Photo analysis offers a simple, noninvasive approach to characterizing and quantifying skin lesions in cetaceans; however, this process involves methodological considerations that have often gone unaddressed or have varied in approach among investigators. Subjectivity associated with classifying skin lesion types of unknown etiology and quantifying measures of skin lesion prevalence and extent from photo data raises questions about observer bias and agreement (i.e., interrater reliability), which are often ignored. The purpose of the present study was to improve upon data quality control and assessment practices when studying skin lesions using only photo data. Specifically, we tested interrater reliability of a skin lesion classification system, compared methods of quantifying skin lesion extent, and determined the validity of the dorsal fin as a proxy for skin lesions on the entire body. Acceptable levels of interrater reliability were achieved for only 7 of 17 defined lesion types, but reliability was high for the two tested measures of lesion extent. Skin lesion extent measured from the dorsal fin alone was not a decent proxy for the whole visible surface; disparities between measures were as high as 43%. We discuss the potential pitfalls discovered and provide recommendations for others attempting similar approaches.

中文翻译：

---

视觉评估自由放养的普通宽吻海豚（Tursiops truncatus）上的皮肤病变：定量工具的可靠性和实用性

照片分析提供了一种简单，非侵入性的方法来表征和量化鲸类动物的皮肤病变。但是，此过程涉及方法学上的考虑，研究人员往往对此未加以解决或方法有所不同。与对病因不明的皮肤病变类型进行分类以及根据照片数据量化皮肤病变患病率和程度的度量相关的主观性提出了有关观察者偏见和一致性（即间位可靠度）的问题，这些问题通常被忽略。本研究的目的是改善仅使用照片数据研究皮肤病变时的数据质量控制和评估实践。具体来说，我们测试了皮肤病变分类系统的界面可靠性，比较了量化皮肤病变程度的方法，并确定背鳍作为全身皮肤病变的替代药物的有效性。在17种已定义的病变类型中，只有7种达到了可接受的间质可靠性，但是对于两种测试的病变程度，其可靠性很高。仅从背鳍上测得的皮肤病变程度并不是整个可见表面的良好代表。措施之间的差异高达43％。我们讨论发现的潜在陷阱，并为其他尝试类似方法的人提供建议。措施之间的差异高达43％。我们讨论发现的潜在陷阱，并为其他尝试类似方法的人提供建议。措施之间的差异高达43％。我们讨论发现的潜在陷阱，并为其他尝试类似方法的人提供建议。