文献速递皮肤肿瘤

卷积神经网络联合三维全身摄影自动分类部位特异性皮肤光损伤

Automated classification of site-specific cutaneous photodamage using a convolutional neural network and three-dimensional total body photography.
PubMed 文献 · Br J Dermatol · 2026-01-01

多任务CNN较单任务CNN的ROC-AUC从0.91提升至0.96(P<0.01),外部验证ROC-AUC为0.93。

要点

  • 多任务CNN自动评估光损伤的ROC-AUC达0.96,优于单任务CNN的0.91。
  • 外部验证ROC-AUC为0.93,轻、中、重度光损伤分类准确率分别为0.97、0.85、0.97。
  • 该工具可从3D TBP生成热图展示光损伤程度,有望用于黑色素瘤风险预测。

核心结论

  • CNN可从3D TBP自动且可重复地报告个体光损伤表型,纳入风险预测模型可能有助于靶向监测建议。

研究设计与样本

  • 开发并验证光损伤光子数值量表,由两名皮肤科研究学生和两名非专业人员比较评分者间可重复性。
  • 采用多任务学习策略将色素沉着作为辅助任务,并与仅处理光损伤的单任务CNN比较。

主要发现

  • 非专业人员使用光子数值量表与皮肤科研究学生评分达到高度至几乎完美一致(κ=0.77-0.83)。
  • 多任务CNN较单任务CNN的ROC-AUC从0.91提升至0.96(P<0.01)。
  • 轻度光损伤分类准确率从0.96提升至0.98(P=0.04),中度从0.85提升至0.92(P<0.01),重度从0.97提升至0.99(P<0.01)。
  • 各身体部位准确率范围为0.86-0.92,外部验证集ROC-AUC为0.93,轻、中、重度准确率分别为0.97、0.85、0.97。

关键数值与推荐

  • 光子数值量表评分者间一致性κ=0.77-0.83。
  • 多任务CNN的ROC-AUC为0.96,单任务CNN为0.91(P<0.01)。
  • 外部验证集ROC-AUC为0.93,轻度准确率0.97、中度0.85、重度0.97。

临床怎么用

  • 可将CNN标记的光损伤以热图形式显示在3D TBP患者虚拟人像上,供临床解读。
  • 该评估可能纳入风险预测模型,以辅助制定针对性监测建议。

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: Early detection of melanoma presents a major public health challenge. Growing evidence supports targeted surveillance of individuals at high risk identified using risk stratification. Skin photodamage is the primary environmental risk factor for melanoma; however, it is inconsistently captured and often relies on self-reporting or subjective observations, resulting in poor reproducibility. The increasing use of total-body photography (TBP) in clinical skin examinations, combined with advances in artificial intelligence technology, presents new opportunities for automated skin assessment of ultraviolet damage. OBJECTIVES: To develop a clinical photonumeric scale for photodamage assessment, use the scale to build a dataset of annotated image tiles, and train a convolutional neural network (CNN) to automate photodamage assessment from three-dimensional (3D) TBP. METHODS: Our photonumeric scale was validated for assessing photodamage and pigmentation from 3D TBP by comparing inter-rater reproducibility between two dermatology research students and two lay people. A total of 24 720 cutaneous image tiles from 56 individuals at high risk and 51 at population risk for melanoma were annotated. Annotated images were used to train a CNN with a multi-task learning (MTL) strategy that incorporated pigmentation as an auxiliary task to increase the performance for photodamage. The MTL-CNN was compared with a single-task CNN that considered photodamage in isolation. RESULTS: Lay people achieved substantial-to-almost perfect agreement with dermatology research students using the photonumeric scale (κ = 0.77-0.83). The MTL-CNN design improved performance compared with the single-task CNN, with receiver operating characteristic area under the curve (ROC-AUC) increasing from 0.91 to 0.96 (P < 0.01). Class-specific accuracy improved for mild (0.96 to 0.98; P = 0.04), moderate (0.85 to 0.92; P < 0.01) and severe (0.97 to 0.99; P < 0.01) photodamage categories, and was maintained across each body site (range 0.86-0.92). Accuracy was reproduced in an external validation set with a ROC-AUC of 0.93, including class-specific accuracies of 0.97 for mild, 0.85 for moderate and 0.97 for severe photodamage. An interface was developed to display CNN-labelled photodamage as heatmaps on 3D TBP patient avatars for clinical interpretation. CONCLUSIONS: Our CNN provides a novel tool to automatically and reproducibly report an individual's photodamage phenotype from 3D TBP. Incorporating this assessment into risk prediction models may inform targeted risk prediction facilitating surveillance recommendations. Sun damage to the skin is the main environmental risk factor for melanoma. However, current risk assessment tools rarely use sun damage as there is no simple or reliable way to measure it across the whole body. Current methods usually need people to report past sun exposure. This includes how much time they spent outdoors or how many severe sunburns they had as a child. Such estimates are often unreliable because they depend on memory. To address this difficulty, we developed a visual scale to help assess sun damage on skin. The study took place in Australia. We tested how reliable the scale was. We compared how clinicians and lay people rated the same skin images. We found high similarity of rating. We then recruited the lay people to assess (‘label’) a large dataset of skin images (24,750 image tiles). This dataset was made from photos using a special three-dimensional (3D) body scanner that uses 2D skin photos to create a 3D map (‘avatar’) of a person. The dataset included photos from 107 volunteers. From each volunteer, over 200 image tiles were collected for the dataset. The labelled image tiles were then used to ‘train’ an artificial intelligence (AI) system to automatically detect sun damage. The AI system reported sun damage with very high accuracy. It performed better when trained to recognize both sun damage and skin pigmentation, versus sun damage alone. We also carried out ‘external validation’ of the AI system. This means we tested it using images from a group of people not part of the original set. The AI results are shown using colour-coded areas on the 3D human avatars (‘heatmaps’). These show areas of mild, moderate or severe sun damage. Overall, we showed our AI system provides an automated way to measure sun damage across the entire body using 3D total-body photography. This tool could be used to improve skin cancer risk assessment and guide more personalized skin cancer screening.

来源:Br J Dermatol,查看原文。版权归原作者所有。

引用

  1. Kahler S, Yan S, Mothershaw A, et al. Automated classification of site-specific cutaneous photodamage using a convolutional neural network and three-dimensional total body photography. Br J Dermatol. 2026. PMID: 41370219. doi:10.1093/bjd/ljaf516

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。