| 摘 要: 儿童颅骨畸形的判断是典型的机器学习分类问题,文章对电子计算机断层扫描(Computed Tomography,CT)图像进行三维建模和数据增强,从中提取9个数值特征并分析其在正常颅骨与畸形颅骨之间的区别,表明所选特征的有效性。选用Stacking(堆叠)方法构建针对颅骨畸形自动判断的集成学习分类模型,将其应用到医院提供的临床CT数据集进行十折交叉验证实验。实验结果表明,集成学习模型优于单一的分类模型,其平均预测准确度约为0.986,F1值(统计学中用于衡量分类模型精确度的指标,同时兼顾了分类模型的准确率和召回率)约为0.982,各项评价指标也优于目前同类模型。 |
| 关键词: 颅骨畸形;集成学习;图像处理;分类预测 |
|
中图分类号: TP311.5
文献标识码: A
|
| 基金项目: 国家自然科学基金(81801797) |
|
| Research on Cranial Deformities Classification of Children Based on Machine Learning |
|
ZHANG Shunyu, HU Jun, LIN Yong
|
(School of Health Science and Engineering, University of Shanghai f or Science and Technology, Shanghai 200093, China)
zsy3216@163.com; hujun@163.com; yong_lynn@163.com
|
| Abstract: Diagnosis of children cranial deformities is a typical classification problem of machine learning. This paper proposes to conduct three-dimensional modeling and data augmentation on CT (Computed Tomography) images, from which nine numerical features are extracted, and their differences between normal and deformed cranial bones are analyzed, demonstrating the effectiveness of the selected features. The stacking method is selected to construct an ensemble learning classification model for automatic detection of cranial deformities, and it is applied to the clinical CT dataset provided by a certain hospital for ten-fold cross-validation experiments. The experimental results show that the ensemble learning model is superior to a single classification model, with an average prediction accuracy of about 0.986 and an F1 value (a statistical indicator used to measure the accuracy of classification models, while also considering the accuracy and recall of classification models) of about 0.982. All of its evaluation indicators are also better than similar models currently available. |
| Keywords: cranial deformities; ensemble learning; image processing; classification predictionn |
