Semi-supervised and Task-Driven Data Augmentation

Abstract

Supervised deep learning methods for segmentation require large amounts of labelled training data, without which they are prone to overfitting, not generalizing well to unseen images. In practice, obtaining a large number of annotations from clinical experts is expensive and time-consuming. One way to address scarcity of annotated examples is data augmentation using random spatial and intensity transformations. Recently, it has been proposed to use generative models to synthesize realistic training examples, complementing the random augmentation. So far, these methods have yielded limited gains over the random augmentation. However, there is potential to improve the approach by (i) explicitly modeling deformation fields (non-affine spatial transformation) and intensity transformations and (ii) leveraging unlabelled data during the generative process. With this motivation, we propose a novel task-driven data augmentation method where to synthesize new training examples, a generative network explicitly models and applies deformation fields and additive intensity masks on existing labelled data, modeling shape and intensity variations, respectively. Crucially, the generative model is optimized to be conducive to the task, in this case segmentation, and constrained to match the distribution of images observed from labelled and unlabelled samples. Furthermore, explicit modeling of deformation fields allow synthesizing segmentation masks and images in exact correspondence by simply applying the generated transformation to an input image and the corresponding annotation. Our experiments on cardiac magnetic resonance images (MRI) showed that, for the task of segmentation in small training data scenarios, the proposed method substantially outperforms conventional augmentation techniques.

Data Augmentation for Brain-Tumor Segmentation: A Review

“…Although there are works which indicate that such aggressive augmentation may deteriorate the performance of the models in brain-tumor delineation , it is still an open issue. Chaitanya et al (2019) showed that visually nonrealistic synthetic examples can improve the segmentation of cardiac MRI and noted that it is slightly counter-intuitive-it may have occurred due to the inherent structural and deformationrelated characteristics of the cardiovascular system. Finally, elastic transformations often benefit from B-splines (Huang and Cohen, 1996;Gu et al, 2014) or random deformations (Castro et al, 2018).…”

Improving Robustness of Deep Learning Based Knee MRI Segmentation: Mixup and Adversarial Domain Adaptation

“…In medical imaging, UDA has been studied in several fields for which multiple diverse datasets are publicly available: brain MRI [20,4], chest X-ray [7], cardiac MRI-CT [13] and others. However, very few studies investigated the use of UDA techniques in knee MRI domain and, more specifically, knee cartilage segmentation.…”

“…Comparison between the baseline and the best performing approaches. Here, means and standard deviations of volumetric DSCs are presented for the subject groups of specific KL-grades(1)(2)(3)(4) and for the full test sets. # shows the number of scans in the specific group.…”

Maschinelles Lernen in der Radiologie

“…Die diesem Konzept zugrundeliegende Idee ist, dass das Aussehen der Objekte in einem Repräsentationsraum weitestgehend kontinuierlich verteilt und die Struktur des Raums durch die initialen Klassen soweit etabliert wird, das neue Klassen eingepasst werden können. Ein anderer Ansatz für das gleiche Problem ist das erlernte Augmentieren von Daten, das heißt, die Generierung von zusätzlichen Trainingsdaten aus wenigen Beobachtungen [5].…”