Popecki Paweł, Jurczyszyn Kamil, Ziętek Marcin, Kozakiewicz Marcin
Department of Oral Surgery, Wroclaw Medical University, Krakowska 26, 50-425 Wroclaw, Poland.
Department of Oncology, Wroclaw Medical University, Plac Hirszfelda 12, 53-413 Wroclaw, Poland.
J Clin Med. 2022 Apr 29;11(9):2505. doi: 10.3390/jcm11092505.
The differential diagnosis of benign nevi (BN), dysplastic nevi (DN), and melanomas (MM) represents a considerable clinical problem. These lesions are similar in clinical examination but have different prognoses and therapeutic management techniques. A texture analysis (TA) is a mathematical and statistical analysis of pixel patterns of a digital image. This study aims to demonstrate the relationship between the TA of digital images of pigmented lesions under polarized and non-polarized light and their histopathological diagnosis. Ninety pigmented lesions of 76 patients were included in this study. We obtained 166 regions of interest (ROI) images for MM, 166 for DN, and 166 for BN. The pictures were taken under polarized and non-polarized light. Selected image texture features (entropy and difference entropy and long-run emphasis) of ROIs were calculated. Those three equations were used to construct the texture index (TI) and bone index (BI). All of the presented features distinguish melanomas, benign and dysplastic lesions under polarized light very well. In non-polarized images, only the long-run emphasis moment and both indices effectively differentiated nevi from melanomas. TA is an objective method of assessing pigmented lesions and can be used in automatic diagnostic systems.
良性痣（BN）、发育异常痣（DN）和黑色素瘤（MM）的鉴别诊断是一个相当大的临床问题。这些病变在临床检查中相似，但预后和治疗管理技术不同。纹理分析（TA）是对数字图像像素模式的数学和统计分析。本研究旨在证明偏振光和非偏振光下色素沉着病变数字图像的TA与其组织病理学诊断之间的关系。本研究纳入了76例患者的90个色素沉着病变。我们获得了166个MM的感兴趣区域（ROI）图像、166个DN的ROI图像和166个BN的ROI图像。这些图片是在偏振光和非偏振光下拍摄的。计算了ROI的选定图像纹理特征（熵、差异熵和长期强调）。使用这三个公式构建纹理指数（TI）和骨指数（BI）。所有呈现的特征在偏振光下能很好地区分黑色素瘤、良性和发育异常病变。在非偏振图像中，只有长期强调矩和两个指数能有效区分痣和黑色素瘤。TA是评估色素沉着病变的一种客观方法，可用于自动诊断系统。
