Improving Breast Cancer Detection with YOLO-v8
Author Information
Author(s): Shia Wei-Chung, Ku Tien-Hsiung, Ekpo Ernest Usang
Primary Institution: Changhua Christian Hospital, Taiwan
Hypothesis
Can the YOLO-v8 object detection algorithm enhance the detection of breast microcalcifications in mammography?
Conclusion
The YOLO-v8 model significantly outperforms traditional detection methods in detecting breast microcalcifications, improving both speed and accuracy.
Supporting Evidence
- YOLO-v8 achieved an mAP50 of 0.921 and an F1 score of 0.82.
- The model's detection accuracy was 0.842 with a recall rate of 0.796.
- Compared to Mask R-CNN, YOLO-v8 showed improvements in speed and accuracy.
Takeaway
This study shows that a new computer program can find tiny signs of breast cancer in X-ray pictures much better and faster than older methods.
Methodology
The study used the YOLO-v8 model to analyze mammograms from 7615 female participants, employing five-fold cross-validation for training and validation.
Limitations
The study is retrospective and relies on existing mammogram data, which may not represent all populations.
Participant Demographics
Female participants aged 35-70 who underwent mammography.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website