A deep convolutional neural network for the classification of imbalanced breast cancer dataset DOI Creative Commons

Robert B. Eshun,

Marwan Bikdash, Amirul Islam

et al.

Healthcare Analytics, Journal Year: 2024, Volume and Issue: 5, P. 100330 - 100330

Published: April 9, 2024

The primary procedures for breast cancer diagnosis involve the assessment of histopathological slide images by skilled patholo-gists. This procedure is prone to human subjectivity and can lead diagnostic errors with adverse implications patient health welfare. Artificial intelligence-based models have yielded promising results in other medical tasks offer tools potentially addressing shortcomings traditional image analysis. BreakHis dataset suffers from insufficient data minority class an imbalance ratio >0.40, which poses challenges deep learning models. To avoid performance degradation, researchers explored a variety augmentation schemes generate adequate samples study designed Deep Convolutional Neural Network (DCGAN) specific generator discriminator architectures mitigate model instability high-quality synthetic class. balanced was passed fine-tuned ResNet50 tumor detection. produced high accuracy diagnosing benign/malignancy at 40X magnification, outperforming state-of-art. demonstrated that methods support effective screening clinical practice.

Language: Английский

Ensemble Deep Learning-Based Image Classification for Breast Cancer Subtype and Invasiveness Diagnosis from Whole Slide Image Histopathology DOI Open Access

A. R. Balasubramanian,

Salah Alheejawi,

Akarsh Singh

et al.

Published: May 10, 2024

Cancer diagnosis and classification are pivotal for effective patient management treatment planning. In this study, we present a comprehensive approach utilizing ensemble deep learning techniques to analyze breast cancer histopathology images. Our datasets based on two widely employed from different centers tasks: BACH BreakHis. Within the Dataset, deployed an strategy incorporating VGG16 ResNet50 architec-tures achieve precise of Introducing novel image patching technique, preprocess high-resolution image, which facilitates focused analysis localized regions interest. The annotated dataset encompasses 400 WSIs across four distinct classes: Normal, Benign, Situ Carcinoma, Invasive Carcinoma. addition, BreakHis dataset, VGG16, ResNet34, models classify mi-croscopic images into eight categories (four benign malignant). For both leveraged five-fold cross-validation rigorous training testing. Preliminary ex-perimental results indicate Patch accuracy 95.31% (on dataset) WSI 98.43% (BreakHis). This research significantly contributes on-going endeavors in harnessing artificial intelligence advance diagnosis, potentially fostering improved outcomes alleviating healthcare burdens.

Language: Английский

Citations

3
Canine Mammary Tumor Histopathological Image Classification via Computer-Aided Pathology: An Available Dataset for Imaging Analysis DOI Creative Commons
Giovanni Pietro Burrai, Andrea Gabrieli, Marta Polinas

et al.

Animals, Journal Year: 2023, Volume and Issue: 13(9), P. 1563 - 1563

Published: May 6, 2023

Histopathology, the gold-standard technique in classifying canine mammary tumors (CMTs), is a time-consuming process, affected by high inter-observer variability. Digital (DP) and Computer-aided pathology (CAD) are emergent fields that will improve overall classification accuracy. In this study, ability of CAD systems to distinguish benign from malignant CMTs has been explored on dataset-namely CMTD-of 1056 hematoxylin eosin JPEG images 20 24 CMTs, with three different based combination convolutional neural network (VGG16, Inception v3, EfficientNet), which acts as feature extractor, classifier (support vector machines (SVM) or stochastic gradient boosting (SGB)), placed top net. Based human breast cancer dataset (i.e., BreakHis) (accuracy 0.86 0.91), our models were applied CMT dataset, showing accuracy 0.63 0.85 across all architectures. The EfficientNet framework coupled SVM resulted best performances an 0.82 0.85. encouraging results obtained use DP provide interesting perspective integration artificial intelligence machine learning technologies cancer-related research.

Language: Английский

Citations

7
A Comparative Study of Javanese Script Classification with GoogleNet, DenseNet, ResNet, VGG16 and VGG19 DOI Creative Commons
Ajib Susanto, Christy Atika Sari, Eko Hari Rachmawanto

et al.

Scientific Journal of Informatics, Journal Year: 2024, Volume and Issue: 11(1), P. 31 - 40

Published: Jan. 12, 2024

Purpose: Javanese script is a legacy of heritage or in Indonesia originating from the island Java needs to be preserved. Therefore, this study, classification and identification process letters will carried out using CNN method. The purpose research able build model which can properly classify script, it help recognizing easily.Methods: In has been used transfer learning Convolutional Neural Network, namely GoogleNet, DenseNet, ResNet, VGG16 VGG19. improve sequential model, processing better optimal because utilizes previously trained model.Result: results obtained after testing study are method, GoogleNet gets an accuracy 88.75%, DenseNet 92%, ResNet 82.75%, 99.25% VGG19 99.50%.Novelty: previous studies, still very rare discuss method most for performing process. had resulted find effective carry optimally.

Language: Английский

Citations

2
Breast Tumor Tissue Image Classification Using Single-Task Meta Learning with Auxiliary Network DOI Open Access

Jiann-Shu Lee,

Wenkai Wu

Cancers, Journal Year: 2024, Volume and Issue: 16(7), P. 1362 - 1362

Published: March 30, 2024

Breast cancer has a high mortality rate among cancers. If the type of breast tumor can be correctly diagnosed at an early stage, survival patients will greatly improved. Considering actual clinical needs, classification model pathology images needs to have ability make correct classification, even in facing image data with different characteristics. The existing convolutional neural network (CNN)-based models for lack requisite generalization capability maintain accuracy when confronted varied Consequently, this study introduces new model, STMLAN (Single-Task Meta Learning Auxiliary Network), which integrates and auxiliary network. Single-Task was proposed endow ability, used enhance feature characteristics images. experimental results demonstrate that improves by least 1.85% challenging multi-classification tasks compared methods. Furthermore, Silhouette Score corresponding features learned increased 31.85%, reflecting learn more discriminative features, overall is also

Language: Английский

Citations

2
A deep convolutional neural network for the classification of imbalanced breast cancer dataset DOI Creative Commons

Robert B. Eshun,

Marwan Bikdash, Amirul Islam

et al.

Healthcare Analytics, Journal Year: 2024, Volume and Issue: 5, P. 100330 - 100330

Published: April 9, 2024

The primary procedures for breast cancer diagnosis involve the assessment of histopathological slide images by skilled patholo-gists. This procedure is prone to human subjectivity and can lead diagnostic errors with adverse implications patient health welfare. Artificial intelligence-based models have yielded promising results in other medical tasks offer tools potentially addressing shortcomings traditional image analysis. BreakHis dataset suffers from insufficient data minority class an imbalance ratio >0.40, which poses challenges deep learning models. To avoid performance degradation, researchers explored a variety augmentation schemes generate adequate samples study designed Deep Convolutional Neural Network (DCGAN) specific generator discriminator architectures mitigate model instability high-quality synthetic class. balanced was passed fine-tuned ResNet50 tumor detection. produced high accuracy diagnosing benign/malignancy at 40X magnification, outperforming state-of-art. demonstrated that methods support effective screening clinical practice.

Language: Английский

Citations

2