Medical Image Segmentation Based on Edge Detection Techniques

Salman, Nassir H.; Ghafour, Bnar M; Hadi, Gullanar M.

doi:10.14738/aivp.32.1006

Cited by 15 publications

(10 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The discontinuity based methods are based on the principle of variation of intensity of pixels in an image. There will be significant changes in intensity levels among neighboring pixels and therefore results in a discontinuity in the image [7].…”

Section: Discontinuity Based Methodsmentioning

confidence: 99%

Energy based Methods for Medical Image Segmentation

Juneja¹,

Kashyap²

2016

IJCA

View full text Add to dashboard Cite

Health care applications have become boon for the healthcare industry. It needs correct segmentation connected with medical images regarding correct diagnosis. An efficient method assures good quality segmentation of medical images. Segmentation methods are classified as edge based, region based, clustering based, Level set methods (LSM) and Energy based methods. In this paper, a survey on all the effective methods those are capable for accurate segmentation is given, however quick process employing correct segments is still difficult. Some existing methods do correction and some badly pertain to deep irregularity in images. The wide range of the problems of computer vision may make good use of image segmentation. This paper studies and evaluate the different methods for segmentation techniques. This study is useful for determining the appropriate use of the image segmentation methods and for improving their accuracy and performance and also works on the main objective, which is designing new algorithms. The main goal is to make the image more simple and meaningful. After a brief description of each method an experimental comparison of some empirical (goodness and discrepancy) methods commonly used is then given to provide a rank of their evaluation abilities. This study is helpful for an appropriate use of existing segmentation methods and for improving their performance as well as for systematically designing new segmentation methods.

show abstract

Section: Discontinuity Based Methodsmentioning

confidence: 99%

Energy based Methods for Medical Image Segmentation

Juneja¹,

Kashyap²

2016

IJCA

View full text Add to dashboard Cite

show abstract

“…To segment computed tomography (CT) and magnetic resonance (MR) medical images, a novel mix of image segmentation techniques, including K-means clustering, watershed transform, region merging, and growth algorithm, was developed by Salman et al , 2015" due to Ref. [16]. Khan et al, 2017" [17], provides an improved K-means method to address the limitations of traditional K-means in the context of image segmentation.…”

Section: Related Workmentioning

confidence: 99%

Region of Interest Extraction using K-Means and Edge Detection for DEXA Images

Z. Mohammed,

George

2023

AJEST

View full text Add to dashboard Cite

Region of interest in the world of medical images, there is a region in each image that contains unique features that distinguish each image from the other or distinguish one group of the image from another group. This paper proposes a new method for extracting the region of interest for DEXA images via the K-means and edge detection. firstly, the noise is reduced by the mean filter then segment the image into two clusters by kmeans, followed by edge detection to identify object boundaries and erosion operation to clarify boundaries and get the correct coordinates of ROI. The results show that the accuracy of the proposed system is 99%. where 174 images cropped correctly out of 176. the dataset used in this work is 'Osteoporosis DEXA Scans Images of Spine from Pakistan'.

show abstract

“…In contrast, our paper investigates a more general problem, where task-specific pre-training is not needed for a new task as we use one robust large-scale pre-trained model trained on ImageNet. [56] considers the robust pre-training on the large-scale ImageNet and its transfer to downstream tasks, but focuses on the performance on clean instead of adversarial images. The Learning-without-Forgetting (LwF) [40] approach for retaining robustness is shown to be effective in the small-scale transfer experiment [58], but is not effective in our experiment setting of transfer of large-scale models.…”

Section: Related Workmentioning

confidence: 99%

“…This problem setting is becoming more important as the standard pretrained models do not learn robust representations from the pre-training data and are substantially weaker than the robust pre-trained counterparts in some challenging downstream tasks, e.g., fine-grained classification as shown in our experiment. Meanwhile, more large-scale robust pretrained models are released (e.g., ResNet [56] and ViT [4]), which makes the robust pre-trained models more accessible. However, naively applying adversarial training to fine-tune from the robustly pre-trained model will lead to suboptimal robustness, since the robust representations learned by the robust pre-trained model are not fully utilized.…”

Section: Introductionmentioning

confidence: 99%

TWINS: A Fine-Tuning Framework for Improved Transferability of Adversarial Robustness and Generalization

Liu¹,

Xu²,

Ji³

et al. 2023

Preprint

View full text Add to dashboard Cite

Recent years have seen the ever-increasing importance of pre-trained models and their downstream training in deep learning research and applications. At the same time, the defense for adversarial examples has been mainly investigated in the context of training from random initialization on simple classification tasks. To better exploit the potential of pre-trained models in adversarial robustness, this paper focuses on the fine-tuning of an adversarially pre-trained model in various classification tasks. Existing research has shown that since the robust pre-trained model has already learned a robust feature extractor, the crucial question is how to maintain the robustness in the pre-trained model when learning the downstream task. We study the modelbased and data-based approaches for this goal and find that the two common approaches cannot achieve the objective of improving both generalization and adversarial robustness. Thus, we propose a novel statistics-based approach, Two-WIng NormliSation (TWINS) fine-tuning framework, which consists of two neural networks where one of them keeps the population means and variances of pre-training data in the batch normalization layers. Besides the robust information transfer, TWINS increases the effective learning rate without hurting the training stability since the relationship between a weight norm and its gradient norm in standard batch normalization layer is broken, resulting in a faster escape from the sub-optimal initialization and alleviating the robust overfitting. Finally, TWINS is shown to be effective on a wide range of image classification datasets in terms of both generalization and robustness. Our code is available at https://github.com/ziquanliu/ CVPR2023-TWINS. (a) CIFAR10 (b) Caltech-256 Figure 1. The performance of fine-tuning robust and non-robust large-scale pre-trained (PT) ResNet50 [27, 56] on CIFAR10 [35] and Caltech-256 [23]. We compare standard adversarial training (AT), Learning without Forgetting (LwF) (model approach) [40], joint fine-tuning with UOT data selection (data approach) [47]and our TWINS fine-tuning. The robust accuracy is evaluated using l∞ norm bounded AutoAttack [12] with ϵ = 8/255. On CIFAR10, the data-based and model-based approach fail to improve clean and robust accuracy. On Caltech, both approaches improve the clean accuracy but hurt the robust accuracy. Our TWINS fine-tuning improves the clean and robust performance on both datasets. The pink triangle denotes the performance of standard AT with the non-robust pre-trained ResNet50, which drops considerably compared with fine-tuning starting from the robust pretrained model.

show abstract

Medical Image Segmentation Based on Edge Detection Techniques

Cited by 15 publications

References 8 publications

Energy based Methods for Medical Image Segmentation

Energy based Methods for Medical Image Segmentation

Region of Interest Extraction using K-Means and Edge Detection for DEXA Images

TWINS: A Fine-Tuning Framework for Improved Transferability of Adversarial Robustness and Generalization

Contact Info

Product

Resources

About