 2D and 3D Magnetic Resonance Images (MRI) showed that 1) the test-time augmentation-based aleatoric uncertainty provides a better uncertainty estimation than calculating the test-time dropout-based model uncertainty alone and helps to reduce overconfident incorrect predictions, and 2) our test-time augmentation outperforms a single-prediction baseline and dropout-based multiple predictions.
•Open-source Python library for preprocessing, augmentation and sampling of medical images for deep learning.•Support for 2D, 3D and 4D images such as X-ray, histopathology, CT, ultrasound and ...diffusion MRI.•Modular design inspired by the deep learning framework PyTorch.•Focus on reproducibility and traceability to encourage open-science practices.•Compatible with related frameworks for medical image processing with deep learning.
Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.
We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.
Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.
TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.
Data-driven machine learning (ML) has emerged as a promising approach for building accurate and robust statistical models from medical data, which is collected in huge volumes by modern healthcare ...systems. Existing medical data is not fully exploited by ML primarily because it sits in data silos and privacy concerns restrict access to this data. However, without access to sufficient data, ML will be prevented from reaching its full potential and, ultimately, from making the transition from research to clinical practice. This paper considers key factors contributing to this issue, explores how federated learning (FL) may provide a solution for the future of digital health and highlights the challenges and considerations that need to be addressed.
Accurate medical image segmentation is essential for diagnosis and treatment planning of diseases. Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance for automatic ...medical image segmentation. However, they are still challenged by complicated conditions where the segmentation target has large variations of position, shape and scale, and existing CNNs have a poor explainability that limits their application to clinical decisions. In this work, we make extensive use of multiple attentions in a CNN architecture and propose a comprehensive attention-based CNN (CA-Net) for more accurate and explainable medical image segmentation that is aware of the most important spatial positions, channels and scales at the same time. In particular, we first propose a joint spatial attention module to make the network focus more on the foreground region. Then, a novel channel attention module is proposed to adaptively recalibrate channel-wise feature responses and highlight the most relevant feature channels. Also, we propose a scale attention module implicitly emphasizing the most salient feature maps among multiple scales so that the CNN is adaptive to the size of an object. Extensive experiments on skin lesion segmentation from ISIC 2018 and multi-class segmentation of fetal MRI found that our proposed CA-Net significantly improved the average segmentation Dice score from 87.77% to 92.08% for skin lesion, 84.79% to 87.08% for the placenta and 93.20% to 95.88% for the fetal brain respectively compared with U-Net. It reduced the model size to around 15 times smaller with close or even better accuracy compared with state-of-the-art DeepLabv3+. In addition, it has a much higher explainability than existing networks by visualizing the attention weight maps. Our code is available at https://github.com/HiLab-git/CA-Net .
Automatic segmentation of brain tumors from medical images is important for clinical assessment and treatment planning of brain tumors. Recent years have seen an increasing use of convolutional ...neural networks (CNNs) for this task, but most of them use either 2D networks with relatively low memory requirement while ignoring 3D context, or 3D networks exploiting 3D features while with large memory consumption. In addition, existing methods rarely provide uncertainty information associated with the segmentation result. We propose a cascade of CNNs to segment brain tumors with hierarchical subregions from multi-modal Magnetic Resonance images (MRI), and introduce a 2.5D network that is a trade-off between memory consumption, model complexity and receptive field. In addition, we employ test-time augmentation to achieve improved segmentation accuracy, which also provides voxel-wise and structure-wise uncertainty information of the segmentation result. Experiments with BraTS 2017 dataset showed that our cascaded framework with 2.5D CNNs was one of the top performing methods (second-rank) for the BraTS challenge. We also validated our method with BraTS 2018 dataset and found that test-time augmentation improves brain tumor segmentation accuracy and that the resulting uncertainty information can indicate potential mis-segmentations and help to improve segmentation accuracy.
•We describe the process of generating histological sections.•We present artefacts and image processing methods to minimise them.•We survey methods for 3D histology reconstruction.•We highlight ...hybrid approaches and discuss remaining challenges in the field.
Display omitted
Histology permits the observation of otherwise invisible structures of the internal topography of a specimen. Although it enables the investigation of tissues at a cellular level, it is invasive and breaks topology due to cutting. Three-dimensional (3D) reconstruction was thus introduced to overcome the limitations of single-section studies in a dimensional scope. 3D reconstruction finds its roots in embryology, where it enabled the visualisation of spatial relationships of developing systems and organs, and extended to biomedicine, where the observation of individual, stained sections provided only partial understanding of normal and abnormal tissues. However, despite bringing visual awareness, recovering realistic reconstructions is elusive without prior knowledge about the tissue shape.
3D medical imaging made such structural ground truths available. In addition, combining non-invasive imaging with histology unveiled invaluable opportunities to relate macroscopic information to the underlying microscopic properties of tissues through the establishment of spatial correspondences; image registration is one technique that permits the automation of such a process and we describe reconstruction methods that rely on it. It is thereby possible to recover the original topology of histology and lost relationships, gain insight into what affects the signals used to construct medical images (and characterise them), or build high resolution anatomical atlases.
This paper reviews almost three decades of methods for 3D histology reconstruction from serial sections, used in the study of many different types of tissue. We first summarise the process that produces digitised sections from a tissue specimen in order to understand the peculiarity of the data, the associated artefacts and some possible ways to minimise them. We then describe methods for 3D histology reconstruction with and without the help of 3D medical imaging, along with methods of validation and some applications. We finally attempt to identify the trends and challenges that the field is facing, many of which are derived from the cross-disciplinary nature of the problem as it involves the collaboration between physicists, histolopathologists, computer scientists and physicians.
Purpose:
The aim of this study was to evaluate the appropriateness of using computed tomography (CT) to cone-beam CT (CBCT) deformable image registration (DIR) for the application of calculating the ...“dose of the day” received by a head and neck patient.
Methods:
NiftyReg is an open-source registration package implemented in our institution. The affine registration uses a Block Matching-based approach, while the deformable registration is a GPU implementation of the popular B-spline Free Form Deformation algorithm. Two independent tests were performed to assess the suitability of our registrations methodology for “dose of the day” calculations in a deformed CT. A geometric evaluation was performed to assess the ability of the DIR method to map identical structures between the CT and CBCT datasets. Features delineated in the planning CT were warped and compared with features manually drawn on the CBCT. The authors computed the dice similarity coefficient (DSC), distance transformation, and centre of mass distance between features. A dosimetric evaluation was performed to evaluate the clinical significance of the registrations errors in the application proposed and to identify the limitations of the approximations used. Dose calculations for the same intensity-modulated radiation therapy plan on the deformed CT and replan CT were compared. Dose distributions were compared in terms of dose differences (DD), gamma analysis, target coverage, and dose volume histograms (DVHs). Doses calculated in a rigidly aligned CT and directly in an extended CBCT were also evaluated.
Results:
A mean value of 0.850 in DSC was achieved in overlap between manually delineated and warped features, with the distance between surfaces being less than 2 mm on over 90% of the pixels. Deformable registration was clearly superior to rigid registration in mapping identical structures between the two datasets. The dose recalculated in the deformed CT is a good match to the dose calculated on a replan CT. The DD is smaller than 2% of the prescribed dose on 90% of the body's voxels and it passes a 2% and 2 mm gamma-test on over 95% of the voxels. Target coverage similarity was assessed in terms of the 95%-isodose volumes. A mean value of 0.962 was obtained for the DSC, while the distance between surfaces is less than 2 mm in 95.4% of the pixels. The method proposed provided adequate dose estimation, closer to the gold standard than the other two approaches. Differences in DVH curves were mainly due to differences in the OARs definition (manual vs warped) and not due to differences in dose estimation (dose calculated in replan CT vs dose calculated in deformed CT).
Conclusions:
Deforming a planning CT to match a daily CBCT provides the tools needed for the calculation of the “dose of the day” without the need to acquire a new CT. The initial clinical application of our method will be weekly offline calculations of the “dose of the day,” and use this information to inform adaptive radiotherapy (ART). The work here presented is a first step into a full implementation of a “dose-driven” online ART.
Learning temporal patterns from multivariate longitudinal data is challenging especially in cases when data is sporadic, as often seen in, e.g., healthcare applications where the data can suffer from ...irregularity and asynchronicity as the time between consecutive data points can vary across features and samples, hindering the application of existing deep learning models that are constructed for complete, evenly spaced data with fixed sequence lengths. In this article, a novel deep learning-based model is developed for modeling multiple temporal features in sporadic data using an integrated deep learning architecture based on a recurrent neural network (RNN) unit and a continuous-time autoregressive (CAR) model. The proposed model, called CARRNN, uses a generalized discrete-time autoregressive (AR) model that is trainable end-to-end using neural networks modulated by time lags to describe the changes caused by the irregularity and asynchronicity. It is applied to time-series regression and classification tasks for Alzheimer's disease progression modeling, intensive care unit (ICU) mortality rate prediction, human activity recognition, and event-based digit recognition, where the proposed model based on a gated recurrent unit (GRU) in all cases achieves significantly better predictive performance than the state-of-the-art methods using RNNs, GRUs, and long short-term memory (LSTM) networks.
•An open-source platform is implemented based on TensorFlow APIs for deep learning in medical imaging domain.•A modular implementation of the typical medical imaging machine learning pipeline ...facilitates (1) warm starts with established pre-trained networks, (2) adapting existing neural network architectures to new problems, and (3) rapid prototyping of new solutions.•Three deep-learning applications, including segmentation, regression, image generation and representation learning, are presented as concrete examples illustrating the platform’s key features.
Medical image analysis and computer-assisted intervention problems are increasingly being addressed with deep-learning-based solutions. Established deep-learning platforms are flexible but do not provide specific functionality for medical image analysis and adapting them for this domain of application requires substantial implementation effort. Consequently, there has been substantial duplication of effort and incompatible infrastructure developed across many research groups. This work presents the open-source NiftyNet platform for deep learning in medical imaging. The ambition of NiftyNet is to accelerate and simplify the development of these solutions, and to provide a common mechanism for disseminating research outputs for the community to use, adapt and build upon.
The NiftyNet infrastructure provides a modular deep-learning pipeline for a range of medical imaging applications including segmentation, regression, image generation and representation learning applications. Components of the NiftyNet pipeline including data loading, data augmentation, network architectures, loss functions and evaluation metrics are tailored to, and take advantage of, the idiosyncracies of medical image analysis and computer-assisted intervention. NiftyNet is built on the TensorFlow framework and supports features such as TensorBoard visualization of 2D and 3D images and computational graphs by default.
We present three illustrative medical image analysis applications built using NiftyNet infrastructure: (1) segmentation of multiple abdominal organs from computed tomography; (2) image regression to predict computed tomography attenuation maps from brain magnetic resonance images; and (3) generation of simulated ultrasound images for specified anatomical poses.
The NiftyNet infrastructure enables researchers to rapidly develop and distribute deep learning solutions for segmentation, regression, image generation and representation learning applications, or extend the platform to new applications.