PDF(Sci-hub)
Abstract：
A Probabilistic Neural Network (PNN) is a statistical algorithm and consists of a grouping of multi-class data. The conventional method of detection of DNA mutations by the human eye may not detect the minute variations in PCR-SSCP bands, which may lead to false positive or false negative results. The detection by photographic images may contain a blare (noise) caused during the time of photography; therefore, image processing techniques were used to reduce image noise. PCR-SSCP gels of T2DM patients (n = 100) and controls (n = 100) were initially photographed with equal ratio of pixels and later subjected to a two-stage analysis: feature extraction and PNN. The evaluation of the results was done by quality training and the accuracy was up to 95%, and the human eye analysis showed 80% mutation detection rate. This study proves to be very reliable and gives accurate and fast detection for mutation analysis in diabetes. This method could be extended for analysis in other human diseases.
摘要：
概率神经网络（PNN）是一种统计算法，由一组多类数据组成。通过人眼检测DNA突变的常规方法可能无法检测到PCR-SSCP条带的微小变化，这可能导致假阳性或假阴性结果。摄影图像的检测可能包含在摄影期间引起的杂音（噪声）；因此，图像处理技术用于降低图像噪声。最初以相等的像素比例拍摄T2DM患者（n=100）和对照组（n=100）的PCR-SSCP凝胶，然后进行两阶段分析：特征提取和PNN。通过质量培训对结果进行评估，准确性高达95％，人眼分析显示80%的突变检出率。这项研究被证明是非常可靠的，并为糖尿病的突变分析提供了准确和快速的检测。该方法可以扩展用于其他人类疾病的分析。
