Transfer Learning-Based Classification of Breast Cancer using Ultrasound Images

Güldoğan, Emek; Ucuzal, Hasan; Küçükakçali, Zeynep; Çolak, Cemil

doi:10.19127/mbsjohs.876667

Cited by 6 publications

(8 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A comparison of the proposed method with previous works using the same dataset is presented in Table 2 . The proposed MSTL showed the best performance compared to all published papers using the Mendeley dataset including, Acevedo et al [ 17 ], Zeebaree et al [ 18 ], and Guldogan et al [ 19 ], with accuracies of 94%, 95.4%, and 97.4%, respectively. In [ 17 , 18 ], the authors implemented classification based on manually collected features, which is how the authors taught the machine a feature to decide corresponding class, whereas in our case, we carried out an end-to-end deep learning where the model itself learns the features of each class and decides on the corresponding class using the rich capability of CNNs.…”

Section: Resultsmentioning

confidence: 86%

“…This results in the merit of having a model that is fast and not computationally complex. In [ 19 ], the authors utilized a conventional transfer learning method whereby an ImageNet pre-trained AlexNet network is used to classify breast ultrasound images. In our case, we used a multistage transfer learning method whereby additional transfer learning using cancer cell lines was carried out on top of ImageNet prior to transfer learning to classify breast ultrasound images.…”

Section: Resultsmentioning

confidence: 99%

“…In our case, we used a multistage transfer learning method whereby additional transfer learning using cancer cell lines was carried out on top of ImageNet prior to transfer learning to classify breast ultrasound images. Due to the fact that cancer cell line images possess resemblance to ultrasound images, superior transfer learning was achieved using our method on the same dataset when compared to [ 19 ].…”

Section: Resultsmentioning

confidence: 99%

“…The US image data used for this study were obtained from the publicly available Mendeley dataset ( (accessed on 8 June 2021)) composed of 250 breast US images, of which 150 are malignant cases and 100 are benign cases [ 37 ]. The dataset has been widely used in various studies [ 17 , 18 , 19 ] and is convenient to use for our purpose. We used 200 images (100 malignant and 100 benign) for preventing bias from using different sizes of data for the two classes.…”

Section: Methodsmentioning

confidence: 99%

“…Furthermore, in [ 17 , 18 ], region of interest (ROI) segmentation was used prior to classification and the ROI (patches) were utilized as input for training the classifiers. In [ 19 ], the authors utilized a transfer learning method whereby an ImageNet pre-trained AlexNet network was used to classify breast ultrasound images. The authors were able to record improved performance using their proposed method.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Novel Multistage Transfer Learning for Ultrasound Breast Cancer Image Classification

et al. 2022

View full text Add to dashboard Cite

Breast cancer diagnosis is one of the many areas that has taken advantage of artificial intelligence to achieve better performance, despite the fact that the availability of a large medical image dataset remains a challenge. Transfer learning (TL) is a phenomenon that enables deep learning algorithms to overcome the issue of shortage of training data in constructing an efficient model by transferring knowledge from a given source task to a target task. However, in most cases, ImageNet (natural images) pre-trained models that do not include medical images, are utilized for transfer learning to medical images. Considering the utilization of microscopic cancer cell line images that can be acquired in large amount, we argue that learning from both natural and medical datasets improves performance in ultrasound breast cancer image classification. The proposed multistage transfer learning (MSTL) algorithm was implemented using three pre-trained models: EfficientNetB2, InceptionV3, and ResNet50 with three optimizers: Adam, Adagrad, and stochastic gradient de-scent (SGD). Dataset sizes of 20,400 cancer cell images, 200 ultrasound images from Mendeley and 400 ultrasound images from the MT-Small-Dataset were used. ResNet50-Adagrad-based MSTL achieved a test accuracy of 99 ± 0.612% on the Mendeley dataset and 98.7 ± 1.1% on the MT-Small-Dataset, averaging over 5-fold cross validation. A p-value of 0.01191 was achieved when comparing MSTL against ImageNet based TL for the Mendeley dataset. The result is a significant improvement in the performance of artificial intelligence methods for ultrasound breast cancer classification compared to state-of-the-art methods and could remarkably improve the early diagnosis of breast cancer in young women.

show abstract

Section: Resultsmentioning

confidence: 86%

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Novel Multistage Transfer Learning for Ultrasound Breast Cancer Image Classification

et al. 2022

View full text Add to dashboard Cite

show abstract

Breast lesion classification using features fusion and selection of ensemble ResNet method

Kilicarslan¹,

Koç

Özyurt

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Medical Imaging with Deep Learning has recently become the most prominent topic in the scientific world. Significant results have been obtained in the classification of medical images using deep learning methods, and there has been an increase in studies on malignant types. The main reason for choosing breast cancer is that breast cancer is one of the critical malignant types that increase the death rate in women. In this study, 1236 ultrasound images were collected from Elazig Fethi Sekin City Hospital, and three different ResNet CNN architectures were used for feature extraction. Data were trained with an SVM classifier. In addition, the three ResNet architectures were combined, and novel fused ResNet architecture was used in this study. In addition, these features were used with three different feature selection techniques, MR‐MR, NCA, and Relieff. These results are 89.3% obtained from ALL‐ResNet architecture and the feature selected with NCA in normal and lesion classification. Normal, malignant, and benign classification best accuracy is 84.9% with ALL‐ResNet NCA. Experimental studies show that MR‐MR, NCA, and Relieff feature selection algorithms reduce features and give more results that are successful. This indicates that the proposed method is more successful than classical deep learning methods.

show abstract

EENet: Application of convolutional neural network‐based deep learning methods in bone tumor pathological diagnosis

Li,

Ding,

Wang

et al. 2024

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Bone tumors are one of the most common diseases in bone and soft tissue tumors, and accurate classification is crucial for developing effective treatment strategies. However, traditional pathological morphology diagnosis is subjective and uncertain, requiring highly specialized knowledge and experience. Therefore, how to efficiently and accurately classify bone tumor types based on pathological images is an urgent problem to be solved. In this study, a lightweight convolutional neural network (CNN) model called Efficient and Enhance Network (EENet) was proposed for the automatic recognition of bone tumor pathological images. The model introduced efficient channel attention (ECA) blocks to improve the efficiency of feature extraction and reduce computational complexity. A comprehensive dataset of bone pathological images, including five types of bone lesion tissues and normal bone tissue, was compiled for the research. Due to the low incidence of bone tumors and the difficulty in obtaining a large amount of data, the transfer learning method was chosen to overcome the problem of limited data volume. The experimental results from the fivefold cross‐validation demonstrate that the proposed model achieves 99.06% accuracy, 99.08% precision, 99.19% recall, 99.81% specificity, and 99.03% F1‐score in the bone tumor classification task, highlighting its potential as a clinical diagnostic tool.

show abstract

Transfer Learning-Based Classification of Breast Cancer using Ultrasound Images

Cited by 6 publications

References 21 publications

A Novel Multistage Transfer Learning for Ultrasound Breast Cancer Image Classification

A Novel Multistage Transfer Learning for Ultrasound Breast Cancer Image Classification

Breast lesion classification using features fusion and selection of ensemble ResNet method

EENet: Application of convolutional neural network‐based deep learning methods in bone tumor pathological diagnosis

Contact Info

Product

Resources

About