通过调整皮肤病状分布差异在临床环境中的差异来弥补AI泛化差距
Rajeev V Rikhye1, Aaron Loh1, Grace Eunhae Hong2
1Google Research, Mountain View, CA, USA.
EBioMedicine
|June 3, 2025
概括
人工智能 (AI) 模型可以通过重新校准或重新培训来适应新的临床环境,从而提高皮肤病例的诊断准确性. 这项研究表明,人工智能可以很好地从远程医疗到临床图像进行概括,有针对性的再培训可以提高性能.
科学领域:
- 医疗人工智能 医疗人工智能
- 皮肤病学AI 人工智能
- 临床人工智能概括
背景情况:
- 将人工智能 (AI) 模型推广到新的临床环境中存在重大挑战.
- 这项研究调查了皮肤病AI模型在不同图像类型中的强度和概括能力.
研究的目的:
- 评估皮肤学AI模型从远程医疗病例到临床环境的通用性,包括患者提交的 (PAT) 和临床医生拍摄的 (CLIN) 照片.
- 了解影响人工智能和皮肤科医生在新临床环境中的表现的因素.
主要方法:
- 来自22家诊所的2500个未见病例 (PAT和CLIN) 的回顾性队列研究 (2015年11月至2021年1月).
- 主要结局:与参考诊断相比,前三位精度 (AI和皮肤科医生) 较高.
- 分析与人工智能错误相关的人口因素和条件类别;重新抽样以匹配人工智能开发数据集分布.
主要成果:
- 对于CLIN (74%的Top-3精度) 和PAT (71%) 图像,AI的表现相似;皮肤科医生在PAT (87%) 和CLIN (79%) 上表现更好.
- 人口因素与错误无关;特定条件类别与AI错误有关.
- 再抽样改善了AI准确度 (CLIN 84%,PAT 79%) 和皮肤科医生准确度的调整 (CLIN 77%,PAT 89%).
- 微调策略,包括端到端和分类层再培训,在不重新采样的情况下实现了高精度 (83-86%).
结论:
- 人工智能模型可以通过重新校准或针对罕见疾病和再培训的有针对性的数据采集来有效地适应新环境.
- 重新训练最终的分类层或端到端的微调显示了相似的性能.
- 这些发现表明,在各种临床环境中部署AI的实际策略.
相关概念视频
01:52Regression Toward the Mean
6.3K
Regression toward the mean (“RTM”) is a phenomenon in which extremely high or low values—for example, and individual’s blood pressure at a particular moment—appear closer to a group’s average upon remeasuring. Although this statistical peculiarity is the result of random error and chance, it has been problematic across various medical, scientific, financial and psychological applications. In particular, RTM, if not taken into account, can interfere when...
6.3K
01:15Receiver Operating Characteristic Plot
106
A ROC (Receiver Operating Characteristic) plot is a graphical tool used to assess the performance of a binary classification model by illustrating the trade-off between sensitivity (true positive rate) and specificity (false positive rate). By plotting sensitivity against 1 - specificity across various threshold settings, the ROC curve shows how well the model distinguishes between classes, with a curve closer to the top-left corner indicating a more accurate model. The area under the ROC curve...
106
01:21Testing a Claim about Mean: Unknown Population SD
3.4K
A complete procedure of testing a hypothesis about a population mean when the population standard deviation is unknown is explained here.
Estimating a population mean requires the samples to be approximately normally distributed. The data should be collected from the randomly selected samples having no sampling bias. There is no specific requirement for sample size. But if the sample size is less than 30, and we don't know the population standard deviation, a different approach is used;...
Estimating a population mean requires the samples to be approximately normally distributed. The data should be collected from the randomly selected samples having no sampling bias. There is no specific requirement for sample size. But if the sample size is less than 30, and we don't know the population standard deviation, a different approach is used;...
3.4K