Anomaly Detection (AD) in images is a fundamental computer vision problem and refers to identifying images and/or image substructures that deviate significantly from the norm. Popular AD algorithms ...commonly try to learn a model of normality from scratch using task specific datasets, but are limited to semi-supervised approaches employing mostly normal data due to the inaccessibility of anomalies on a large scale combined with the ambiguous nature of anomaly appearance. We follow an alternative approach and demonstrate that deep feature representations learned by discriminative models on large natural image datasets are well suited to describe normality and detect even subtle anomalies in a transfer learning setting. Our model of normality is established by fitting a multivariate Gaussian (MVG) to deep feature representations of classification networks trained on ImageNet using normal data only. By subsequently applying the Mahalanobis distance as the anomaly score we outperform the current state of the art on the public MVTec AD dataset, achieving an Area Under the Receiver Operating Characteristic curve of 95.8 ± 1.2% (mean ± SEM) over all 15 classes. We further investigate why the learned representations are discriminative to the AD task using Principal Component Analysis. We find that the principal components containing little variance in normal data are the ones crucial for discriminating between normal and anomalous instances. This gives a possible explanation to the often subpar performance of AD approaches trained from scratch using normal data only. By selectively fitting a MVG to these most relevant components only, we are able to further reduce model complexity while retaining AD performance. We also investigate setting the working point by selecting acceptable False Positive Rate thresholds based on the MVG assumption. Code is publicly available at https://github.com/ORippler/gaussian-ad-mvtec.
Identifying image features that are robust with respect to segmentation variability is a tough challenge in radiomics. So far, this problem has mainly been tackled in test-retest analyses. In this ...work we analyse radiomics feature reproducibility in two phases: first with manual segmentations provided by four expert readers and second with probabilistic automated segmentations using a recently developed neural network (PHiseg). We test feature reproducibility on three publicly available datasets of lung, kidney and liver lesions. We find consistent results both over manual and automated segmentations in all three datasets and show that there are subsets of radiomic features which are robust against segmentation variability and other radiomic features which are prone to poor reproducibility under differing segmentations. By providing a detailed analysis of robustness of the most common radiomics features across several datasets, we envision that more reliable and reproducible radiomic models can be built in the future based on this work.
The Medical Segmentation Decathlon Antonelli, Michela; Reinke, Annika; Bakas, Spyridon ...
Nature communications,
07/2022, Volume:
13, Issue:
1
Journal Article
Peer reviewed
Open access
International challenges have become the de facto standard for comparative assessment of image analysis algorithms. Although segmentation is the most widely investigated medical image processing ...task, the various challenges have been organized to focus only on specific clinical tasks. We organized the Medical Segmentation Decathlon (MSD)-a biomedical image analysis challenge, in which algorithms compete in a multitude of both tasks and modalities to investigate the hypothesis that a method capable of performing well on multiple tasks will generalize well to a previously unseen task and potentially outperform a custom-designed solution. MSD results confirmed this hypothesis, moreover, MSD winner continued generalizing well to a wide range of other clinical problems for the next two years. Three main conclusions can be drawn from this study: (1) state-of-the-art image segmentation algorithms generalize well when retrained on unseen tasks; (2) consistent algorithmic performance across multiple tasks is a strong surrogate of algorithmic generalizability; (3) the training of accurate AI segmentation models is now commoditized to scientists that are not versed in AI model training.
Anomaly detection (AD) in images is a fundamental computer vision problem and refers to identifying images that deviate significantly from normality. State-of-the-art AD algorithms commonly learn a ...model of normality from scratch using task-specific datasets in either semisupervised or self-supervised manner. We follow an alternative approach and model the distribution of normal data in deep feature representations learned from ImageNet via a multivariate Gaussian (MVG). This lightweight approach achieves a new state of the art in AD on the public MVTec AD dataset. In addition to the empirical benefits, we give a clear motivation for the seemingly simplistic approach via the ties between deep generative and discriminative modeling revealed recently. We further elucidate why ImageNet representations are discriminative in the transfer learning AD setting using the principal component analysis. Here, we find that the principal components containing little variance in normal data are the ones crucial for discriminating between normal and anomalous instances, giving an explanation for the unreasonable effectiveness of our approach. We also investigate setting the working point of our approach by selecting acceptable false-positive rate thresholds based on the MVG assumption and the resistance of our approach to unlabeled anomalies in the dataset. Finally, we investigate whether our approach is prone to exploiting spurious correlations using explainable AI techniques. Code is publicly available at https://github.com/ORippler/gaussian-ad-mvtec .
Spectral reconstruction from RGB or spectral super-resolution (SSR) offers a cheap alternative to otherwise costly and more complex spectral imaging devices. In recent years, deep learning based ...methods consistently achieved the best reconstruction quality in terms of spectral error metrics. However, there are important properties that are not maintained by deep neural networks. This work is primarily dedicated to scale invariance, also known as brightness invariance or exposure invariance. When RGB signals only differ in their absolute scale, they should lead to identical spectral reconstructions apart from the scaling factor. Scale invariance is an essential property that signal processing must guarantee for a wide range of practical applications. At the moment, scale invariance can only be achieved by relying on a diverse database during network training that covers all possibly occurring signal intensities. In contrast, we propose and evaluate a fundamental approach for deep learning based SSR that holds the property of scale invariance by design and is independent of the training data. The approach is independent of concrete network architectures and instead focuses on reevaluating what neural networks should actually predict. The key insight is that signal magnitudes are irrelevant for acquiring spectral reconstructions from camera signals and are only useful for a potential signal denoising.
Global Plants, a collaborative between JSTOR and some 300 herbaria, now contains about 2.48 million high-resolution images of plant specimens, a number that continues to grow, and collections that ...are digitizing their specimens at high resolution are allocating considerable recourses to the maintenance of computer hardware (e.g., servers) and to acquiring digital storage space. We here apply machine learning, specifically the training of a Support-Vector-Machine, to classify specimen images into categories, ideally at the species level, using the 26 most common tree species in Germany as a test case.
We designed an analysis pipeline and classification system consisting of segmentation, normalization, feature extraction, and classification steps and evaluated the system in two test sets, one with 26 species, the other with 17, in each case using 10 images per species of plants collected between 1820 and 1995, which simulates the empirical situation that most named species are represented in herbaria and databases, such as JSTOR, by few specimens. We achieved 73.21% accuracy of species assignments in the larger test set, and 84.88% in the smaller test set.
The results of this first application of a computer vision algorithm trained on images of herbarium specimens shows that despite the problem of overlapping leaves, leaf-architectural features can be used to categorize specimens to species with good accuracy. Computer vision is poised to play a significant role in future rapid identification at least for frequently collected genera or species in the European flora.
Fabric anomaly detection (AD) tries to detect anomalies (i.e., defects) in fabrics, and fabric AD approaches are continuously improved with respect to their AD performance. However, developed ...solutions are known to generalize poorly to previously unseen fabrics, posing a crucial limitation to their applicability. Moreover, current research focuses on adapting converged models to previously unseen fabrics in a post hoc manner, rather than training models that generalize better in the first place. In our work, we explore this potential for the first time. Specifically, we propose that previously unseen fabrics can be regarded as shifts in the underlying data distribution. We therefore argue that factors which reportedly improve a model’s resistance to distribution shifts should also improve the performance of supervised fabric AD methods on unseen fabrics. Hence, we assess the potential benefits of: (I) vicinal risk minimization (VRM) techniques adapted to the fabric AD use-case, (II) different loss functions, (III) ImageNet pre-training, (IV) dataset diversity, and (V) model architecture as well as model complexity. The subsequently performed large-scale analysis reveals that (I) only the VRM technique, AugMix, consistently improves performance on unseen fabrics; (II) hypersphere classifier outperforms other loss functions when combined with AugMix and (III) ImageNet pre-training, which is already beneficial on its own; (IV) increasing dataset diversity improves performance on unseen fabrics; and (V) architectures with better ImageNet performance also perform better on unseen fabrics, yet the same does not hold for more complex models. Notably, the results show that not all factors and techniques which reportedly improve a model’s resistance to distribution shifts in natural images also improve the generalization of supervised fabric AD methods to unseen fabrics, demonstrating the necessity of our work. Additionally, we also assess whether the performance gains of models which generalize better propagate to post hoc adaptation methods and show this to be the case. Since no suitable fabric dataset was publicly available at the time of this work, we acquired our own fabric dataset, called OLP, as the basis for the above experiments. OLP consists of 38 complex, patterned fabrics, more than 6400 images in total, and is made publicly available.
We present a system that utilizes a range of image processing algorithms to allow fully automated thermal face analysis under both laboratory and real-world conditions. We implement methods for face ...detection, facial landmark detection, face frontalization and analysis, combining all of these into a fully automated workflow. The system is fully modular and allows implementing own additional algorithms for improved performance or specialized tasks. Our suggested pipeline contains a histogtam of oriented gradients support vector machine (HOG-SVM) based face detector and different landmark detecion methods implemented using feature-based active appearance models, deep alignment networks and a deep shape regression network. Face frontalization is achieved by utilizing piecewise affine transformations. For the final analysis, we present an emotion recognition system that utilizes HOG features and a random forest classifier and a respiratory rate analysis module that computes average temperatures from an automatically detected region of interest. Results show that our combined system achieves a performance which is comparable to current stand-alone state-of-the-art methods for thermal face and landmark datection and a classification accuracy of 65.75% for four basic emotions.
Nephropathologic analyses provide important outcomes-related data in experiments with the animal models that are essential for understanding kidney disease pathophysiology. Precision medicine ...increases the demand for quantitative, unbiased, reproducible, and efficient histopathologic analyses, which will require novel high-throughput tools. A deep learning technique, the convolutional neural network, is increasingly applied in pathology because of its high performance in tasks like histology segmentation.
We investigated use of a convolutional neural network architecture for accurate segmentation of periodic acid-Schiff-stained kidney tissue from healthy mice and five murine disease models and from other species used in preclinical research. We trained the convolutional neural network to segment six major renal structures: glomerular tuft, glomerulus including Bowman's capsule, tubules, arteries, arterial lumina, and veins. To achieve high accuracy, we performed a large number of expert-based annotations, 72,722 in total.
Multiclass segmentation performance was very high in all disease models. The convolutional neural network allowed high-throughput and large-scale, quantitative and comparative analyses of various models. In disease models, computational feature extraction revealed interstitial expansion, tubular dilation and atrophy, and glomerular size variability. Validation showed a high correlation of findings with current standard morphometric analysis. The convolutional neural network also showed high performance in other species used in research-including rats, pigs, bears, and marmosets-as well as in humans, providing a translational bridge between preclinical and clinical studies.
We developed a deep learning algorithm for accurate multiclass segmentation of digital whole-slide images of periodic acid-Schiff-stained kidneys from various species and renal disease models. This enables reproducible quantitative histopathologic analyses in preclinical models that also might be applicable to clinical studies.