2D and 3D Segmentation of Uncertain Local Collagen Fiber Orientations in SHG Microscopy

Schmarje, Lars; Zelenka, Claudius; Geisen, Ulf; Glüer, Claus‐C.; Koch, Reinhard

doi:10.1007/978-3-030-33676-9_26

Cited by 15 publications

(17 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While the issue of labels is present in multiple datasets [ 12 , 13 , 14 , 15 , 16 , 17 ], they are not well suited for evaluations. If we want to quantify the performance on labels, we need a dataset with very good ground-truth.…”

Section: Methodsmentioning

confidence: 99%

“…This issue of different annotations is also known as intra- and inter-observer variability [ 11 ] and is common in many biological and medical application fields [ 8 , 9 , 12 , 13 , 14 , 15 , 16 , 17 ]. Even in a curated dataset [ 1 ], we quote Tarling et al who state ”there will very likely be inaccuracies, bias, and even inconsistencies in the labeling which will have affected the training capacity of the model and lead to discrepancies between predictions and ground truths” [ 18 ].…”

Section: Introductionmentioning

confidence: 99%

“…In contrast to us, most methods [ 28 , 29 , 30 ] and literature surveys [ 11 , 26 , 27 ] interpret labels as corrupted labels. We argue that labels are valid signals derived from ambiguous images and that it is important to discover the substructures for real-world data handling [ 12 , 13 , 14 , 15 , 16 , 17 ].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy

Schmarje

Brünger

Santarossa

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

Deep learning has been successfully applied to many classification problems including underwater challenges. However, a long-standing issue with deep learning is the need for large and consistently labeled datasets. Although current approaches in semi-supervised learning can decrease the required amount of annotated data by a factor of 10 or even more, this line of research still uses distinct classes. For underwater classification, and uncurated real-world datasets in general, clean class boundaries can often not be given due to a limited information content in the images and transitional stages of the depicted objects. This leads to different experts having different opinions and thus producing fuzzy labels which could also be considered ambiguous or divergent. We propose a novel framework for handling semi-supervised classifications of such fuzzy labels. It is based on the idea of overclustering to detect substructures in these fuzzy labels. We propose a novel loss to improve the overclustering capability of our framework and show the benefit of overclustering for fuzzy labels. We show that our framework is superior to previous state-of-the-art semi-supervised methods when applied to real-world plankton data with fuzzy labels. Moreover, we acquire 5 to 10% more consistent predictions of substructures.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy

Schmarje

Brünger

Santarossa

et al. 2021

Sensors

Self Cite

View full text Add to dashboard Cite

show abstract

“…A common strategy for dealing with this problem is transfer learning. This strategy improves results even on small and specialized datasets like medical imaging [10]. This might be a practical workaround for some applications but the fundamental issue remains: Unlike humans, supervised learning needs enormous amounts of labeled data.…”

Section: Introductionmentioning

confidence: 99%

A Survey on Semi-, Self- and Unsupervised Learning for Image Classification

et al. 2021

Self Cite

View full text Add to dashboard Cite

While deep learning strategies achieve outstanding results in computer vision tasks, one issue remains: The current strategies rely heavily on a huge amount of labeled data. In many real-world problems, it is not feasible to create such an amount of labeled training data. Therefore, it is common to incorporate unlabeled data into the training process to reach equal results with fewer labels. Due to a lot of concurrent research, it is difficult to keep track of recent developments. In this survey, we provide an overview of often used ideas and methods in image classification with fewer labels. We compare 34 methods in detail based on their performance and their commonly used ideas rather than a fine-grained taxonomy. In our analysis, we identify three major trends that lead to future research opportunities. 1. State-of-the-art methods are scaleable to real-world applications in theory but issues like class imbalance, robustness, or fuzzy labels are not considered. 2. The degree of supervision which is needed to achieve comparable results to the usage of all labels is decreasing and therefore methods need to be extended to settings with a variable number of classes. 3. All methods share some common ideas but we identify clusters of methods that do not share many ideas. We show that combining ideas from different clusters can lead to better performance.

show abstract

“…By learning high-level features of the stacked layers in deep learning models, the models can recognize the essential features from the medical imaging modalities [13], [14]. The convolutional neural network (CNN) has become the most common deep learning model, and various studies have employed CNN for classification and segmentation of the medical images [9], [15], [16]. However, to the best of our knowledge, no study has utilized the extracted features from a CNN model to quantify and characterize the fibrous collagen in cutaneous scar tissue.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning for Analysis of Collagen Fiber Organization in Scar Tissue

et al. 2021

View full text Add to dashboard Cite

The evaluation of the morphology and organization of collagen fibers is critical in understanding wound healing and tissue remodeling after a thermal injury of the skin. However, histological analysis conducted by pathologists is often labor-intensive and limited to qualitative evaluations and scoring within a narrow field of view. In this study, we propose a convolutional neural network (CNN) model to classify Masson's trichrome (MT)-stained histology images of burn-induced scar tissue and to characterize the microstructures of normal tissue and scar tissue in a quantitative manner. The scar tissue is created on in vivo rodent models and prepared for MT-stained histology slides after wound healing. A CNN model is developed, trained, and tested with various sizes of the histology images for classification and characterization. The proposed model classifies both normal tissue (i.e., without burn, as the control) and scar tissue at various scales with over 97% accuracy. The features acquired from the proposed CNN model visually characterizes the density and directional variance of the collagen fibers distributed in the dermal layers from whole histology images. The proposed deep learning technique can provide an objective and reliable method to rapidly assess and quantify wound repair and remodeling.INDEX TERMS Deep learning, histology image, collagen fiber characterization, scar tissue classification.

show abstract

2D and 3D Segmentation of Uncertain Local Collagen Fiber Orientations in SHG Microscopy

Cited by 15 publications

References 24 publications

Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy

Fuzzy Overclustering: Semi-Supervised Classification of Fuzzy Labels with Overclustering and Inverse Cross-Entropy

A Survey on Semi-, Self- and Unsupervised Learning for Image Classification

Deep Learning for Analysis of Collagen Fiber Organization in Scar Tissue

Contact Info

Product

Resources

About