| 摘 要: 提出基于计算机视觉与机器学习的医学影像处理法。利用图像处理技术预处理X光图像,借助VGG16预训练模型提取高级特征,用贝叶斯和KNN分类器训练模型,进行不同数据集评估。实验显示,模型准确率约为0.9,能精准判别图像有无病变,经像素差异计算与高斯模糊处理,可直观展现变化区域。随着数据量增大,F1值超过0.8,召回率超0.9,表明该方法适配常规影像任务。即便影像噪声多、病变变化小,也能给出精确分析结果,辅助医生识别病变。 |
| 关键词: 医学影像分析 X光图像 图像差异检测 HOG特征 VGG16 贝叶斯分类器 KNN分类器 |
|
中图分类号: TP391.4
文献标识码: A
|
| 基金项目: 甘肃省重点研发计划项目(21YF5GA088) |
|
| Research on Medical Image Difference Detection and Classification Method Based on Multi-feature Fusion and Machine Learning |
|
XIAO Lili, JIANG Haibo
|
(College of Science, Nanjing University of Posts and Telecommunications, Nanjing 210003, China)
xll@njupt.edu.cn; 20230081@njupt.edu.cn
|
| Abstract: This study proposes a medical image processing approach based on computer vision and machine learning. X-ray images are first preprocessed using image processing techniques, followed by extracting high-level features via the VGG16 pre-trained model. Bayesian and KNN classifiers are then employed to train the model, which is evaluated across diverse datasets. Experiments demonstrate that the model achieves an accuracy of approximately 0.9,enabling precise identification of lesions in images. Through pixel difference calculation and Gaussian blur processing,it visually highlights regions of change. As data volume increases, the F1-score exceeds 0.8 and recall surpasses 0.9.This indicates the method’s suitability for routine imaging tasks, providing accurate analysis even for noisy images or subtle lesion variations, thereby assisting doctors in lesion identification. |
| Keywords: medical image analysis X-ray images image difference detection HOG features VGG16 Bayesian
classifier KNN classifier |
