Applying Deep Learning for the Detection of Abnormalities in Mammograms

Wessels, Steven; Haar, Dustin van der

doi:10.1007/978-981-15-1465-4_21

Cited by 4 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, we differentiated between only two classes, i.e. normal and abnormal, which is consistent with most existing works on the MIAS dataset as reported in [23,[68][69][70][71][72]. Moreover, the ROI was cropped using the co-ordinates of the center and the radius of the abnormality provided by the dataset.…”

Section: Dataset Selectionmentioning

confidence: 75%

A framework for breast cancer classification using Multi-DCNNs

Ragab

Attallah

Sharkas

et al. 2021

Computers in Biology and Medicine

115

View full text Add to dashboard Cite

Section: Dataset Selectionmentioning

confidence: 75%

A framework for breast cancer classification using Multi-DCNNs

Ragab

Attallah

Sharkas

et al. 2021

Computers in Biology and Medicine

115

View full text Add to dashboard Cite

“…A modified Gaussian Noise Attack (mGNA) attack on an MNIST dataset trained on CapsNet [34] DLA reduced classification accuracy from 99.6% to 52.4%. We also discovered that the modified Jacobian-based Saliency Map attack (mJSMA) on the Cifar10 dataset trained on the MiniVGGNet [35] model has the highest robustness, with classification accuracy dropping from 81.2% to 57.8%.…”

Section: B Novel Contributionsmentioning

confidence: 99%

StAIn: Stealthy Avenues of Attacks on Horizontally Collaborated Convolutional Neural Network Inference and Their Mitigation

et al. 2023

View full text Add to dashboard Cite

With significant potential improvement in device-to-device (D2D) communication due to improved wireless link capacity (e.g., 5G and NextG systems), a collaboration of multiple edge devices (called horizontal collaboration (HC)) is becoming a reality for real-time Edge Intelligence (EI). The distributed nature of HC offers an advantage against traditional adversarial attacks because the adversary does not have access to the entire deep learning architecture (DLA). Due to the involvement of multiple untrusted edge devices in HC environment, the possibility of malicious devices cannot be eliminated. In this paper, we unearth some attacks that are very effective and stealthy even when the attacker has minimal knowledge of the DLA as is the case in HC-based DLA. We are also providing novel filtering methods to mitigate such attacks. Our novel attacks leverage local information available on output feature maps (FMs) of a targeted edge device to modify the regular adversarial attacks (e.g. Fast Gradient Signed Method (FGSM) and Jacobian-based Saliency Map Attack (JSMA)). Similarly, a customized convolutional neural network (CNN) based filter is empirically designed, developed, and tested. Four different CNN models (LeNet, CapsuleNet, MiniVGGNet, and VGG16) are used to validate the proposed attacks and defense methodologies. Our three attacks on four different CNN models (with two variations of each attack) show a substantial accuracy drop of 62% on average. The proposed filtering approach is able to mitigate the attack by recovering the actual accuracy back to 75.1% on average. To the best of our knowledge, this is the first work that investigates the security vulnerability of DLA in the HC environment, and all three of our attacks are scalable and agnostic to the partition location within the DLA.

show abstract

“…For handling this condition, the Computer-Aided Diagnosis (CAD) technique has been widely employed for minimising the false negative rates and increasing the true positive rate of BC [7]. Various scientists have explored the classification and detection of BC cells and projected distinct automatic solutions that depend on AI-based ML methods, such as Vector Quantisation, ANN, SVM, NB, RF, DT, etc., [8]. In a remote condition, where there is an absence of cancer specialists/medical expert, the service given through IoMT could be utilised by offering data through cytology images via mobile devices to e-Health care expert schemes for detecting and classifying cancer cells.…”

Section: Introductionmentioning

confidence: 99%

Cognitive Computing-Based Mammographic Image Classification on an Internet of Medical

Mansour¹,

Althobaiti²

2022

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

Recently, the Internet of Medical Things (IoMT) has become a research hotspot due to its various applicability in medical field. However, the data analysis and management in IoMT remain challenging owing to the existence of a massive number of devices linked to the server environment, generating a massive quantity of healthcare data. In such cases, cognitive computing can be employed that uses many intelligent technologies-machine learning (ML), deep learning (DL), artificial intelligence (AI), natural language processing (NLP) and others-to comprehend data expansively. Furthermore, breast cancer (BC) has been found to be a major cause of mortality among ladies globally. Earlier detection and classification of BC using digital mammograms can decrease the mortality rate. This paper presents a novel deep learning-enabled multi-objective mayfly optimization algorithm (DL-MOMFO) for BC diagnosis and classification in the IoMT environment. The goal of this paper is to integrate deep learning (DL) and cognitive computing-based techniques for e-healthcare applications as a part of IoMT technology to detect and classify BC. The proposed DL-MOMFO algorithm involved Adaptive Weighted Mean Filter (AWMF)-based noise removal and contrast-limited adaptive histogram equalisation (CLAHE)-based contrast improvement techniques to improve the quality of the digital mammograms. In addition, a U-Net architecture-based segmentation method was utilised to detect diseased regions in the mammograms. Moreover, a SqueezeNet-based feature extraction and a fuzzy support vector machine (FSVM) classifier were used in the presented technique. To enhance the diagnostic performance of the presented method, the MOMFO algorithm was used to effectively tune the parameters of the SqueezeNet and FSVM techniques. The DL-MOMFO technique was tested on the MIAS database, and the experimental outcomes revealed that the DL-MOMFO technique outperformed existing techniques.

show abstract

Applying Deep Learning for the Detection of Abnormalities in Mammograms

Cited by 4 publications

References 7 publications

A framework for breast cancer classification using Multi-DCNNs

A framework for breast cancer classification using Multi-DCNNs

StAIn: Stealthy Avenues of Attacks on Horizontally Collaborated Convolutional Neural Network Inference and Their Mitigation

Cognitive Computing-Based Mammographic Image Classification on an Internet of Medical

Contact Info

Product

Resources

About