Nat Rev Clin Oncol, IF: 34.106
Google’s lung cancer AI: a promising tool that needs further validation
Researchers from Google AI have presented results obtained using a deep learning model for the detection of lung cancer in screening CT images. The authors report a level of performance similar to, or better than, that of radiologists. However, these claims are currently too strong. The model is promising but needs further validation and could only be implemented if screening guidelines were adjusted to accept recommendations from black- box proprietary AI systems.
來自谷歌AI的研究人員展示了使用深度學習模型在篩查CT圖像中檢測肺癌的結果。作者報告說，它們的表現水平與放射科醫生相似，甚至更好。然而，這些論斷還為時過早。該模型很有前景，但需要進一步驗證，只有調整篩選指南，以接受黑箱專有人工智慧系統的建議，才能實現。
A black-box AI-based system that overrules well established clinical guidelines that prompt radiologists to base management decisions on the size and growth rate of nodules is unlikely to be quickly accepted.
一個基於人工智慧的黑箱系統不太可能很快被接受，因為它推翻了已經建立的臨床指南，這些指南使放射科醫生根據結節的大小和生長速度做出治療決定。
Thus, if and how Google intends to translate this tool, one of several it has developed in recent years in the field of medical image analysis, into a product currently remains to be seen. All of the solutions developed by Google AI so far are proprietary. In the Nature Medicine letter, the authors state that their code has 「dependencies on internal tooling, infrastructure and hardware, and its release is therefore not feasible. However, all experiments and implementation details are described in sufficient detail […] to allow independent replication」.
因此，谷歌是否以及打算如何推出這一工具（近年來在醫學圖像分析領域開發的幾種產品之一）目前還有待觀察。到目前為止，谷歌AI開發的所有解決方案都是專有的。在《自然醫學》雜誌的信中，作者指出他們的代碼「依賴於內部工具、基礎設施和硬體，因此其不大會發布」。然而，所有的實驗和實現細節都被描述得足夠詳細[…]，以允許獨立的複製。
This description, however, lacks many technical detailsthat might be crucial for obtaining a good level of performance. Therefore, we consider Google’s code availability statements as only paying lip service to the principle of reproducible science.
然而，這種描述缺乏許多技術細節，而這些細節對於獲得良好的性能水平可能是至關重要的。因此，我們認為谷歌的代碼可用性聲明只是口頭上支持可重複科學的原則。
（此文為轉載）