【作者】 曲龙；

【导师】 孙贵范；

【作者基本信息】 中国医科大学 ， 劳动卫生与环境卫生， 2004， 硕士

【摘要】 实验目的 鉴于目前彗星试验图像分析软件存在的种种问题与不足，如它对于分析结果无法进行必要的人为校正、对于高度损伤的细胞可能分析错误、程序鲁棒性较差、内部文档不足、不易操作等，本课题在Helma等人的宏程序基础上进行了改进，着重解决高度损伤细胞的分析不准确甚至完全错误的问题，同时增加手动校正功能，以尽可能保证试验结果的真实性、准确性，并完善了该宏的容错性能与内外部文档，提高其鲁棒性，使用户更易操作。 实验方法 在对图像分析错误的改进中，本研究采用了以梯度阈值为基础的综合性彗星头中心点确定方法，代替原来计算彗星头中心的固定阈值(80％)方法。本研究还增加了对分析错误的数据进行人为校正的功能，由操作者重新指定正确的彗星头中心，从而把错误数据纠正为正确数据，并入统计分析。在容错性能的改善上，本研究在所有可能被用户调用的宏过程的开始处均加入了错误检查和异常处理程序段，可以暂停程序执行并给出正确的操作方法说明。同时，本研究应用改进后软件进行标准彗星试验检测，以100μM过氧化氢染毒小鼠外周血单个核细胞，并经历0、10、20、40min不同时间的孵育，获得不同损伤级别的细胞，并以改进后的软件进行分析。比较改进前后的尾长、尾矩、尾DNA百分含量上的差异、改进前后分析正确率差异、以及改进后的软件以不同DNA损伤级别为截断值判定阳性结果的灵敏度、特异度、Youden指数、粗一致性、调整一致性、阳性预测值、阴性预测值、阳性似然比和阴性似然比。实验结果 改进后的软件对于原软件分析错误的图像在大多数情况下都可以正确分析，同时拥有了错误数据的人工校正功能，程序鲁棒性有所增强，更加方便易用。对于0一2级损伤的彗星细胞来说，原软件与改进后的软件在尾长、尾矩、尾DNA百分含量上的分析具有统计学非常显著意义，但实际差别量不大，没有实际生物学意义;而对3一4级损伤的彗星细胞来说，改进后的软件比原软件在统计学上更加接近真值，且实际差别量足够大，同时具有统计学非常显著意义和生物学意义;改进后的软件对于1级及1级以上彗星细胞的判定来说，灵敏度97.70%、特异度67.42%、Youden指数0.“、粗一致性79.45%、调整一致性82.33%、阳性预测值“.41%、阳性似然比3.00、阴性预测值97.80%、阴性似然比0.03;而改进后的软件对于2级及2级以上彗星细胞的判定来说，灵敏度95.38%、特异度89.89%、Youden指数0.85、粗一致性93.150/0、调整一致性 92.88%、阳性预测值93.23%、阳性似然比9.43、阴性预测值93.02%、阴性似然比0.05。结论 彗星试验图像分析软件经改进后，分析准确性得到提高，程序鲁棒性增强，而且易用性改善，并增加了由操作者进行必要校正的功能;所得分析结果与人为指定的分析结果相比，并不同时存在统计学显著差异和生物学显著差异，可作为真实可靠的彗星数据分析系统应用;改进后的系统在拖尾细胞判定上具有很高灵敏度，可用于筛查试验项目;在2级及2级以上损伤细胞的判定上同时具有较高的灵敏度和特异度，是实验室研究中的优良指标。更多还原

【Abstract】 To solve the problem of wrong or inaccurate analyses of severely damaged Comet cells by the Comet Assay image analysis macro software, to implement the function of manual correction to the macro, and to improve the robustness and user friendliness of the software.MethodsA comprehensive method was used for the determination of Comet head center based on gradually increased thresholds in sequential Particle Analysis of Scion Image. Several macros were added to implement the manual correction function. And at the beginning of each macro procedure, operation check processes were inserted to minimize the user interference. A standard Comet Assay was conducted using the improved comet analysis macro, with the intoxication by 100//M hydrogen peroxide to mononuclear cells of mouse peripheral blood for 0, 10, 20, 40min respectively to get different DNA damage grades. Tail Length, Tail Moment, Tail DNA Percentage, and correction rate were compared between the improved software and the original one, and epidemiological indices for the improved software were also calculated and analyzed, including sensitivity, specificity, Youden index, crude agreement, adjusted agreement, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio.Results The improved software can correctly analyze most of the images towhich the original one cannot give right results. Moreover, the improved one has manual correction function, as well as better robustness and user friendliness. For Comet cells with 0~2 DNA damage grades, the results of the original and improved software were statistically different, however, with no biological significance. For cells with 3~4 damage grades, the improved software were both statistically and biologically different with the original one. In determining cells with 1 damage grades, the improved software gave the sensitivity of 97.70%, specificity 67.42%, Youden Index 0.65, crude agreement 79.45%, adjusted agreement 82.33%, positive predictive value 66.41%, positive likelihood ratio 3.00, negative predictive value 97.80%, and negative likelihood ratio 0.03. While in determining cells with 2 damage grades, the improved software gave the sensitivity of 95.38%, specificity 89.89%, Youden Index 0.85, crude agreement 93.15%, adjusted agreement 92.88%, positive predictive value 93.23%, positive likelihood ratio 9.43, negative predictive value 93.02%, and negative likelihood ratio 0.05.ConclusionThe improved Comet Assay image analysis software is better than the original one, in respect to analysis accuracy, robustness, user friendliness, and manual correction. The results given by the improved software are not both statistically and biologically different from manually specified results, and thus can be used as an effective alternative for the latter. It has a very high sensitivity in determining cells with a tail, meeting the needs of screening tests. It has both a high sensitivity and a high specificity in determining cells with 2 damage grade, indicating a potentially good laboratory index.更多还原