Metric Learning for Image Registration

Deep learning based analysis of dynamic video ultrasonography for predicting cervical lymph node metastasis in papillary thyroid carcinoma

BACKGROUND

Cervical lymph node metastasis (CLNM) is the most common form of thyroid cancer metastasis. Accurate preoperative CLNM diagnosis is of more importance in patients with papillary thyroid cancer (PTC). However, there is currently no unified methods to objectively predict CLNM risk from ultrasonography in PTC patients.This study aimed to develop a deep learning (DL) model to help clinicians more accurately determine the existence of CLNM risk in patients with PTC and then assist them with treatment decisions.

METHODS

Ultrasound dynamic videos of 388 patients with 717 thyroid nodules were retrospectively collected from Zhejiang Cancer Hospital between January 2020 and June 2022. Five deep learning (DL) models were investigated to examine its efficacy for predicting CLNM risks and their performances were also compared with those predicted using two-dimensional ultrasound static images.

RESULTS

In the testing dataset (n = 78), the DenseNet121 model trained on ultrasound dynamic videos outperformed the other four DL models as well as the DL model trained using the two-dimensional (2D) static images across all metrics. Specifically, using DenseNet121, the comparison between the 3D model and 2D model for all metrics are shown as below: AUROC: 0.903 versus 0.828, sensitivity: 0.877 versus 0.871, specificity: 0.865 versus 0.659.

CONCLUSIONS

This study demonstrated that the DenseNet121 model has the greatest potential in distinguishing CLNM from non-CLNM in patients with PTC. Dynamic videos also offered more information about the disease states which have proven to be more efficient and robust in identifying CLNM compared to statis images.

A survey on deep learning in medical image registration: New technologies, uncertainty, evaluation metrics, and beyond

Deep learning technologies have dramatically reshaped the field of medical image registration over the past decade. The initial developments, such as regression-based and U-Net-based networks, established the foundation for deep learning in image registration. Subsequent progress has been made in various aspects of deep learning-based registration, including similarity measures, deformation regularizations, network architectures, and uncertainty estimation. These advancements have not only enriched the field of image registration but have also facilitated its application in a wide range of tasks, including atlas construction, multi-atlas segmentation, motion estimation, and 2D-3D registration. In this paper, we present a comprehensive overview of the most recent advancements in deep learning-based image registration. We begin with a concise introduction to the core concepts of deep learning-based image registration. Then, we delve into innovative network architectures, loss functions specific to registration, and methods for estimating registration uncertainty. Additionally, this paper explores appropriate evaluation metrics for assessing the performance of deep learning models in registration tasks. Finally, we highlight the practical applications of these novel techniques in medical imaging and discuss the future prospects of deep learning-based image registration.

Riemannian geometry for efficient analysis of protein dynamics data

An increasingly common viewpoint is that protein dynamics datasets reside in a nonlinear subspace of low conformational energy. Ideal data analysis tools should therefore account for such nonlinear geometry. The Riemannian geometry setting can be suitable for a variety of reasons. First, it comes with a rich mathematical structure to account for a wide range of geometries that can be modeled after an energy landscape. Second, many standard data analysis tools developed for data in Euclidean space can be generalized to Riemannian manifolds. In the context of protein dynamics, a conceptual challenge comes from the lack of guidelines for constructing a smooth Riemannian structure based on an energy landscape. In addition, computational feasibility in computing geodesics and related mappings poses a major challenge. This work considers these challenges. The first part of the paper develops a local approximation technique for computing geodesics and related mappings on Riemannian manifolds in a computationally feasible manner. The second part constructs a smooth manifold and a Riemannian structure that is based on an energy landscape for protein conformations. The resulting Riemannian geometry is tested on several data analysis tasks relevant for protein dynamics data. In particular, the geodesics with given start- and end-points approximately recover corresponding molecular dynamics trajectories for proteins that undergo relatively ordered transitions with medium-sized deformations. The Riemannian protein geometry also gives physically realistic summary statistics and retrieves the underlying dimension even for large-sized deformations within seconds on a laptop.

A Review of Medical Image Registration for Different Modalities

Medical image registration has become pivotal in recent years with the integration of various imaging modalities like X-ray, ultrasound, MRI, and CT scans, enabling comprehensive analysis and diagnosis of biological structures. This paper provides a comprehensive review of registration techniques for medical images, with an in-depth focus on 2D-2D image registration methods. While 3D registration is briefly touched upon, the primary emphasis remains on 2D techniques and their applications. This review covers registration techniques for diverse modalities, including unimodal, multimodal, interpatient, and intra-patient. The paper explores the challenges encountered in medical image registration, including geometric distortion, differences in image properties, outliers, and optimization convergence, and discusses their impact on registration accuracy and reliability. Strategies for addressing these challenges are highlighted, emphasizing the need for continual innovation and refinement of techniques to enhance the accuracy and reliability of medical image registration systems. The paper concludes by emphasizing the importance of accurate medical image registration in improving diagnosis.

Artificial intelligence-based automated segmentation and radiotherapy dose mapping for thoracic normal tissues

Background and purpose

Objective assessment of delivered radiotherapy (RT) to thoracic organs requires fast and accurate deformable dose mapping. The aim of this study was to implement and evaluate an artificial intelligence (AI) deformable image registration (DIR) and organ segmentation-based AI dose mapping (AIDA) applied to the esophagus and the heart.

Materials and methods

AIDA metrics were calculated for 72 locally advanced non-small cell lung cancer patients treated with concurrent chemo-RT to 60 Gy in 2 Gy fractions in an automated pipeline. The pipeline steps were: (i) automated rigid alignment and cropping of planning CT to week 1 and week 2 cone-beam CT (CBCT) field-of-views, (ii) AI segmentation on CBCTs, and (iii) AI-DIR-based dose mapping to compute dose metrics. AIDA dose metrics were compared to the planned dose and manual contour dose mapping (manual DA).

Results

AIDA required ∼2 min/patient. Esophagus and heart segmentations were generated with a mean Dice similarity coefficient (DSC) of 0.80±0.15 and 0.94±0.05, a Hausdorff distance at 95th percentile (HD95) of 3.9±3.4 mm and 14.1±8.3 mm, respectively. AIDA heart dose was significantly lower than the planned heart dose (p = 0.04). Larger dose deviations (>=1Gy) were more frequently observed between AIDA and the planned dose (N = 26) than with manual DA (N = 6).

Conclusions

Rapid estimation of RT dose to thoracic tissues from CBCT is feasible with AIDA. AIDA-derived metrics and segmentations were similar to manual DA, thus motivating the use of AIDA for RT applications.

Hierarchical cumulative network for unsupervised medical image registration

Unsupervised deep learning techniques have gained increasing popularity in deformable medical image registration However, existing methods usually overlook the optimal similarity position between moving and fixed images To tackle this issue, we propose a novel hierarchical cumulative network (HCN), which explicitly considers the optimal similarity position with an effective Bidirectional Asymmetric Registration Module (BARM). The BARM simultaneously learns two asymmetric displacement vector fields (DVFs) to optimally warp both moving images and fixed images to their optimal similar shape along the geodesic path. Furthermore, we incorporate the BARM into a Laplacian pyramid network with hierarchical recursion, in which the moving image at the lowest level of the pyramid is warped successively for aligning to the fixed image at the lowest level of the pyramid to capture multiple DVFs. We then accumulate these DVFs and up-sample them to warp the moving images at higher levels of the pyramid to align to the fixed image of the top level. The entire system is end-to-end and jointly trained in an unsupervised manner. Extensive experiments were conducted on two public 3D Brain MRI datasets to demonstrate that our HCN outperforms both the traditional and state-of-the-art registration methods. To further evaluate the performance of our HCN, we tested it on the validation set of the MICCAI Learn2Reg 2021 challenge. Additionally, a cross-dataset evaluation was conducted to assess the generalization of our HCN. Experimental results showed that our HCN is an effective deformable registration method and achieves excellent generalization performance.

Automated Learning for Deformable Medical Image Registration by Jointly Optimizing Network Architectures and Objective Functions

Deformable image registration plays a critical role in various tasks of medical image analysis. A successful registration algorithm, either derived from conventional energy optimization or deep networks, requires tremendous efforts from computer experts to well design registration energy or to carefully tune network architectures with respect to medical data available for a given registration task/scenario. This paper proposes an automated learning registration algorithm (AutoReg) that cooperatively optimizes both architectures and their corresponding training objectives, enabling non-computer experts to conveniently find off-the-shelf registration algorithms for various registration scenarios. Specifically, we establish a triple-level framework to embrace the searching for both network architectures and objectives with a cooperating optimization. Extensive experiments on multiple volumetric datasets and various registration scenarios demonstrate that AutoReg can automatically learn an optimal deep registration network for given volumes and achieve state-of-the-art performance. The automatically learned network also improves computational efficiency over the mainstream UNet architecture from 0.558 to 0.270 seconds for a volume pair on the same configuration.

SequenceMorph: A Unified Unsupervised Learning Framework for Motion Tracking on Cardiac Image Sequences

Modern medical imaging techniques, such as ultrasound (US) and cardiac magnetic resonance (MR) imaging, have enabled the evaluation of myocardial deformation directly from an image sequence. While many traditional cardiac motion tracking methods have been developed for the automated estimation of the myocardial wall deformation, they are not widely used in clinical diagnosis, due to their lack of accuracy and efficiency. In this paper, we propose a novel deep learning-based fully unsupervised method, SequenceMorph, for in vivo motion tracking in cardiac image sequences. In our method, we introduce the concept of motion decomposition and recomposition. We first estimate the inter-frame (INF) motion field between any two consecutive frames, by a bi-directional generative diffeomorphic registration neural network. Using this result, we then estimate the Lagrangian motion field between the reference frame and any other frame, through a differentiable composition layer. Our framework can be extended to incorporate another registration network, to further reduce the accumulated errors introduced in the INF motion tracking step, and to refine the Lagrangian motion estimation. By utilizing temporal information to perform reasonable estimations of spatio-temporal motion fields, this novel method provides a useful solution for image sequence motion tracking. Our method has been applied to US (echocardiographic) and cardiac MR (untagged and tagged cine) image sequences; the results show that SequenceMorph is significantly superior to conventional motion tracking methods, in terms of the cardiac motion tracking accuracy and inference efficiency.

Deep learning predicts malignancy and metastasis of solid pulmonary nodules from CT scans

In the clinic, it is difficult to distinguish the malignancy and aggressiveness of solid pulmonary nodules (PNs). Incorrect assessments may lead to delayed diagnosis and an increased risk of complications. We developed and validated a deep learning-based model for the prediction of malignancy as well as local or distant metastasis in solid PNs based on CT images of primary lesions during initial diagnosis. In this study, we reviewed the data from multiple patients with solid PNs at our institution from 1 January 2019 to 30 April 2022. The patients were divided into three groups: benign, Ia-stage lung cancer, and T1-stage lung cancer with metastasis. Each cohort was further split into training and testing groups. The deep learning system predicted the malignancy and metastasis status of solid PNs based on CT images, and then we compared the malignancy prediction results among four different levels of clinicians. Experiments confirmed that human-computer collaboration can further enhance diagnostic accuracy. We made a held-out testing set of 134 cases, with 689 cases in total. Our convolutional neural network model reached an area under the ROC (AUC) of 80.37% for malignancy prediction and an AUC of 86.44% for metastasis prediction. In observer studies involving four clinicians, the proposed deep learning method outperformed a junior respiratory clinician and a 5-year respiratory clinician by considerable margins; it was on par with a senior respiratory clinician and was only slightly inferior to a senior radiologist. Our human-computer collaboration experiment showed that by simply adding binary human diagnosis into model prediction probabilities, model AUC scores improved to 81.80-88.70% when combined with three out of four clinicians. In summary, the deep learning method can accurately diagnose the malignancy of solid PNs, improve its performance when collaborating with human experts, predict local or distant metastasis in patients with T1-stage lung cancer, and facilitate the application of precision medicine.

VWI-APP: Vessel wall imaging-dedicated automated processing pipeline for intracranial atherosclerotic plaque quantification

BACKGROUND

Quantitative plaque assessment based on 3D magnetic resonance (MR) vessel wall imaging (VWI) has been shown to provide valuable numerical markers of the burden and risk of intracranial atherosclerotic disease (ICAD). However, plaque quantification is currently time-consuming and observer-dependent due to the demand for heavy manual effort. A VWI-dedicated automated processing pipeline (VWI-APP) is desirable.

PURPOSE

To develop and evaluate a VWI-APP for end-to-end quantitative analysis of intracranial atherosclerotic plaque.

METHODS

We retrospectively enrolled 91 subjects with ICAD (80 for pipeline development, 10 for an end-to-end pipeline evaluation, and 1 for demonstrating longitudinal plaque assessment) who had undergone VWI and MR angiography. In an end-to-end evaluation, diameter stenosis (DS), normalized wall index (NWI), remodeling ratio (RR), plaque wall contrast ratio (CR), and total plaque volume (TPV) were quantified at each culprit lesion using the developed VWI-APP and a computer-aided manual approach by a neuroradiologist, respectively. The time consumed in each quantification approach was recorded. Two-sided paired t-tests and intraclass correlation coefficient (ICC) were used to determine the difference and agreement in each plaque metric between VWI-APP and manual quantification approaches.

RESULTS

There was no significant difference between VWI-APP and manual quantification in each plaque metric. The ICC was 0.890, 0.813, 0.827, 0.891, and 0.991 for DS, NWI, RR, CR, and TPV, respectively, suggesting good to excellent accuracy of the pipeline method in plaque quantification. Quantitative analysis of each culprit lesion on average took 675.7 s using the manual approach but shortened to 238.3 s with the aid of VWI-APP.

CONCLUSIONS

VWI-APP is an accurate and efficient approach to intracranial atherosclerotic plaque quantification. Further clinical assessment of this automated tool is warranted to establish its utility in the risk assessment of ICAD lesions.

Value of deep learning models based on ultrasonic dynamic videos for distinguishing thyroid nodules

Objective

This study was designed to distinguish benign and malignant thyroid nodules by using deep learning(DL) models based on ultrasound dynamic videos.

Methods

Ultrasound dynamic videos of 1018 thyroid nodules were retrospectively collected from 657 patients in Zhejiang Cancer Hospital from January 2020 to December 2020 for the tests with 5 DL models.

Results

In the internal test set, the area under the receiver operating characteristic curve (AUROC) was 0.929(95% CI: 0.888,0.970) for the best-performing model LSTM Two radiologists interpreted the dynamic video with AUROC values of 0.760 (95% CI: 0.653, 0.867) and 0.815 (95% CI: 0.778, 0.853). In the external test set, the best-performing DL model had AUROC values of 0.896(95% CI: 0.847,0.945), and two ultrasound radiologist had AUROC values of 0.754 (95% CI: 0.649,0.850) and 0.833 (95% CI: 0.797,0.869).

Conclusion

This study demonstrates that the DL model based on ultrasound dynamic videos performs better than the ultrasound radiologists in distinguishing thyroid nodules.

Learning Deformable Image Registration From Optimization: Perspective, Modules, Bilevel Training and Beyond

Conventional deformable registration methods aim at solving an optimization model carefully designed on image pairs and their computational costs are exceptionally high. In contrast, recent deep learning-based approaches can provide fast deformation estimation. These heuristic network architectures are fully data-driven and thus lack explicit geometric constraints which are indispensable to generate plausible deformations, e.g., topology-preserving. Moreover, these learning-based approaches typically pose hyper-parameter learning as a black-box problem and require considerable computational and human effort to perform many training runs. To tackle the aforementioned problems, we propose a new learning-based framework to optimize a diffeomorphic model via multi-scale propagation. Specifically, we introduce a generic optimization model to formulate diffeomorphic registration and develop a series of learnable architectures to obtain propagative updating in the coarse-to-fine feature space. Further, we propose a new bilevel self-tuned training strategy, allowing efficient search of task-specific hyper-parameters. This training strategy increases the flexibility to various types of data while reduces computational and human burdens. We conduct two groups of image registration experiments on 3D volume datasets including image-to-atlas registration on brain MRI data and image-to-image registration on liver CT data. Extensive results demonstrate the state-of-the-art performance of the proposed method with diffeomorphic guarantee and extreme efficiency. We also apply our framework to challenging multi-modal image registration, and investigate how our registration to support the down-streaming tasks for medical image analysis including multi-modal fusion and image segmentation.

A voting-based ensemble feature network for semiconductor wafer defect classification

Semiconductor wafer defects severely affect product development. In order to reduce the occurrence of defects, it is necessary to identify why they occur, and it can be inferred by analyzing the patterns of defects. Automatic defect classification (ADC) is used to analyze large amounts of samples. ADC can reduce human resource requirements for defect inspection and improve inspection quality. Although several ADC systems have been developed to identify and classify wafer surfaces, the conventional ML-based ADC methods use numerous image recognition features for defect classification and tend to be costly, inefficient, and time-consuming. Here, an ADC technique based on a deep ensemble feature framework (DEFF) is proposed that classifies different kinds of wafer surface damage automatically. DEFF has an ensemble feature network and the final decision network layer. The feature network learns features using multiple pre-trained convolutional neural network (CNN) models representing wafer defects and the ensemble features are computed by concatenating these features. The decision network layer decides the classification labels using the ensemble features. The classification performance is further enhanced by using a voting-based ensemble learning strategy in combination with the deep ensemble features. We show the efficacy of the proposed strategy using the real-world data from SK Hynix.

SymReg-GAN: Symmetric Image Registration With Generative Adversarial Networks

Symmetric image registration estimates bi-directional spatial transformations between images while enforcing an inverse-consistency. Its capability of eliminating bias introduced inevitably by generic single-directional image registration allows more precise analysis in different interdisciplinary applications of image registration, e.g., computational anatomy and shape analysis. However, most existing symmetric registration techniques especially for multimodal images are limited by low speed from the commonly-used iterative optimization, hardship in exploring inter-modality relations or high labor cost for labeling data. We propose SymReg-GAN to shatter these limits, which is a novel generative adversarial networks (GAN) based approach to symmetric image registration. We formulate symmetric registration of unimodal/multimodal images as a conditional GAN and train it with a semi-supervised strategy. The registration symmetry is realized by introducing a loss for encouraging that the cycle composed of the geometric transformation from one image to another and its reverse should bring an image back. The semi-supervised learning enables both the precious labeled data and large amounts of unlabeled data to be fully exploited. Experimental results from six public brain magnetic resonance imaging (MRI) datasets and 1 our own computed tomography (CT) and MRI dataset demonstrate the superiority of SymReg-GAN to several existing state-of-the-art methods.

Joint synthesis and registration network for deformable MR-CBCT image registration for neurosurgical guidance

Objective.The accuracy of navigation in minimally invasive neurosurgery is often challenged by deep brain deformations (up to 10 mm due to egress of cerebrospinal fluid during neuroendoscopic approach). We propose a deep learning-based deformable registration method to address such deformations between preoperative MR and intraoperative CBCT.Approach.The registration method uses a joint image synthesis and registration network (denoted JSR) to simultaneously synthesize MR and CBCT images to the CT domain and perform CT domain registration using a multi-resolution pyramid. JSR was first trained using a simulated dataset (simulated CBCT and simulated deformations) and then refined on real clinical images via transfer learning. The performance of the multi-resolution JSR was compared to a single-resolution architecture as well as a series of alternative registration methods (symmetric normalization (SyN), VoxelMorph, and image synthesis-based registration methods).Main results.JSR achieved median Dice coefficient (DSC) of 0.69 in deep brain structures and median target registration error (TRE) of 1.94 mm in the simulation dataset, with improvement from single-resolution architecture (median DSC = 0.68 and median TRE = 2.14 mm). Additionally, JSR achieved superior registration compared to alternative methods-e.g. SyN (median DSC = 0.54, median TRE = 2.77 mm), VoxelMorph (median DSC = 0.52, median TRE = 2.66 mm) and provided registration runtime of less than 3 s. Similarly in the clinical dataset, JSR achieved median DSC = 0.72 and median TRE = 2.05 mm.Significance.The multi-resolution JSR network resolved deep brain deformations between MR and CBCT images with performance superior to other state-of-the-art methods. The accuracy and runtime support translation of the method to further clinical studies in high-precision neurosurgery.

Light-Weight Deformable Registration Using Adversarial Learning With Distilling Knowledge

Deformable registration is a crucial step in many medical procedures such as image-guided surgery and radiation therapy. Most recent learning-based methods focus on improving the accuracy by optimizing the non-linear spatial correspondence between the input images. Therefore, these methods are computationally expensive and require modern graphic cards for real-time deployment. In this paper, we introduce a new Light-weight Deformable Registration network that significantly reduces the computational cost while achieving competitive accuracy. In particular, we propose a new adversarial learning with distilling knowledge algorithm that successfully leverages meaningful information from the effective but expensive teacher network to the student network. We design the student network such as it is light-weight and well suitable for deployment on a typical CPU. The extensively experimental results on different public datasets show that our proposed method achieves state-of-the-art accuracy while significantly faster than recent methods. We further show that the use of our adversarial learning algorithm is essential for a time-efficiency deformable registration method. Finally, our source code and trained models are available at https://github.com/aioz-ai/LDR_ALDK.

Diffeomorphic transforms for data augmentation of highly variable shape and texture objects

BACKGROUND AND OBJECTIVE

Training a deep convolutional neural network (CNN) for automatic image classification requires a large database with images of labeled samples. However, in some applications such as biology and medicine only a few experts can correctly categorize each sample. Experts are able to identify small changes in shape and texture which go unnoticed by untrained people, as well as distinguish between objects in the same class that present drastically different shapes and textures. This means that currently available databases are too small and not suitable to train deep learning models from scratch. To deal with this problem, data augmentation techniques are commonly used to increase the dataset size. However, typical data augmentation methods introduce artifacts or apply distortions to the original image, which instead of creating new realistic samples, obtain basic spatial variations of the original ones.

METHODS

We propose a novel data augmentation procedure which generates new realistic samples, by combining two samples that belong to the same class. Although the idea behind the method described in this paper is to mimic the variations that diatoms experience in different stages of their life cycle, it has also been demonstrated in glomeruli and pollen identification problems. This new data augmentation procedure is based on morphing and image registration methods that perform diffeomorphic transformations.

RESULTS

The proposed technique achieves an increase in accuracy over existing techniques of 0.47%, 1.47%, and 0.23% for diatom, glomeruli and pollen problems respectively.

CONCLUSIONS

For the Diatom dataset, the method is able to simulate the shape changes in different diatom life cycle stages, and thus, images generated resemble newly acquired samples with intermediate shapes. In fact, the other methods compared obtained worse results than those which were not using data augmentation. For the Glomeruli dataset, the method is able to add new samples with different shapes and degrees of sclerosis (through different textures). This is the case where our proposed DA method is more beneficial, when objects highly differ in both shape and texture. Finally, for the Pollen dataset, since there are only small variations between samples in a few classes and this dataset has other features such as noise which are likely to benefit other existing DA techniques, the method still shows an improvement of the results.

MDReg-Net: Multi-resolution diffeomorphic image registration using fully convolutional networks with deep self-supervision

We present a diffeomorphic image registration algorithm to learn spatial transformations between pairs of images to be registered using fully convolutional networks (FCNs) under a self-supervised learning setting. Particularly, a deep neural network is trained to estimate diffeomorphic spatial transformations between pairs of images by maximizing an image-wise similarity metric between fixed and warped moving images, similar to those adopted in conventional image registration algorithms. The network is implemented in a multi-resolution image registration framework to optimize and learn spatial transformations at different image resolutions jointly and incrementally with deep self-supervision in order to better handle large deformation between images. A spatial Gaussian smoothing kernel is integrated with the FCNs to yield sufficiently smooth deformation fields for diffeomorphic image registration. The spatial transformations learned at coarser resolutions are utilized to warp the moving image, which is subsequently used as input to the network for learning incremental transformations at finer resolutions. This procedure proceeds recursively to the full image resolution and the accumulated transformations serve as the final transformation to warp the moving image at the finest resolution. Experimental results for registering high-resolution 3D structural brain magnetic resonance (MR) images have demonstrated that image registration networks trained by our method obtain robust, diffeomorphic image registration results within seconds with improved accuracy compared with state-of-the-art image registration algorithms.

Development and Validation of a Novel Radiomics-Based Nomogram With Machine Learning to Preoperatively Predict Histologic Grade in Pancreatic Neuroendocrine Tumors

BACKGROUD

Tumor grade is the determinant of the biological aggressiveness of pancreatic neuroendocrine tumors (PNETs) and the best current tool to help establish individualized therapeutic strategies. A noninvasive way to accurately predict the histology grade of PNETs preoperatively is urgently needed and extremely limited.

METHODS

The models training and the construction of the radiomic signature were carried out separately in three-phase (plain, arterial, and venous) CT. Mann-Whitney U test and least absolute shrinkage and selection operator (LASSO) were applied for feature preselection and radiomic signature construction. SVM-linear models were trained by incorporating the radiomic signature with clinical characteristics. An optimal model was then chosen to build a nomogram.

RESULTS

A total of 139 PNETs (including 83 in the training set and 56 in the independent validation set) were included in the present study. We build a model based on an eight-feature radiomic signature (group 1) to stratify PNET patients into grades 1 and 2/3 groups with an AUC of 0.911 (95% confidence intervals (CI), 0.908-0.914) and 0.837 (95% CI, 0.827-0.847) in the training and validation cohorts, respectively. The nomogram combining the radiomic signature of plain-phase CT with T stage and dilated main pancreatic duct (MPD)/bile duct (BD) (group 2) showed the best performance (training set: AUC = 0.919, 95% CI = 0.916-0.922; validation set: AUC = 0.875, 95% CI = 0.867-0.883).

CONCLUSIONS

Our developed nomogram that integrates radiomic signature with clinical characteristics could be useful in predicting grades 1 and 2/3 PNETs preoperatively with powerful capability.

LDVoxelMorph: A precise loss function and cascaded architecture for unsupervised diffeomorphic large displacement registration

PURPOSE

The traditional learning-based non-rigid registration methods for medical images are trained by an invariant smoothness regularization parameter, which cannot satisfy the registration accuracy and diffeomorphic property simultaneously. The diffeomorphic property reflects the credibility of the registration results.

METHOD

To improve the diffeomorphic property in 3D medical image registration, we propose a diffeomorphic cascaded network based on the compressed loss, named LDVoxelMorph. the proposed network has several constituent U-Nets and is trained with deep supervision, which uses a different spatial smoothness regularization parameter in each constituent U-Nets for training. This cascade-variant smoothness regularization parameter can maintain the diffeomorphic property in top cascades with large displacement and achieves precise registration in bottom cascades. Besides, we develop the compressed loss (CL) as a penalty for the velocity field, which can accurately limit the velocity field that causes the deformation field overlap after the velocity field integration.

RESULTS

In our registration experiments, the dice scores of our method were 0.892±0.040 on liver CT datasets SLIVER³⁷ , 0.848±0.044 on liver CT datasets LiTS³⁸ , 0.689±0.014 on brain MRI datasets LPBA³⁸ , and the number of overlapping voxels of deformation field were 325, 159 and 0, respectively. Ablation study shows the compressed loss improves the diffeomorphic property more effectively than increases.

CONCLUSION

Experiments results show that our method can achieve higher registration accuracy assessed by dice scores and overlapping voxels while maintaining the diffeomorphic property for large deformation. This article is protected by copyright. All rights reserved.

A holistic overview of deep learning approach in medical imaging

Medical images are a rich source of invaluable necessary information used by clinicians. Recent technologies have introduced many advancements for exploiting the most of this information and use it to generate better analysis. Deep learning (DL) techniques have been empowered in medical images analysis using computer-assisted imaging contexts and presenting a lot of solutions and improvements while analyzing these images by radiologists and other specialists. In this paper, we present a survey of DL techniques used for variety of tasks along with the different medical image's modalities to provide critical review of the recent developments in this direction. We have organized our paper to provide significant contribution of deep leaning traits and learn its concepts, which is in turn helpful for non-expert in medical society. Then, we present several applications of deep learning (e.g., segmentation, classification, detection, etc.) which are commonly used for clinical purposes for different anatomical site, and we also present the main key terms for DL attributes like basic architecture, data augmentation, transfer learning, and feature selection methods. Medical images as inputs to deep learning architectures will be the mainstream in the coming years, and novel DL techniques are predicted to be the core of medical images analysis. We conclude our paper by addressing some research challenges and the suggested solutions for them found in literature, and also future promises and directions for further developments.

A review of deep learning-based three-dimensional medical image registration methods

Medical image registration is a vital component of many medical procedures, such as image-guided radiotherapy (IGRT), as it allows for more accurate dose-delivery and better management of side effects. Recently, the successful implementation of deep learning (DL) in various fields has prompted many research groups to apply DL to three-dimensional (3D) medical image registration. Several of these efforts have led to promising results. This review summarized the progress made in DL-based 3D image registration over the past 5 years and identify existing challenges and potential avenues for further research. The collected studies were statistically analyzed based on the region of interest (ROI), image modality, supervision method, and registration evaluation metrics. The studies were classified into three categories: deep iterative registration, supervised registration, and unsupervised registration. The studies are thoroughly reviewed and their unique contributions are highlighted. A summary is presented following a review of each category of study, discussing its advantages, challenges, and trends. Finally, the common challenges for all categories are discussed, and potential future research topics are identified.

Deformation trajectory prediction using a neural network trained on finite element data-application to library of CTVs creation for cervical cancer

Purpose. We propose a neural network for fast prediction of realistic, time-parametrized deformations between pairs of input segmentations. The proposed method was used to generate a library of planning CTVs for cervical cancer radiotherapy.Methods.A 3D convolutional neural network (CNN) was introduced to predict a stationary velocity field given the distance maps of the cervix CTV in empty and full bladder anatomy. Diffeomorphic deformation trajectories between the two states were obtained by time integration. Intermediate deformation states were used to populate a library of cervix CTVs. The network was trained on cervix CTV deformations of 20 patients generated by finite element modeling (FEM). Validation was performed on FEM data of 9 healthy volunteers. Additionally, for these subjects, CTV deformations were observed in a series of repeat MR scans as the bladder filled from empty to full. Predicted and FEM libraries were compared, and benchmarked against the observed deformations. Finally, for an independent test set of 20 patients the predicted libraries were evaluated clinically, and compared to the current method.Results.The median Dice score over the validation subjects between the predicted and FEM libraries was >0.95 throughout the deformation, with a median 90 percentile surface distance of <3 mm. The ability to cover observed CTVs was similar for both the FEM-based and the proposed method, with residual offsets being about twice as large as the difference between the two methods. Clinical evaluation showed improved library properties over the method currently used in clinic.Conclusions.We proposed a CNN trained on FEM deformations, which predicts the deformation trajectory between two input states of the cervix CTV in one forward pass. We applied this to CTV library prediction for cervical cancer. The network is able to mimic FEM deformations well, while being much faster and simpler in use.

ICON: Learning Regular Maps Through Inverse Consistency

Learning maps between data samples is fundamental. Applications range from representation learning, image translation and generative modeling, to the estimation of spatial deformations. Such maps relate feature vectors, or map between feature spaces. Well-behaved maps should be regular, which can be imposed explicitly or may emanate from the data itself. We explore what induces regularity for spatial transformations, e.g., when computing image registrations. Classical optimization-based models compute maps between pairs of samples and rely on an appropriate regularizer for well-posedness. Recent deep learning approaches have attempted to avoid using such regularizers altogether by relying on the sample population instead. We explore if it is possible to obtain spatial regularity using an inverse consistency loss only and elucidate what explains map regularity in such a context. We find that deep networks combined with an inverse consistency loss and randomized off-grid interpolation yield well behaved, approximately diffeomorphic, spatial transformations. Despite the simplicity of this approach, our experiments present compelling evidence, on both synthetic and real data, that regular maps can be obtained without carefully tuned explicit regularizers, while achieving competitive registration performance.

S3Reg: Superfast Spherical Surface Registration Based on Deep Learning

Cortical surface registration is an essential step and prerequisite for surface-based neuroimaging analysis. It aligns cortical surfaces across individuals and time points to establish cross-sectional and longitudinal cortical correspondences to facilitate neuroimaging studies. Though achieving good performance, available methods are either time consuming or not flexible to extend to multiple or high dimensional features. Considering the explosive availability of large-scale and multimodal brain MRI data, fast surface registration methods that can flexibly handle multimodal features are desired. In this study, we develop a Superfast Spherical Surface Registration (S3Reg) framework for the cerebral cortex. Leveraging an end-to-end unsupervised learning strategy, S3Reg offers great flexibility in the choice of input feature sets and output similarity measures for registration, and meanwhile reduces the registration time significantly. Specifically, we exploit the powerful learning capability of spherical Convolutional Neural Network (CNN) to directly learn the deformation fields in spherical space and implement diffeomorphic design with "scaling and squaring" layers to guarantee topology-preserving deformations. To handle the polar-distortion issue, we construct a novel spherical CNN model using three orthogonal Spherical U-Nets. Experiments are performed on two different datasets to align both adult and infant multimodal cortical features. Results demonstrate that our S3Reg shows superior or comparable performance with state-of-the-art methods, while improving the registration time from 1 min to 10 sec.

Experiments of Image Classification Using Dissimilarity Spaces Built with Siamese Networks

Traditionally, classifiers are trained to predict patterns within a feature space. The image classification system presented here trains classifiers to predict patterns within a vector space by combining the dissimilarity spaces generated by a large set of Siamese Neural Networks (SNNs). A set of centroids from the patterns in the training data sets is calculated with supervised k-means clustering. The centroids are used to generate the dissimilarity space via the Siamese networks. The vector space descriptors are extracted by projecting patterns onto the similarity spaces, and SVMs classify an image by its dissimilarity vector. The versatility of the proposed approach in image classification is demonstrated by evaluating the system on different types of images across two domains: two medical data sets and two animal audio data sets with vocalizations represented as images (spectrograms). Results show that the proposed system's performance competes competitively against the best-performing methods in the literature, obtaining state-of-the-art performance on one of the medical data sets, and does so without ad-hoc optimization of the clustering methods on the tested data sets.

Anatomical Data Augmentation via Fluid-based Image Registration

We introduce a fluid-based image augmentation method for medical image analysis. In contrast to existing methods, our framework generates anatomically meaningful images via interpolation from the geodesic subspace underlying given samples. Our approach consists of three steps: 1) given a source image and a set of target images, we construct a geodesic subspace using the Large Deformation Diffeomorphic Metric Mapping (LDDMM) model; 2) we sample transformations from the resulting geodesic subspace; 3) we obtain deformed images and segmentations via interpolation. Experiments on brain (LPBA) and knee (OAI) data illustrate the performance of our approach on two tasks: 1) data augmentation during training and testing for image segmentation; 2) one-shot learning for single atlas image segmentation. We demonstrate that our approach generates anatomically meaningful data and improves performance on these tasks over competing approaches. Code is available at https://github.com/uncbiag/easyreg.

A Deep Network for Joint Registration and Reconstruction of Images with Pathologies

Registration of images with pathologies is challenging due to tissue appearance changes and missing correspondences caused by the pathologies. Moreover, mass effects as observed for brain tumors may displace tissue, creating larger deformations over time than what is observed in a healthy brain. Deep learning models have successfully been applied to image registration to offer dramatic speed up and to use surrogate information (e.g., segmentations) during training. However, existing approaches focus on learning registration models using images from healthy patients. They are therefore not designed for the registration of images with strong pathologies for example in the context of brain tumors, and traumatic brain injuries. In this work, we explore a deep learning approach to register images with brain tumors to an atlas. Our model learns an appearance mapping from images with tumors to the atlas, while simultaneously predicting the transformation to atlas space. Using separate decoders, the network disentangles the tumor mass effect from the reconstruction of quasi-normal images. Results on both synthetic and real brain tumor scans show that our approach outperforms cost function masking for registration to the atlas and that reconstructed quasi-normal images can be used for better longitudinal registrations.

Unsupervised Learning for Spherical Surface Registration

Current spherical surface registration methods achieve good performance on alignment and spatial normalization of cortical surfaces across individuals in neuroimaging analysis. However, they are computationally intensive, since they have to optimize an objective function independently for each pair of surfaces. In this paper, we present a fast learning-based algorithm that makes use of the recent development in spherical Convolutional Neural Networks (CNNs) for spherical cortical surface registration. Given a set of surface pairs without supervised information such as ground truth deformation fields or anatomical landmarks, we formulate the registration as a parametric function and learn its parameters by enforcing the feature similarity between one surface and the other one warped by the estimated deformation field using the function. Then, given a new pair of surfaces, we can quickly infer the spherical deformation field registering one surface to the other one. We model this parametric function using three orthogonal Spherical U-Nets and use spherical transform layers to warp the spherical surfaces, while imposing smoothness constraints on the deformation field. All the layers in the network are well-defined and differentiable, thus the parameters can be effectively learned. We show that our method achieves accurate cortical alignment results on 102 subjects, comparable to two state-of-the-art methods: Spherical Demons and MSM, while runs much faster.

Region-specific Diffeomorphic Metric Mapping

We introduce a region-specific diffeomorphic metric mapping (RDMM) registration approach. RDMM is non-parametric, estimating spatio-temporal velocity fields which parameterize the sought-for spatial transformation. Regularization of these velocity fields is necessary. In contrast to existing non-parametric registration approaches using a fixed spatially-invariant regularization, for example, the large displacement diffeomorphic metric mapping (LDDMM) model, our approach allows for spatially-varying regularization which is advected via the estimated spatio-temporal velocity field. Hence, not only can our model capture large displacements, it does so with a spatio-temporal regularizer that keeps track of how regions deform, which is a more natural mathematical formulation. We explore a family of RDMM registration approaches: 1) a registration model where regions with separate regularizations are pre-defined (e.g., in an atlas space or for distinct foreground and background regions), 2) a registration model where a general spatially-varying regularizer is estimated, and 3) a registration model where the spatially-varying regularizer is obtained via an end-to-end trained deep learning (DL) model. We provide a variational derivation of RDMM, showing that the model can assure diffeomorphic transformations in the continuum, and that LDDMM is a particular instance of RDMM. To evaluate RDMM performance we experiment 1) on synthetic 2D data and 2) on two 3D datasets: knee magnetic resonance images (MRIs) of the Osteoarthritis Initiative (OAI) and computed tomography images (CT) of the lung. Results show that our framework achieves comparable performance to state-of-the-art image registration approaches, while providing additional information via a learned spatio-temporal regularizer. Further, our deep learning approach allows for very fast RDMM and LDDMM estimations. Code is available at https://github.com/uncbiag/registration.