Cengizler Çağlar, Kabakci Ayse Gul, Bozkır Dursun Murat, Sire Eren Dilek, Bozkır Memduha Gülhal
Biomedical Device Technology Program, Vocational School of Health Services, Izmir Democracy University, Izmir, Turkey.
Department of Anatomy, Faculty of Medicine, Cukurova University, Adana, Turkey.
Clin Cosmet Investig Dermatol. 2023 Sep 18;16:2537-2546. doi: 10.2147/CCID.S425797. eCollection 2023.
Dark circles and pigmentation around the eyes are common reasons people see dermatologists. An effective assessment of the severity of infraorbital color and texture differences is critical for determining appropriate treatment. Evaluation of the visual severity of cases is mostly based on visual inspection. Treatment efficiency is often measured using patient questionnaires in many cases. The subjectivity of assessments may lead to a prolonged healing process, misdiagnosis, and excessive use of drugs or chemicals.
In this study, a computer-aided objective evaluation approach was proposed for grading periorbital facial rejuvenation. This approach is based on the analysis of numerical features extracted from different facial regions in digital images. Healing was objectively graded by evaluating data clusters formed using the extracted features. Accordingly, an increase in the visual similarity between paired facial regions is accepted as an indicator of healing, which directly affects the form of data clusters. An intracluster validity index was measured to evaluate the clusters as dense and well separated. A total of 144 facial regions were extracted and examined, and the automatically calculated grades were compared with expert evaluations.
The cosmetic effects of the experimental drug were evaluated during the experiments, and expert grades were accepted as the ground truth. The results show that the proposed automated grading approach can evaluate rejuvenation with an accuracy of up to 0.91 accuracy in the upper orbital region.
This study concluded that the proposed data-clustering-based approach is promising and can be functional without any special instruments.
黑眼圈和眼周色素沉着是人们看皮肤科医生的常见原因。有效评估眶下颜色和质地差异的严重程度对于确定适当的治疗方法至关重要。对病例视觉严重程度的评估大多基于目视检查。在许多情况下，治疗效果通常使用患者问卷来衡量。评估的主观性可能导致愈合过程延长、误诊以及药物或化学品的过度使用。
在本研究中，提出了一种用于眶周面部年轻化分级的计算机辅助客观评估方法。该方法基于对从数字图像中不同面部区域提取的数值特征的分析。通过评估使用提取特征形成的数据簇来客观地对愈合情况进行分级。因此，配对面部区域之间视觉相似度的增加被视为愈合的指标，这直接影响数据簇的形式。测量簇内有效性指数以评估簇的密集程度和良好分离度。总共提取并检查了144个面部区域，并将自动计算的分级与专家评估进行比较。
在实验过程中评估了实验药物的美容效果，并将专家分级作为基准事实。结果表明，所提出的自动分级方法在上眼眶区域评估年轻化的准确率高达0.91。
本研究得出结论，所提出的基于数据聚类的方法很有前景，并且无需任何特殊仪器即可发挥作用。
