杜兴氏肌营养不良(DMD)是一种严重的儿童腿部神经肌肉罕见病。传统的诊断和检测方案一般为有创手段, 会带给患儿极大的痛苦。基于受试者的磁共振图像(MRI), 采用计算机辅助检测手段探索了有效的无创检测方法。实验分别选用sym4和db4两种小波基函数, 对患儿组和健康对照组的MRI进行三种尺度的小波分解, 从所得的分解图像中提取12个纹理特征参数, 并利用人工神经网络(ANN)算法对图像参数进行分类识别。结果显示: 在受试者的两类MRI加权图像(T1和T2)中, T1图像能更好地区分患儿与健康儿童; 利用db4函数对图像进行小波分解, 其效果略优于sym4函数, 且在三种小波分解尺度中, 以二层分解最优; 利用ANN算法对图像进行分类识别, 其灵敏度、特异度和准确率分别高达98.5%、97.3%和97.9%。该处理方法有望为临床提供客观有效的辅助诊断手段, 可作为DMD疾病无创检测的尝试探索。

Duchenne muscular dystrophy (DMD) is a severe orphan neuromuscular disease in the part of legs. The conventional treatment is invasive, which incurs great sufferings. Therefore, with the aid of computers, a non-invasive detection method is explored on the basis of magnetic resonance images (MRI) of the patients. Two wavelet basis function, sym4 and db4 are used and wavelet decomposition is conducted for three levels of MRI from both the sick and the healthy. 12 texture parameters are extracted from the decomposed images. In the end, classification and recognition of these images are carried out by using Artificial Neural Network on the basis of these texture feature parameters. Conclusions of the study are as follows: (1) In the two kinds of weighed images (T1 and T2) of the sick, T1 is better in distinguishing the sick from the healthy. (2) The results of wavelet decomposition with db4 function are better than those with sym4 function, and in the three levels decomposed, the second level is the best. (3) With the optimal wavelet function and decomposition level, classification and recognition can produce very excellent outcomes. The sensitivity, specificity and accuracy rate might reach as high as 98.5%, 97.3%, 97.9% respectively. This method, as a pilot for non-invasive treatment of DMD, could be expected to provide an objective and effective auxiliary method for clinical diagnoses.