目的本研究旨在评估HIrisPlex-S系统在法医DNA表型推断(forensic DNA phenotyping,FDP)中预测眼睛颜色、头发颜色和肤色的准确性。方法本研究共纳入49名无关个体作为研究对象。首先,基于所采集的面部照片及问卷信息,对其实际表型特征(...目的本研究旨在评估HIrisPlex-S系统在法医DNA表型推断(forensic DNA phenotyping,FDP)中预测眼睛颜色、头发颜色和肤色的准确性。方法本研究共纳入49名无关个体作为研究对象。首先,基于所采集的面部照片及问卷信息,对其实际表型特征(包括眼睛颜色、头发颜色和肤色)进行人工评估与分类。随后,采集研究对象的外周血样本并提取DNA,采用SNaPshot方法结合毛细管电泳技术对HIrisPlex-S系统进行基因分型。利用在线推断平台(https://HIrisPlex.erasmusmc.nl/)对获得的单核苷酸多态性(single nucleotide polymorphisms,SNP)位点进行分析,以预测上述表型特征。进一步采用两种不同方法解释基因型预测结果(方法1:直接选取概率值最高的表型特征作为预测表型;方法2:Manfred Kayser团队所提出的预测指南),结合人工判读的实际表型数据,分别计算预测的灵敏度和特异性,以比较两种解释方法在上述性能指标方面的差异。此外,本研究还纳入两例真实案例样本进行表型推断分析,以验证系统在实际应用中的有效性。结果研究对象中眼睛颜色为“棕色”的约占55%,“中间色”占比较少;头发颜色为“黑色”的约占45%;皮肤颜色为“白色”的约占39%。经对单碱基延伸(single base extension,SBE)引物浓度进行优化调整后,HIrisPlex-S系统中多数SNP位点的峰值更加均衡,但其在实验室间的重复性仍有待提高。在眼睛颜色的预测方面,方法1的灵敏度(0.8750、0.0000、1.0000)略高于方法2(0.8125、0.0000、0.9259),而方法1的特异性(0.8485、1.0000、0.8182)略低于方法2(0.9091、1.0000、0.9091)。在头发颜色预测中,两种方法的灵敏度一致(均为0.7727、0.6154、0.8571),方法2的特异性(1.0000、0.9444、0.7142)略优于方法1(1.0000、0.9167、0.7429)。在肤色预测方面,方法2的灵敏度(0.0000、0.7895、1.0000、0.3333、0.8750)和特异性(1.0000、0.9667、0.8333、0.9767、0.8780)均略高于方法1(灵敏度:0.0000、0.5789、0.9231、0.1667、0.7500;特异性:1.0000、0.9667、0.6944、0.9535、0.8780)。然而,总体来看,两种方法在表型预测的灵敏度与特异性方面差异均无统计学意义(P>0.05)。此外,两例真实案例样本的表型特征预测结果与遗传祖先分析一致。结论HIrisPlex-S系统可用于人群样本的眼睛和头发颜色预测,但其肤色预测的准确性有待提高。展开更多
The pervasive use of photo editing applications such as Photoshop and FaceTune has significantly altered societal beauty standards, particularly for individuals with skin of color, often leading to unrealistic expecta...The pervasive use of photo editing applications such as Photoshop and FaceTune has significantly altered societal beauty standards, particularly for individuals with skin of color, often leading to unrealistic expectations regarding skin appearance and health. These tools allow users to smooth skin textures, lighten skin tones, and erase imperfections, perpetuating Eurocentric beauty ideals that frequently marginalize the natural diversity of skin tones and textures. Consequently, individuals with skin of color may seek dermatological interventions—such as skin lightening treatments, aggressive acne scar revisions, and other cosmetic procedures—aimed at achieving appearances that align more closely with digitally manipulated images. This pursuit of an unattainable aesthetic can result in increased dissatisfaction with common skin conditions like hyperpigmentation and keloids, which are often misrepresented in edited photos. Additionally, the psychological impact of these alterations can exacerbate feelings of inadequacy, contributing to conditions such as anxiety and body dysmorphic disorder. Dermatologists face the dual challenge of addressing patients’ clinical needs while also managing their expectations shaped by digital enhancements. To combat this, it is essential for dermatologists to integrate patient education that emphasizes the beauty of diverse skin tones and the discrepancies between digital images and authentic skin health. By fostering an understanding of realistic outcomes and promoting the acceptance of natural skin characteristics, dermatologists can empower individuals with skin of color to prioritize authentic skin health over digitally influenced ideals, ultimately leading to more satisfying dermatological care and improved self-image.展开更多
针对传统高斯肤色模型在肤色和光照变化较大情况下不能有效提取肤色区域的问题,提出一种改进的高斯肤色模型,并将其应用于人脸检测中。模型参数采用一种自适应更新的参数选择方法,通过对相似度人脸和灰度人脸在对应像素点加权相乘的方式...针对传统高斯肤色模型在肤色和光照变化较大情况下不能有效提取肤色区域的问题,提出一种改进的高斯肤色模型,并将其应用于人脸检测中。模型参数采用一种自适应更新的参数选择方法,通过对相似度人脸和灰度人脸在对应像素点加权相乘的方式,得到将肤色相似度信息和灰度分布信息有效结合的人脸肤色模型,并结合Adaboost算法设计了人脸检测方法。在FERET(facial recognition technology database)、LFW(labeled faces in the wild)、GTFD(Georgia Tech face database)和多人脸图库上的实验结果表明,该模型的肤色提取正确率比传统高斯肤色模型提高了27.1%,提出的人脸检测方法的检测率比Adaboost算法提高了5.5%。展开更多
文摘目的本研究旨在评估HIrisPlex-S系统在法医DNA表型推断(forensic DNA phenotyping,FDP)中预测眼睛颜色、头发颜色和肤色的准确性。方法本研究共纳入49名无关个体作为研究对象。首先,基于所采集的面部照片及问卷信息,对其实际表型特征(包括眼睛颜色、头发颜色和肤色)进行人工评估与分类。随后,采集研究对象的外周血样本并提取DNA,采用SNaPshot方法结合毛细管电泳技术对HIrisPlex-S系统进行基因分型。利用在线推断平台(https://HIrisPlex.erasmusmc.nl/)对获得的单核苷酸多态性(single nucleotide polymorphisms,SNP)位点进行分析,以预测上述表型特征。进一步采用两种不同方法解释基因型预测结果(方法1:直接选取概率值最高的表型特征作为预测表型;方法2:Manfred Kayser团队所提出的预测指南),结合人工判读的实际表型数据,分别计算预测的灵敏度和特异性,以比较两种解释方法在上述性能指标方面的差异。此外,本研究还纳入两例真实案例样本进行表型推断分析,以验证系统在实际应用中的有效性。结果研究对象中眼睛颜色为“棕色”的约占55%,“中间色”占比较少;头发颜色为“黑色”的约占45%;皮肤颜色为“白色”的约占39%。经对单碱基延伸(single base extension,SBE)引物浓度进行优化调整后,HIrisPlex-S系统中多数SNP位点的峰值更加均衡,但其在实验室间的重复性仍有待提高。在眼睛颜色的预测方面,方法1的灵敏度(0.8750、0.0000、1.0000)略高于方法2(0.8125、0.0000、0.9259),而方法1的特异性(0.8485、1.0000、0.8182)略低于方法2(0.9091、1.0000、0.9091)。在头发颜色预测中,两种方法的灵敏度一致(均为0.7727、0.6154、0.8571),方法2的特异性(1.0000、0.9444、0.7142)略优于方法1(1.0000、0.9167、0.7429)。在肤色预测方面,方法2的灵敏度(0.0000、0.7895、1.0000、0.3333、0.8750)和特异性(1.0000、0.9667、0.8333、0.9767、0.8780)均略高于方法1(灵敏度:0.0000、0.5789、0.9231、0.1667、0.7500;特异性:1.0000、0.9667、0.6944、0.9535、0.8780)。然而,总体来看,两种方法在表型预测的灵敏度与特异性方面差异均无统计学意义(P>0.05)。此外,两例真实案例样本的表型特征预测结果与遗传祖先分析一致。结论HIrisPlex-S系统可用于人群样本的眼睛和头发颜色预测,但其肤色预测的准确性有待提高。
文摘The pervasive use of photo editing applications such as Photoshop and FaceTune has significantly altered societal beauty standards, particularly for individuals with skin of color, often leading to unrealistic expectations regarding skin appearance and health. These tools allow users to smooth skin textures, lighten skin tones, and erase imperfections, perpetuating Eurocentric beauty ideals that frequently marginalize the natural diversity of skin tones and textures. Consequently, individuals with skin of color may seek dermatological interventions—such as skin lightening treatments, aggressive acne scar revisions, and other cosmetic procedures—aimed at achieving appearances that align more closely with digitally manipulated images. This pursuit of an unattainable aesthetic can result in increased dissatisfaction with common skin conditions like hyperpigmentation and keloids, which are often misrepresented in edited photos. Additionally, the psychological impact of these alterations can exacerbate feelings of inadequacy, contributing to conditions such as anxiety and body dysmorphic disorder. Dermatologists face the dual challenge of addressing patients’ clinical needs while also managing their expectations shaped by digital enhancements. To combat this, it is essential for dermatologists to integrate patient education that emphasizes the beauty of diverse skin tones and the discrepancies between digital images and authentic skin health. By fostering an understanding of realistic outcomes and promoting the acceptance of natural skin characteristics, dermatologists can empower individuals with skin of color to prioritize authentic skin health over digitally influenced ideals, ultimately leading to more satisfying dermatological care and improved self-image.
文摘针对传统高斯肤色模型在肤色和光照变化较大情况下不能有效提取肤色区域的问题,提出一种改进的高斯肤色模型,并将其应用于人脸检测中。模型参数采用一种自适应更新的参数选择方法,通过对相似度人脸和灰度人脸在对应像素点加权相乘的方式,得到将肤色相似度信息和灰度分布信息有效结合的人脸肤色模型,并结合Adaboost算法设计了人脸检测方法。在FERET(facial recognition technology database)、LFW(labeled faces in the wild)、GTFD(Georgia Tech face database)和多人脸图库上的实验结果表明,该模型的肤色提取正确率比传统高斯肤色模型提高了27.1%,提出的人脸检测方法的检测率比Adaboost算法提高了5.5%。
