计算机算法有望提高皮肤科医生诊断皮肤黑色素瘤的准确性:国际皮肤成像协作组织2017年结果
该横断面研究使用150张皮肤镜图像(50例黑色素瘤、50例痣、50例脂溢性角化病)及23个算法结果,比较8名皮肤科医生和9名住院医师的分类准确性。排名第一的算法AUC为0.87,高于皮肤科医生(0.74)和住院医师(0.66)(P < .001)。在皮肤科医生总体敏感性76.0%时,算法特异性更优(85.0% vs. 72.6%,P = .001);将算法分类代入低置信度评估(占26.6%)后,敏感性从76.0%升至80.8%,特异性从72.6%升至72.8%。
要点
- 顶级算法AUC 0.87优于皮肤科医生0.74和住院医师0.66(P < .001)。
- 医生敏感性76.0%时,算法特异性85.0%高于医生72.6%(P = .001)。
- 算法辅助低置信度评估可提升敏感性至80.8%、特异性至72.8%。
核心结论
- 深神经网络可高精度分类黑色素瘤及其良性模仿者的皮肤图像,并可能提升人类表现。
研究设计与样本
- 横断面研究,使用150张皮肤镜图像(50例黑色素瘤、50例痣、50例脂溢性角化病)来自黑色素瘤检测挑战测试集,以及23个团队的算法结果。
- 8名皮肤科医生和9名皮肤科住院医师参与在线读片研究,对皮肤镜病变图像进行分类并提供置信度。
主要发现
- 排名第一的计算机算法受试者工作特征曲线下面积为0.87,高于皮肤科医生(0.74)和住院医师(0.66)(所有比较P < .001)。
- 在皮肤科医生总体敏感性76.0%时,算法特异性为85.0%,优于医生的72.6%(P = .001)。
- 将计算机算法分类代入低置信度评分的皮肤科医生评估(占26.6%的评估)后,皮肤科医生敏感性从76.0%升至80.8%,特异性从72.6%升至72.8%。
关键数值与推荐
- 算法AUC 0.87;皮肤科医生AUC 0.74;住院医师AUC 0.66(所有比较P < .001)。
- 皮肤科医生总体敏感性76.0%时,算法特异性85.0% vs. 医生72.6%(P = .001)。
- 低置信度评估占26.6%,代入算法后敏感性76.0%→80.8%,特异性72.6%→72.8%。
临床怎么用
- 计算机算法可作为辅助工具,尤其在医生低置信度时提升黑色素瘤诊断敏感性。
局限性
- 人工研究设置缺乏完整病变谱和临床元数据。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Computer vision has promise in image-based cutaneous melanoma diagnosis but clinical utility is uncertain. OBJECTIVE: To determine if computer algorithms from an international melanoma detection challenge can improve dermatologists' accuracy in diagnosing melanoma. METHODS: In this cross-sectional study, we used 150 dermoscopy images (50 melanomas, 50 nevi, 50 seborrheic keratoses) from the test dataset of a melanoma detection challenge, along with algorithm results from 23 teams. Eight dermatologists and 9 dermatology residents classified dermoscopic lesion images in an online reader study and provided their confidence level. RESULTS: The top-ranked computer algorithm had an area under the receiver operating characteristic curve of 0.87, which was higher than that of the dermatologists (0.74) and residents (0.66) (P < .001 for all comparisons). At the dermatologists' overall sensitivity in classification of 76.0%, the algorithm had a superior specificity (85.0% vs. 72.6%, P = .001). Imputation of computer algorithm classifications into dermatologist evaluations with low confidence ratings (26.6% of evaluations) increased dermatologist sensitivity from 76.0% to 80.8% and specificity from 72.6% to 72.8%. LIMITATIONS: Artificial study setting lacking the full spectrum of skin lesions as well as clinical metadata. CONCLUSION: Accumulating evidence suggests that deep neural networks can classify skin images of melanoma and its benign mimickers with high accuracy and potentially improve human performance.
引用
- Marchetti MA, Liopyris K, Dusza SW, et al. Computer algorithms show potential for improving dermatologists' accuracy to diagnose cutaneous melanoma: Results of the International Skin Imaging Collaboration 2017. J Am Acad Dermatol. 2019. PMID: 31306724. doi:10.1016/j.jaad.2019.07.016
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。