DeepInsight-Net: a CBAM-enhanced ResNet50 framework with focal loss for robust cervical cancer classification on multi-center datasets - Report - MDSpire
Advertisement
DeepInsight-Net: a CBAM-enhanced ResNet50 framework with focal loss for robust cervical cancer classification on multi-center datasets
Clinical Report: DeepInsight-Net for Cervical Cancer Classification
Overview
DeepInsight-Net, a novel deep learning framework, achieves a classification accuracy of 99.63% for cervical cancer, significantly outperforming existing models. The integration of Convolutional Block Attention Modules (CBAM) and Focal Loss addresses critical challenges in cervical cytology, including spatial irrelevance and class imbalance.
Background
Cervical cancer is a major global health issue, being the fourth most common cancer in women and a leading cause of cancer-related mortality. Traditional cytological screening methods are limited by inter-observer variability and high false-negative rates. The advent of deep learning technologies offers a promising avenue for improving diagnostic accuracy and efficiency in cervical cancer screening.
Data Highlights
Model
Accuracy
DeepInsight-Net
99.63%
Cross-dataset (LBC)
98.62%
Key Findings
DeepInsight-Net integrates CBAM with a ResNet50 backbone to enhance feature discrimination.
Utilizes Focal Loss to address class imbalance by prioritizing hard-to-classify instances.
Achieved a state-of-the-art accuracy of 99.63% on the SiPaKMeddataset.
Demonstrated robust cross-dataset generalization with an accuracy of 98.62% on an independent LBC dataset.
Visual interpretability analyses confirm the model's focus on biologically relevant cellular regions.
Clinical Implications
The DeepInsight-Net framework presents a significant advancement in the automation of cervical cancer screening, potentially reducing the workload on cytologists and improving diagnostic accuracy. Its ability to effectively handle class imbalance and spatial irrelevance makes it a valuable tool for enhancing the reliability of cervical cancer diagnostics in clinical practice.
Conclusion
DeepInsight-Net represents a promising development in the field of cervical cancer classification, combining advanced deep learning techniques to improve diagnostic performance. Its application could lead to better outcomes in cervical cancer screening and management.
