论文部分内容阅读
目的 基于乳腺癌微钙化的重要早期征象,研究乳腺X线图像计算机辅助诊断的检测与算法.方法 提出了一种新的微钙化自动检测技术:它采用离散小波变换分解图像的高频分量,在小波域中对疑似的微钙化进行检测,并使用自适应神经网络模糊演绎系统(ANFIS)进行自适应调节;然后使用基于多层感知器(MLP)的分类器对疑似的微钙化进行筛选,以降低检测的假阳性率.结果 在使用FROC曲线中段的参数组进行实际检测时,本算法的真阳性检出率达到了96.9%,每张图像的假阳性个数为0.2个.结论 由于拥有自适应调节能力,本算法与传统的微钙化检测算法相比,具有了更高的检测精度和稳定性.“,”Objective To find a novel computer aided diagnosis algorithm for automatic detection of important signs of micro-calcifications in breast cancer mammogram. Methods Discrete wavelet transform was proposed to decompose the high-frequency information of image, detect the suspicious micro-calcifications in wavelet-domain , and apply ANFIS to adjust the parameters adaptively. Finally a multi-layer perceptrons (MLP) -based classifier was used to suppress the false positives. Results With the parameters set corresponding to the middle of the free receiver operating characteristic ( FROC) curve, the proposed algorithm achieved a true positive rate of 96.9% , with 0.2 false positive per image. Conclusion Due to the adaptive adjustment, the proposed algorithm obtains higher detection precision and stability than the traditional detection methods for micro-calcifications.