Data-Driven Deep Supervision for Medical Image Segmentation

Mishra, Suraj; Zhang, Yizhe; Chen, Danny Z.; Hu, Xiaobo Sharon

doi:10.1109/tmi.2022.3143371

Cited by 28 publications

(16 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Novel ThresholdNet [10] is designed for automatic segmentation of polyp in endoscopy images such that segmentation threshold is learnable. Proposed LEFM-Net framework is directly applicable to methods proposed in [4]- [8], [10], [19]. In this paper, however, we demonstrated performance improvement of the LEFM-Net framework using well know deep networks: DeepLabv3+, UNet, UNet++ and MA-net.…”

Section: Related Work a Brief Overview Of Some Specialized Deep Neura...mentioning

confidence: 71%

“…Herein, we briefly comment on recently proposed deep networks with task-or imaging modality specific architectures with an emphasis on histopathological image analysis. New deep CNN was proposed in [4] for medical image segmentation. It exploits specific attributes in the input datasets.…”

Section: Related Work a Brief Overview Of Some Specialized Deep Neura...mentioning

confidence: 99%

“…Thus, we need to find monomials (6) using all combinations of q such that d i=1 q i ≤ m, ∀q i ≥ 0. Thus, the number of terms in this setting becomes (4). Optionally, if the LEFM layer is followed by batch normalization layer [35], two parameters per each term (mean and standard deviation) are introduced.…”

Section: A Data Adaptive Feature Expansion Layermentioning

confidence: 99%

See 2 more Smart Citations

LEFM-Nets: Learnable Explicit Feature Map Deep Networks for Segmentation of Histopathological Images of Frozen Sections

Sitnik¹,

Kopriva²

2022

Preprint

View full text Add to dashboard Cite

Accurate segmentation of medical images is essential for diagnosis and treatment of diseases. These problems are solved by highly complex models, such as deep networks (DN), requiring a large amount of labeled data for training. Thereby, many DNs possess task-or imaging modality specific architectures with a decisionmaking process that is often hard to explain and interpret. Here, we propose a framework that embeds existing DNs into a low-dimensional subspace induced by the learnable explicit feature map (LEFM) layer. Compared to the existing DN, the framework adds one hyperparameter and only modestly increase the number of learnable parameters. The method is aimed at, but not limited to, segmentation of lowdimensional medical images, such as color histopathological images of stained frozen sections. Since features in the LEFM layer are polynomial functions of the original features, proposed LEFM-Nets contribute to the interpretability of network decisions. In this work, we combined LEFM with the known networks: DeepLabv3+, UNet, UNet++ and MA-net. New LEFM-Nets are applied to the segmentation of adenocarcinoma of a colon in a liver from images of hematoxylin and eosin (H&E) stained frozen sections. LEFM-Nets are also tested on nuclei segmentation from images of H&E stained frozen sections of ten human organs. On the first problem, LEFM-Nets achieved statistically significant performance improvement in terms of micro balanced accuracy and F 1 score than original networks. When averaged over ten runs, LEFM-MA-net achieved balanced accuracy of 89.36% ± 1.28% compared to 88.02% ± 1.22% by the MA-net. Corresponding results for F 1 score are 84.96% ± 1.14% and 82.75% ± 1.10%. On the second problem, LEFM-Nets achieved only better performance in comparison with the original networks. LEFM-MA-net achieved balanced accuracy of 89.41% ± 0.29% compared to 89.30% ± 0.44% by the original MA-net. Results for F 1 score are 85.35% ± 0.25% and 85.12% ± 0.51%. The source code is available at https://github.com/dsitnik/lefm

show abstract

Section: Related Work a Brief Overview Of Some Specialized Deep Neura...mentioning

confidence: 71%

Section: Related Work a Brief Overview Of Some Specialized Deep Neura...mentioning

confidence: 99%

See 1 more Smart Citation

LEFM-Nets: Learnable Explicit Feature Map Deep Networks for Segmentation of Histopathological Images of Frozen Sections

Sitnik¹,

Kopriva²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…But, where and how to apply deep supervision still remains an active research topic. Recently, Mishra et al [17] studied data-driven deep supervision for medical image segmentation utilizing target object labels (or masks) as guidance for deep supervision. Yet, for image classification problems in general, the key issue of deep supervision location selection in a CNN becomes challenging due to the absence of object-level labels, since in classification problems, only image-level labels are normally available.…”

Section: Introductionmentioning

confidence: 99%

“…Previous work [15,17] exploited the observation that if the receptive field (RF) or field-of-view of a segmentation CNN matches in size the target object size in the input images, then the segmentation accuracy can be improved. Relative to the CNN's RF, smaller objects are lost in the network's sub-sampling operations while robust global features of larger objects are not well captured due to a smaller RF [17]. We extend the observation/technique of deep supervision to tackle various skin lesion classification tasks, proposing data-driven deep supervision.…”

Section: Introductionmentioning

confidence: 99%

Data-Driven Deep Supervision for Skin Lesion Classification

Mishra

Zhang

et al. 2022

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Automatic classification of pigmented, non-pigmented, and depigmented non-melanocytic skin lesions have garnered lots of attention in recent years. However, imaging variations in skin texture, lesion shape, depigmentation contrast, lighting condition, etc. hinder robust feature extraction, affecting classification accuracy. In this paper, we propose a new deep neural network that exploits input data for robust feature extraction. Specifically, we analyze the convoutional network's behavior (field-of-view) to find the location of deep supervision for improved feature extraction. To achieve this, first we perform activation mapping to generate an object mask, highlighting the input regions most critical for classification output generation. Then the network layer whose layerwise effective receptive field matches the approximated object shape in the object mask is selected as our focus for deep supervision. Utilizing different types of convolutional feature extractors and classifiers on three melanoma detection datasets and two vitiligo detection datasets, we verify the effectiveness of our new method.

show abstract