Flatness of the loss curve around a model at hand has been shown to empirically correlate with its generalization ability. Optimizing for flatness has been proposed as early as 1994 by Hochreiter and Schmidthuber, and was followed by more recent successful sharpness-aware optimization techniques. Their widespread adoption in practice, though, is dubious because of the lack of theoretically grounded connection between flatness and generalization, in particular in light of the reparameterization curse - certain reparameterizations of a neural network change most flatness measures but do not change generalization. Recent theoretical work suggests that a particular relative flatness measure can be connected to generalization and solves the reparameterization curse. In this paper, we derive a regularizer based on this relative flatness that is easy to compute, fast, efficient, and works with arbitrary loss functions. It requires computing the Hessian only of a single layer of the network, which makes it applicable to large neural networks, and with it avoids an expensive mapping of the loss surface in the vicinity of the model. In an extensive empirical evaluation we show that this relative flatness aware minimization (FAM) improves generalization in a multitude of applications and models, both in finetuning and standard training. We make the code available at github.
International benchmarking competitions have become fundamental for the comparative performance assessment of image analysis methods. However, little attention has been given to investigating what can be learnt from these competitions. Do they really generate scientific progress? What are common and successful participation strategies? What makes a solution superior to a competing method? To address this gap in the literature, we performed a multi-center study with all 80 competitions that were conducted in the scope of IEEE ISBI 2021 and MICCAI 2021. Statistical analyses performed based on comprehensive descriptions of the submitted algorithms linked to their rank as well as the underlying participation strategies revealed common characteristics of winning solutions. These typically include the use of multi-task learning (63%) and/or multi-stage pipelines (61%), and a focus on augmentation (100%), image preprocessing (97%), data curation (79%), and postprocessing (66%). The "typical" lead of a winning team is a computer scientist with a doctoral degree, five years of experience in biomedical image analysis, and four years of experience in deep learning. Two core general development strategies stood out for highly-ranked teams: the reflection of the metrics in the method design and the focus on analyzing and handling failure cases. According to the organizers, 43% of the winning algorithms exceeded the state of the art but only 11% completely solved the respective domain problem. The insights of our study could help researchers (1) improve algorithm development strategies when approaching new problems, and (2) focus on open research questions revealed by this work.
The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps.
We present a deep learning-based approach for skull reconstruction for MONAI, which has been pre-trained on the MUG500+ skull dataset. The implementation follows the MONAI contribution guidelines, hence, it can be easily tried out and used, and extended by MONAI users. The primary goal of this paper lies in the investigation of open-sourcing codes and pre-trained deep learning models under the MONAI framework. Nowadays, open-sourcing software, especially (pre-trained) deep learning models, has become increasingly important. Over the years, medical image analysis experienced a tremendous transformation. Over a decade ago, algorithms had to be implemented and optimized with low-level programming languages, like C or C++, to run in a reasonable time on a desktop PC, which was not as powerful as today's computers. Nowadays, users have high-level scripting languages like Python, and frameworks like PyTorch and TensorFlow, along with a sea of public code repositories at hand. As a result, implementations that had thousands of lines of C or C++ code in the past, can now be scripted with a few lines and in addition executed in a fraction of the time. To put this even on a higher level, the Medical Open Network for Artificial Intelligence (MONAI) framework tailors medical imaging research to an even more convenient process, which can boost and push the whole field. The MONAI framework is a freely available, community-supported, open-source and PyTorch-based framework, that also enables to provide research contributions with pre-trained models to others. Codes and pre-trained weights for skull reconstruction are publicly available at: https://github.com/Project-MONAI/research-contributions/tree/master/SkullRec
The reconstruction loss and the Kullback-Leibler divergence (KLD) loss in a variational autoencoder (VAE) often play antagonistic roles, and tuning the weight of the KLD loss in $\beta$-VAE to achieve a balance between the two losses is a tricky and dataset-specific task. As a result, current practices in VAE training often result in a trade-off between the reconstruction fidelity and the continuity$/$disentanglement of the latent space, if the weight $\beta$ is not carefully tuned. In this paper, we present intuitions and a careful analysis of the antagonistic mechanism of the two losses, and propose, based on the insights, a simple yet effective two-stage method for training a VAE. Specifically, the method aggregates a learned Gaussian posterior $z \sim q_{\theta} (z|x)$ with a decoder decoupled from the KLD loss, which is trained to learn a new conditional distribution $p_{\phi} (x|z)$ of the input data $x$. Experimentally, we show that the aggregated VAE maximally satisfies the Gaussian assumption about the latent space, while still achieves a reconstruction error comparable to when the latent space is only loosely regularized by $\mathcal{N}(\mathbf{0},I)$. The proposed approach does not require hyperparameter (i.e., the KLD weight $\beta$) tuning given a specific dataset as required in common VAE training practices. We evaluate the method using a medical dataset intended for 3D skull reconstruction and shape completion, and the results indicate promising generative capabilities of the VAE trained using the proposed method. Besides, through guided manipulation of the latent variables, we establish a connection between existing autoencoder (AE)-based approaches and generative approaches, such as VAE, for the shape completion problem. Codes and pre-trained weights are available at https://github.com/Jianningli/skullVAE
The HoloLens (Microsoft Corp., Redmond, WA), a head-worn, optically see-through augmented reality display, is the main player in the recent boost in medical augmented reality research. In medical settings, the HoloLens enables the physician to obtain immediate insight into patient information, directly overlaid with their view of the clinical scenario, the medical student to gain a better understanding of complex anatomies or procedures, and even the patient to execute therapeutic tasks with improved, immersive guidance. In this systematic review, we provide a comprehensive overview of the usage of the first-generation HoloLens within the medical domain, from its release in March 2016, until the year of 2021, were attention is shifting towards it's successor, the HoloLens 2. We identified 171 relevant publications through a systematic search of the PubMed and Scopus databases. We analyze these publications in regard to their intended use case, technical methodology for registration and tracking, data sources, visualization as well as validation and evaluation. We find that, although the feasibility of using the HoloLens in various medical scenarios has been shown, increased efforts in the areas of precision, reliability, usability, workflow and perception are necessary to establish AR in clinical practice.
Data has become the most valuable resource in today's world. With the massive proliferation of data-driven algorithms, such as deep learning-based approaches, the availability of data is of great interest. In this context, high-quality training, validation and testing datasets are particularly needed. Volumetric data is a very important resource in medicine, as it ranges from disease diagnoses to therapy monitoring. When the dataset is sufficient, models can be trained to help doctors with these tasks. Unfortunately, there are scenarios and applications where large amounts of data is unavailable. For example, in the medical field, rare diseases and privacy issues can lead to restricted data availability. In non-medical fields, the high cost of obtaining a sufficient amount of high-quality data can also be a concern. A solution to these problems can be the generation of synthetic data to perform data augmentation in combination with other more traditional methods of data augmentation. Therefore, most of the publications on 3D Generative Adversarial Networks (GANs) are within the medical domain. The existence of mechanisms to generate realistic synthetic data is a good asset to overcome this challenge, especially in healthcare, as the data must be of good quality and close to reality, i.e. realistic, and without privacy issues. In this review, we provide a summary of works that generate realistic 3D synthetic data using GANs. We therefore outline GAN-based methods in these areas with common architectures, advantages and disadvantages. We present a novel taxonomy, evaluations, challenges and research opportunities to provide a holistic overview of the current state of GANs in medicine and other fields.
Designing implants for large and complex cranial defects is a challenging task, even for professional designers. Current efforts on automating the design process focused mainly on convolutional neural networks (CNN), which have produced state-of-the-art results on reconstructing synthetic defects. However, existing CNN-based methods have been difficult to translate to clinical practice in cranioplasty, as their performance on complex and irregular cranial defects remains unsatisfactory. In this paper, a statistical shape model (SSM) built directly on the segmentation masks of the skulls is presented. We evaluate the SSM on several cranial implant design tasks, and the results show that, while the SSM performs suboptimally on synthetic defects compared to CNN-based approaches, it is capable of reconstructing large and complex defects with only minor manual corrections. The quality of the resulting implants is examined and assured by experienced neurosurgeons. In contrast, CNN-based approaches, even with massive data augmentation, fail or produce less-than-satisfactory implants for these cases. Codes are publicly available at https://github.com/Jianningli/ssm
Objective: Surveillance imaging of chronic aortic diseases, such as dissections, relies on obtaining and comparing cross-sectional diameter measurements at predefined aortic landmarks, over time. Due to a lack of robust tools, the orientation of the cross-sectional planes is defined manually by highly trained operators. We show how manual annotations routinely collected in a clinic can be efficiently used to ease this task, despite the presence of a non-negligible interoperator variability in the measurements. Impact: Ill-posed but repetitive imaging tasks can be eased or automated by leveraging imperfect, retrospective clinical annotations. Methodology: In this work, we combine convolutional neural networks and uncertainty quantification methods to predict the orientation of such cross-sectional planes. We use clinical data randomly processed by 11 operators for training, and test on a smaller set processed by 3 independent operators to assess interoperator variability. Results: Our analysis shows that manual selection of cross-sectional planes is characterized by 95% limits of agreement (LOA) of $10.6^\circ$ and $21.4^\circ$ per angle. Our method showed to decrease static error by $3.57^\circ$ ($40.2$%) and $4.11^\circ$ ($32.8$%) against state of the art and LOA by $5.4^\circ$ ($49.0$%) and $16.0^\circ$ ($74.6$%) against manual processing. Conclusion: This suggests that pre-existing annotations can be an inexpensive resource in clinics to ease ill-posed and repetitive tasks like cross-section extraction for surveillance of aortic dissections.
Medical images, especially volumetric images, are of high resolution and often exceed the capacity of standard desktop GPUs. As a result, most deep learning-based medical image analysis tasks require the input images to be downsampled, often substantially, before these can be fed to a neural network. However, downsampling can lead to a loss of image quality, which is undesirable especially in reconstruction tasks, where the fine geometric details need to be preserved. In this paper, we propose that high-resolution images can be reconstructed in a coarse-to-fine fashion, where a deep learning algorithm is only responsible for generating a coarse representation of the image, which consumes moderate GPU memory. For producing the high-resolution outcome, we propose two novel methods: learned voxel rearrangement of the coarse output and hierarchical image synthesis. Compared to the coarse output, the high-resolution counterpart allows for smooth surface triangulation, which can be 3D-printed in the highest possible quality. Experiments of this paper are carried out on the dataset of AutoImplant 2021 (https://autoimplant2021.grand-challenge.org/), a MICCAI challenge on cranial implant design. The dataset contains high-resolution skulls that can be viewed as 2D manifolds embedded in a 3D space. Codes associated with this study can be accessed at https://github.com/Jianningli/voxel_rearrangement.