Breast Cancer Histopathological Image Classification Based on Graph Assisted Global Reasoning

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 22, 2025

Breast cancer ranks as the most prevalent among women globally. Histopathological image analysis stands one of reliable methods for tumor detection. This study aims to utilize deep learning extract histopathological features and automatically identify information, thereby assisting doctors in high-precision pathological diagnosis. proposes a dual-stream global–local network (DSGLNet) breast classification. The proposed DSGLNet employs feature extraction architecture that leverages convolutional local graph mapping construct global interaction space capturing information. By deeply integrating both features, achieves precise Additionally, preprocessing through engineering normalizes colors enhances details tissue cell boundaries. model was thoroughly evaluated on publicly available BreakHis dataset, encompassing different magnification levels identification nature types. 40 images achieved best diagnostic results, with reaching 0.966 accuracy 0.973 precision, outperforming other advanced methods.

Language: Английский

---

Leveraging Attention-Based Deep Learning in Binary Classification for Early-Stage Breast Cancer Diagnosis

Diagnostics, Journal Year: 2025, Volume and Issue: 15(6), P. 718 - 718

Published: March 13, 2025

Background: Breast cancer diagnosis is a global health challenge, requiring innovative methods to improve early detection accuracy and efficiency. This study investigates the integration of attention-based deep learning models with traditional machine (ML) classify histopathological breast images. Specifically, Efficient Channel-Spatial Attention Network (ECSAnet) utilized, optimized for binary classification by leveraging advanced attention mechanisms enhance feature extraction across spatial channel dimensions. Methods: Experiments were conducted using BreakHis dataset, which includes images tumors categorized as benign or malignant four magnification levels: 40×, 100×, 200×, 400×. ECSAnet was evaluated independently in combination ML models, such Decision Trees Logistic Regression. The also analyzed impact levels on accuracy, robustness, generalization. Results: Lower consistently outperformed higher magnifications terms generalization, particularly tasks. Additionally, combining improved performance, especially at lower magnifications. These findings highlight diagnostic strengths importance aligning objectives. Conclusions: demonstrates potential ECSAnet, diagnostics when integrated methods. emphasize utility provide foundation future research into hybrid architectures multimodal approaches further diagnosis.

Language: Английский

---

AMRM: Attention-based Mask Reconstruction Module for Multi-Classification of Breast Cancer Histopathological Images

Medical Engineering & Physics, Journal Year: 2025, Volume and Issue: unknown, P. 104335 - 104335

Published: April 1, 2025

Language: Английский

---

Analysis of breast cancer classification and segmentation techniques: A comprehensive review

Informatics in Medicine Unlocked, Journal Year: 2025, Volume and Issue: unknown, P. 101642 - 101642

Published: May 1, 2025

Language: Английский

---

Prognostic and predictive value of pathohistological features in gastric cancer and identification of SLITRK4 as a potential biomarker for gastric cancer

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Nov. 25, 2024

The aim of this study was to develop a quantitative feature-based model from histopathologic images assess the prognosis patients with gastric cancer. Whole slide image (WSI) H&E-stained histologic specimens cancer Cancer Genome Atlas were included and randomly assigned training test groups in 7:3 ratio. A systematic preprocessing approach employed as well non-overlapping segmentation method that combined patch-level prediction multi-instance learning integrate features across images. Subjects categorized into high- or low-risk based on median risk score derived model, significance stratification assessed using log-rank test. In addition, combining transcriptomic data other large cohort studies, we further searched for genes associated pathological their prognostic value. total 165 training, 26 integrated through learning, each process generating 11 probabilistic 2 predictive labeling features. We applied 10-fold Lasso-Cox regression achieve dimensionality reduction these accuracy verified Kaplan-Meyer (KM) curves consistency index 0.741 set 0.585 set. Deep learning-based resultant supervised pathohistological have potential superior patients, transforming pixels an effective labor-saving tool optimize clinical management patients. Also, SLITRK4 identified marker

Language: Английский

---

Deep learning approaches for detection, classification, and localization of breast cancer using microscopic images: A review and bibliometric analysis

Research on Biomedical Engineering, Journal Year: 2024, Volume and Issue: 41(1)

Published: Dec. 27, 2024

Language: Английский

---