Cervical Cancer Screening Using Cell Aggregate Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cervical cancer screening technologies, particularly those using general-purpose object detection schemes based on deep learning, suffer from low precision, resulting in a high rate of false negatives and inadequate detection of abnormal cells in multi-cell cytodiagnostic images, which is insufficient for effective cytodiagnosis.
Innovation Solution
A cervical cancer screening support system that includes an image acquisition unit, a cell aggregate recognition unit using the YOLO algorithm, and estimation units generated through machine learning, which recognize and classify cell aggregates and estimate the position and class of abnormal cells with high precision, employing a combination of object detection and image classification algorithms to improve diagnostic accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If a general-purpose object detection scheme based on deep learning is used for cervical cancer screening, then the screening process can be automated and the workload reduced, but the precision is low resulting in high false negative rates
Solution Approach 1:
The system segments the detection task into two distinct stages: first, a cell aggregate recognition unit identifies and localizes cell aggregates using YOLO algorithm; second, an estimation unit performs detailed classification of abnormal cells within those aggregates. This segmentation allows each stage to be optimized for its specific function, improving overall precision while maintaining automation.
Solution Approach 2:
The cell aggregate recognition unit acts as an intermediary between the initial image acquisition and the final abnormal cell classification. By first identifying cell aggregates and their locations, this intermediate step provides focused regions for the estimation unit to analyze, thereby improving detection precision without sacrificing automation.
2Productivity
If a general-purpose object detection scheme is applied to multi-cell cytodiagnostic images, then processing speed can be maintained, but the recall factor is less than 70% leading to increased overlooking of positive findings
Solution Approach 1:
The system divides the processing workflow into sequential specialized stages: rapid cell aggregate detection using YOLO algorithm maintains processing speed, while the subsequent estimation unit performs thorough classification to improve recall. This segmentation allows speed and reliability to be optimized at different stages respectively.
Solution Approach 2:
The system maintains continuous processing by using the YOLO algorithm for real-time cell aggregate recognition, then seamlessly transitions to the estimation unit for classification without interrupting the workflow. This continuous action ensures both high processing speed and improved reliability through thorough multi-stage analysis.
3Device complexity
If cell aggregates are present in the micrograph, then the complexity of the image increases making detection more difficult, but the system should still be able to identify abnormal cells accurately
Solution Approach 1:
The system segments the complex image analysis task into manageable components: the cell aggregate recognition unit first identifies and localizes aggregates regardless of their complexity, then the estimation unit focuses specifically on classifying abnormal cells within those localized regions. This segmentation breaks down the complexity while maintaining precision.
Solution Approach 2:
The system applies different processing qualities to different regions: the cell aggregate recognition unit provides broad coverage to identify all aggregates, while the estimation unit applies detailed local analysis specifically to the regions containing cell aggregates. This local quality approach handles image complexity effectively while maintaining detection precision.
Data Source
AI summary
A cervical cancer screening support system includes: an image acquisition unit that acquires a micrograph of a cell for cytodiagnosis of a cervix of uterus; a cell aggregate recognition unit that recognizes a cell aggregate in the micrograph; and an output unit that outputs a class applicable to a cell belonging to the cell aggregate.