3 results
Now showing1 - 3 of 3
- Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Skin Lesion Classification Using CNNs with Patch-Based Attention and Diagnosis-Guided Loss WeightingOBJECTIVE: This work addresses two key problems of skin lesion classification. The first problem is the effective use of high-resolution images with pretrained standard architectures for image classification. The second problem is the high class imbalance encountered in real-world multi-class datasets. METHODS: To use high-resolution images, we propose a novel patch-based attention architecture that provides global context between small, high-resolution patches. We modify three pretrained architectures and study the performance of patch-based attention. To counter class imbalance problems, we compare oversampling, balanced batch sampling, and class-specific loss weighting. Additionally, we propose a novel diagnosis-guided loss weighting method which takes the method used for ground-truth annotation into account. RESULTS: Our patch-based attention mechanism outperforms previous methods and improves the mean sensitivity by 7%. Class balancing significantly improves the mean sensitivity and we show that our diagnosis-guided loss weighting method improves the mean sensitivity by 3% over normal loss balancing. CONCLUSION: The novel patch-based attention mechanism can be integrated into pretrained architectures and provides global context between local patches while outperforming other patch-based methods. Hence, pretrained architectures can be readily used with high resolution images without downsampling. The new diagnosis-guided loss weighting method outperforms other methods and allows for effective training when facing class imbalance. SIGNIFICANCE: The proposed methods improve automatic skin lesion classification. They can be extended to other clinical applications where high-resolution image data and class imbalance are relevant.Publicationtype: Journal Article172 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: without files , Skin lesion diagnosis using ensembles, unscaled multi-crop evaluation and loss weighting(2018-08-06); ; ; ; In this paper we present the methods of our submission to the ISIC 2018 challenge for skin lesion diagnosis (Task 3). The dataset consists of 10000 images with seven image-level classes to be distinguished by an automated algorithm. We employ an ensemble of convolutional neural networks for this task. In particular, we fine-tune pretrained state-of-the-art deep learning models such as Densenet, SENet and ResNeXt. We identify heavy class imbalance as a key problem for this challenge and consider multiple balancing approaches such as loss weighting and balanced batch sampling. Another important feature of our pipeline is the use of a vast amount of unscaled crops for evaluation. Last, we consider meta learning approaches for the final predictions. Our team placed second at the challenge while being the best approach using only publicly available data.Publicationtype: Preprint76 - Some of the metrics are blocked by yourconsent settings
Item type:Publication, Access status: with files , Analysis of the influence of imaging-related uncertainties on cerebral aneurysm deformation quantification using a no-deformation physical flow phantom(Nature Publishing Group, 2018-07-20); ; ; ; Cardiac-cycle related pulsatile aneurysm motion and deformation is assumed to provide valuable information for assessing cerebral aneurysm rupture risk. Accordingly, numerous studies addressed quantification of cerebral aneurysm wall motion and deformation. Most of them utilized in vivo imaging data, but image-based aneurysm deformation quantification is subject to pronounced uncertainties: unknown ground-truth deformation; image resolution in the order of the expected deformation; direct interplay between contrast agent inflow and image intensity. To analyze the impact of the uncertainties on deformation quantification, a multi-imaging modality ground-truth phantom study is performed. A physical flow phantom was designed that allowed simulating pulsatile flow through a variety of modeled cerebral vascular structures. The phantom was imaged using different modalities [MRI, CT, 3D-RA] and mimicking physiologically realistic flow conditions. Resulting image data was analyzed by an established registration-based approach for automated wall motion quantification. The data reveals severe dependency between contrast media inflow-related image intensity changes and the extent of estimated wall deformation. The study illustrates that imaging-related uncertainties affect the accuracy of cerebral aneurysm deformation quantification, suggesting that in vivo imaging studies have to be accompanied by ground-truth phantom experiments to foster data interpretation and to prove plausibility of the applied image analysis algorithms.Publicationtype: Journal ArticleScopus© Citations 5 354 308