arXiv subjects

Dorit Merhof

Publications and source records attributed to Dorit Merhof.

89 records · Page 5Linked to original sources

arXiv2021

An Asymmetric Cycle-Consistency Loss for Dealing with Many-to-One Mappings in Image Translation: A Study on Thigh MR Scans

Generative adversarial networks using a cycle-consistency loss facilitate unpaired training of image-translation models and thereby exhibit a very high potential in manifold medical applications. However, the fact that images in one domain potentially map to more than one image in another domain (e.g. in case of pathological changes) exhibits a major challenge for training the networks. In this work, we offer a solution to improve the training process in case of many-to-one mappings by modifying the cycle-consistency loss. We show formally and empirically that the proposed method improves the performance significantly without radically changing the architecture and without increasing the overall complexity. We evaluate our method on thigh MRI scans with the final goal of segmenting the muscle in fat-infiltrated patients' data.

eess.IV↗

arXiv2020

Modeling the Distribution of Normal Data in Pre-Trained Deep Features for Anomaly Detection

Anomaly Detection (AD) in images is a fundamental computer vision problem and refers to identifying images and 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 AUROC value of $95.8 \pm 1.2$ (mean $\pm$ 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 sub-par 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 available at https://github.com/ORippler/gaussian-ad-mvtec

cs.CV↗

arXiv2020

Improving Pixel Embedding Learning through Intermediate Distance Regression Supervision for Instance Segmentation

As a proposal-free approach, instance segmentation through pixel embedding learning and clustering is gaining more emphasis. Compared with bounding box refinement approaches, such as Mask R-CNN, it has potential advantages in handling complex shapes and dense objects. In this work, we propose a simple, yet highly effective, architecture for object-aware embedding learning. A distance regression module is incorporated into our architecture to generate seeds for fast clustering. At the same time, we show that the features learned by the distance regression module are able to promote the accuracy of learned object-aware embeddings significantly. By simply concatenating features of the distance regression module to the images as inputs of the embedding module, the mSBD scores on the CVPPP Leaf Segmentation Challenge can be further improved by more than 8% compared to the identical set-up without concatenation, yielding the best overall result amongst the leaderboard at CodaLab.

cs.CV↗

arXiv2020

AutoML Segmentation for 3D Medical Image Data: Contribution to the MSD Challenge 2018

Fueled by recent advances in machine learning, there has been tremendous progress in the field of semantic segmentation for the medical image computing community. However, developed algorithms are often optimized and validated by hand based on one task only. In combination with small datasets, interpreting the generalizability of the results is often difficult. The Medical Segmentation Decathlon challenge addresses this problem, and aims to facilitate development of generalizable 3D semantic segmentation algorithms that require no manual parametrization. Such an algorithm was developed and is presented in this paper. It consists of a 3D convolutional neural network with encoder-decoder architecture employing residual-connections, skip-connections and multi-level generation of predictions. It works on anisotropic voxel-geometries and has anisotropic depth, i.e., the number of downsampling steps is a task-specific parameter. These depths are automatically inferred for each task prior to training. By combining this flexible architecture with on-the-fly data augmentation and little-to-no pre-- or postprocessing, promising results could be achieved. The code developed for this challenge will be available online after the final deadline at: https://github.com/ORippler/MSD_2018

eess.IV↗

arXiv2020

Instance Segmentation of Biomedical Images with an Object-aware Embedding Learned with Local Constraints

Automatic instance segmentation is a problem that occurs in many biomedical applications. State-of-the-art approaches either perform semantic segmentation or refine object bounding boxes obtained from detection methods. Both suffer from crowded objects to varying degrees, merging adjacent objects or suppressing a valid object. In this work, we assign an embedding vector to each pixel through a deep neural network. The network is trained to output embedding vectors of similar directions for pixels from the same object, while adjacent objects are orthogonal in the embedding space, which effectively avoids the fusion of objects in a crowd. Our method yields state-of-the-art results even with a light-weighted backbone network on a cell segmentation (BBBC006 + DSB2018) and a leaf segmentation data set (CVPPP2017). The code and model weights are public available.

cs.CV↗

arXiv2020

MixNet: Multi-modality Mix Network for Brain Segmentation

Automated brain structure segmentation is important to many clinical quantitative analysis and diagnoses. In this work, we introduce MixNet, a 2D semantic-wise deep convolutional neural network to segment brain structure in multi-modality MRI images. The network is composed of our modified deep residual learning units. In the unit, we replace the traditional convolution layer with the dilated convolutional layer, which avoids the use of pooling layers and deconvolutional layers, reducing the number of network parameters. Final predictions are made by aggregating information from multiple scales and modalities. A pyramid pooling module is used to capture spatial information of the anatomical structures at the output end. In addition, we test three architectures (MixNetv1, MixNetv2 and MixNetv3) which fuse the modalities differently to see the effect on the results. Our network achieves the state-of-the-art performance. MixNetv2 was submitted to the MRBrainS challenge at MICCAI 2018 and won the 3rd place in the 3-label task. On the MRBrainS2018 dataset, which includes subjects with a variety of pathologies, the overall DSC (Dice Coefficient) of 84.7% (gray matter), 87.3% (white matter) and 83.4% (cerebrospinal fluid) were obtained with only 7 subjects as training data.

eess.IV↗

arXiv2020

Radiomic Feature Stability Analysis based on Probabilistic Segmentations

Identifying image features that are robust with respect to segmentation variability and domain shift is a tough challenge in radiomics. So far, this problem has mainly been tackled in test-retest analyses. In this work we analyze radiomics feature stability based on probabilistic segmentations. Based on a public lung cancer dataset, we generate an arbitrary number of plausible segmentations using a Probabilistic U-Net. From these segmentations, we extract a high number of plausible feature vectors for each lung tumor and analyze feature variance with respect to the segmentations. Our results suggest that there are groups of radiomic features that are more (e.g. statistics features) and less (e.g. gray-level size zone matrix features) robust against segmentation variability. Finally, we demonstrate that segmentation variance impacts the performance of a prognostic lung cancer survival model and propose a new and potentially more robust radiomics feature selection workflow.

eess.IV↗

arXiv2019

Multi Scale Curriculum CNN for Context-Aware Breast MRI Malignancy Classification

Classification of malignancy for breast cancer and other cancer types is usually tackled as an object detection problem: Individual lesions are first localized and then classified with respect to malignancy. However, the drawback of this approach is that abstract features incorporating several lesions and areas that are not labelled as a lesion but contain global medically relevant information are thus disregarded: especially for dynamic contrast-enhanced breast MRI, criteria such as background parenchymal enhancement and location within the breast are important for diagnosis and cannot be captured by object detection approaches properly. In this work, we propose a 3D CNN and a multi scale curriculum learning strategy to classify malignancy globally based on an MRI of the whole breast. Thus, the global context of the whole breast rather than individual lesions is taken into account. Our proposed approach does not rely on lesion segmentations, which renders the annotation of training data much more effective than in current object detection approaches. Achieving an AUROC of 0.89, we compare the performance of our approach to Mask R-CNN and Retina U-Net as well as a radiologist. Our performance is on par with approaches that, in contrast to our method, rely on pixelwise segmentations of lesions.

eess.IV↗

arXiv2019

Super-realtime facial landmark detection and shape fitting by deep regression of shape model parameters

We present a method for highly efficient landmark detection that combines deep convolutional neural networks with well established model-based fitting algorithms. Motivated by established model-based fitting methods such as active shapes, we use a PCA of the landmark positions to allow generative modeling of facial landmarks. Instead of computing the model parameters using iterative optimization, the PCA is included in a deep neural network using a novel layer type. The network predicts model parameters in a single forward pass, thereby allowing facial landmark detection at several hundreds of frames per second. Our architecture allows direct end-to-end training of a model-based landmark detection method and shows that deep neural networks can be used to reliably predict model parameters directly without the need for an iterative optimization. The method is evaluated on different datasets for facial landmark detection and medical image segmentation. PyTorch code is freely available at https://github.com/justusschock/shapenet

cs.CV↗

arXiv2018

Image-based Survival Analysis for Lung Cancer Patients using CNNs

Traditional survival models such as the Cox proportional hazards model are typically based on scalar or categorical clinical features. With the advent of increasingly large image datasets, it has become feasible to incorporate quantitative image features into survival prediction. So far, this kind of analysis is mostly based on radiomics features, i.e. a fixed set of features that is mathematically defined a priori. To capture highly abstract information, it is desirable to learn the feature extraction using convolutional neural networks. However, for tomographic medical images, model training is difficult because on the one hand, only few samples of 3D image data fit into one batch at once and on the other hand, survival loss functions are essentially ordering measures that require large batch sizes. In this work, we show that by simplifying survival analysis to median survival classification, convolutional neural networks can be trained with small batch sizes and learn features that predict survival equally well as end-to-end hazard prediction networks. Our approach outperforms the previous state of the art in a publicly available lung cancer dataset.

cs.CV↗

arXiv2018

Spherical Harmonic Residual Network for Diffusion Signal Harmonization

Diffusion imaging is an important method in the field of neuroscience, as it is sensitive to changes within the tissue microstructure of the human brain. However, a major challenge when using MRI to derive quantitative measures is that the use of different scanners, as used in multi-site group studies, introduces measurement variability. This can lead to an increased variance in quantitative metrics, even if the same brain is scanned. Contrary to the assumption that these characteristics are comparable and similar, small changes in these values are observed in many clinical studies, hence harmonization of the signals is essential. In this paper, we present a method that does not require additional preprocessing, such as segmentation or registration, and harmonizes the signal based on a deep learning residual network. For this purpose, a training database is required, which consist of the same subjects, scanned on different scanners. The results show that harmonized signals are significantly more similar to the ground truth signal compared to no harmonization, but also improve in comparison to another deep learning method. The same effect is also demonstrated in commonly used metrics derived from the diffusion MRI signal.

cs.CV↗

arXiv2018

DELIMIT PyTorch - An extension for Deep Learning in Diffusion Imaging

DELIMIT is a framework extension for deep learning in diffusion imaging, which extends the basic framework PyTorch towards spherical signals. Based on several novel layers, deep learning can be applied to spherical diffusion imaging data in a very convenient way. First, two spherical harmonic interpolation layers are added to the extension, which allow to transform the signal from spherical surface space into the spherical harmonic space, and vice versa. In addition, a local spherical convolution layer is introduced that adds the possibility to include gradient neighborhood information within the network. Furthermore, these extensions can also be utilized for the preprocessing of diffusion signals.

cs.LG↗

arXiv2018

Unsupervisedly Training GANs for Segmenting Digital Pathology with Automatically Generated Annotations

Recently, generative adversarial networks exhibited excellent performances in semi-supervised image analysis scenarios. In this paper, we go even further by proposing a fully unsupervised approach for segmentation applications with prior knowledge of the objects' shapes. We propose and investigate different strategies to generate simulated label data and perform image-to-image translation between the image and the label domain using an adversarial model. Specifically, we assess the impact of the annotation model's accuracy as well as the effect of simulating additional low-level image features. For experimental evaluation, we consider the segmentation of the glomeruli, an application scenario from renal pathology. Experiments provide proof of concept and also confirm that the strategy for creating the simulated label data is of particular relevance considering the stability of GAN trainings.

cs.CV↗

arXiv2017

Context-based Normalization of Histological Stains using Deep Convolutional Features

While human observers are able to cope with variations in color and appearance of histological stains, digital pathology algorithms commonly require a well-normalized setting to achieve peak performance, especially when a limited amount of labeled data is available. This work provides a fully automated, end-to-end learning-based setup for normalizing histological stains, which considers the texture context of the tissue. We introduce Feature Aware Normalization, which extends the framework of batch normalization in combination with gating elements from Long Short-Term Memory units for normalization among different spatial regions of interest. By incorporating a pretrained deep neural network as a feature extractor steering a pixelwise processing pipeline, we achieve excellent normalization results and ensure a consistent representation of color and texture. The evaluation comprises a comparison of color histogram deviations, structural similarity and measures the color volume obtained by the different methods.

cs.CV↗

arXiv2017

CNN Cascades for Segmenting Whole Slide Images of the Kidney

Due to the increasing availability of whole slide scanners facilitating digitization of histopathological tissue, there is a strong demand for the development of computer based image analysis systems. In this work, the focus is on the segmentation of the glomeruli constituting a highly relevant structure in renal histopathology, which has not been investigated before in combination with CNNs. We propose two different CNN cascades for segmentation applications with sparse objects. These approaches are applied to the problem of glomerulus segmentation and compared with conventional fully-convolutional networks. Overall, with the best performing cascade approach, single CNNs are outperformed and a pixel-level Dice similarity coefficient of 0.90 is obtained. Combined with qualitative and further object-level analyses the obtained results are assessed as excellent also compared to recent approaches. In conclusion, we can state that especially one of the proposed cascade networks proved to be a highly powerful tool for segmenting the renal glomeruli providing best segmentation accuracies and also keeping the computing time at a low level.

cs.CV↗

arXiv2016

Recycle deep features for better object detection

Aiming at improving the performance of existing detection algorithms developed for different applications, we propose a region regression-based multi-stage class-agnostic detection pipeline, whereby the existing algorithms are employed for providing the initial detection proposals. Better detection is obtained by exploiting the power of deep learning in the region regress scheme while avoiding the requirement on a huge amount of reference data for training deep neural networks. Additionally, a novel network architecture with recycled deep features is proposed, which provides superior regression results compared to the commonly used architectures. As demonstrated on a data set with ~1200 samples of different classes, it is feasible to successfully train a deep neural network in our proposed architecture and use it to obtain the desired detection performance. Since only slight modifications are required to common network architectures and since the deep neural network is trained using the standard hyperparameters, the proposed detection is well accessible and can be easily adopted to a broad variety of detection tasks.

cs.CV↗

arXiv2011

Automated Image Processing for the Analysis of DNA Repair Dynamics

The efficient repair of cellular DNA is essential for the maintenance and inheritance of genomic information. In order to cope with the high frequency of spontaneous and induced DNA damage, a multitude of repair mechanisms have evolved. These are enabled by a wide range of protein factors specifically recognizing different types of lesions and finally restoring the normal DNA sequence. This work focuses on the repair factor XPC (xeroderma pigmentosum complementation group C), which identifies bulky DNA lesions and initiates their removal via the nucleotide excision repair pathway. The binding of XPC to damaged DNA can be visualized in living cells by following the accumulation of a fluorescent XPC fusion at lesions induced by laser microirradiation in a fluorescence microscope. In this work, an automated image processing pipeline is presented which allows to identify and quantify the accumulation reaction without any user interaction. The image processing pipeline comprises a preprocessing stage where the image stack data is filtered and the nucleus of interest is segmented. Afterwards, the images are registered to each other in order to account for movements of the cell, and then a bounding box enclosing the XPC-specific signal is automatically determined. Finally, the time-dependent relocation of XPC is evaluated by analyzing the intensity change within this box. Comparison of the automated processing results with the manual evaluation yields qualitatively similar results. However, the automated analysis provides more accurate, reproducible data with smaller standard errors. The image processing pipeline presented in this work allows for an efficient analysis of large amounts of experimental data with no user interaction required.

cs.CV↗