Tognetti L, Cartocci A, Cinotti E, Moscarella E, Farnetani F, Lallas A, Tiodorovic D, Carrera C, Longo C, Puig S, Perrot J L, Argenziano G, Pellacani G, Cataldo G, Balistreri A, Cevenini G, Rubegni P
Dermatology Unit, Department of Medical, Surgical and Neurosciences, University of Siena, Siena, Italy.
Bioengineering & Biomedical Data Science Lab, Department of Medical Biotechnology, University of Siena, Siena, Italy.
J Eur Acad Dermatol Venereol. 2021 Mar;35(3):650-657. doi: 10.1111/jdv.16847. Epub 2020 Aug 24.
The anatomical location of atypical melanocytic skin lesion (aMSL) was never combined into an algorithm for discriminating early melanomas (EM) from atypical nevi (AN).
To investigate the impact of body location on the intuitive diagnosis performed in teledermoscopy by dermatologists of different skill levels. A further aim was to evaluate how the integration of the body location could improve an algorithm-aided diagnosis.
We retrospectively collected 980 standardized dermoscopic images of aMSL cases (663 AN, 317 EM): data on the anatomical location were collected according to 15 body sites classified into 4 macro-areas of chronically/frequently/seldom/rarely exposure. Through a teledermatology web platform, 111 variously skilled dermoscopists performed either the intuitive diagnosis and 3 algorithm-assisted diagnostic tests (i.e. iDScore, 7-point checklist, ABCD rule) on each case, for a total of 3330 examinations.
In the rarely photoexposed area (side, bottom, abdomen), AN were the most tricky (i.e. highest quote of false positives), due to a frequent recognition of dermoscopic features usually considered as suggestive for melanoma in these lesions; the EM at these sites received the highest quote of false negatives, being generally interpreted as 'featureless' according to these traditional parameters, that were more frequently displayed on the chronically photoexposed area. In rarely and seldom photoexposed area, intuitive diagnosis fails to achieve adequate accuracy for all aMSLs, as the ABCD rule and the 7-point checklist; by applying the iDScore algorithm the diagnostic performance was increased by 15% in young and 17% in experts.
The body location of an aMSL can affect the quality of intuitive dermoscopic diagnosis, especially in sun-protected areas. Accuracy can be improved by using the iDScore algorithm that assigns a different partial score of each body site.
非典型黑素细胞性皮肤病变（aMSL）的解剖位置从未被纳入用于区分早期黑色素瘤（EM）和非典型痣（AN）的算法中。
研究身体部位对不同技术水平的皮肤科医生在远程皮肤镜检查中进行直观诊断的影响。另一个目的是评估身体部位的纳入如何改善算法辅助诊断。
我们回顾性收集了980例aMSL病例的标准化皮肤镜图像（663例AN，317例EM）：根据15个身体部位收集解剖位置数据，这些部位分为4个长期/频繁/偶尔/极少暴露的宏观区域。通过远程皮肤病学网络平台，111名技术水平各异的皮肤镜检查医生对每个病例进行直观诊断和3种算法辅助诊断测试（即iDScore、7分检查表、ABCD规则），共进行了3330次检查。
在极少暴露于阳光的区域（侧面、底部、腹部），AN最难诊断（即假阳性率最高），因为这些病变中通常被认为提示黑色素瘤的皮肤镜特征经常被识别；这些部位的EM假阴性率最高，根据这些传统参数通常被解释为“无特征”，而这些参数在长期暴露于阳光的区域更频繁出现。在极少和偶尔暴露于阳光的区域，直观诊断以及ABCD规则和7分检查表对所有aMSL均未能达到足够的准确性；应用iDScore算法后，年轻医生的诊断性能提高了15%，专家提高了17%。
aMSL的身体部位会影响直观皮肤镜诊断的质量，尤其是在防晒区域。使用为每个身体部位分配不同部分分数的iDScore算法可以提高准确性。
