aetherAI’s initiative in a whole-slide training approach is globally recognized

Updated: Dec 17, 2021

Feb. 2021 published by Commercial Times


The article, An annotation-free whole-slide training approach to pathological classification of lung cancer types using deep learning in the international journal, co-authored by Taipei Medical University and aetherAI was published by Nature Communications. It recognizes the proposed method breakthrough, which is ahead of other well-known AI brands, and solves pain points in digital pathology AI development.


Deep learning for digital pathology is hindered by the extremely high spatial resolution of whole-slide images (WSIs). Most studies have employed patch-based methods, which often require detailed annotation of image patches. This typically involves laborious free-hand contouring on WSIs. To alleviate the burden of such contouring and obtain benefits from scaling up training with numerous WSIs, it develops a method for training neural networks on entire WSIs using only slide-level diagnoses.


Fig. 4, both the MIL model and whole-slide model could discover representative information, which was highlighted by heatmaps after iteratively learning from slide-level diagnosis.

Furthermore, the method demonstrates higher classification performance than multiple-instance learning as well as strong localization results for small lesions through class activation mapping. It is expected that this proposed method can realize AI applications empowered by digital pathology development.


Full article


雲象科技今日發表與臺北醫學大學合作的數位病理AI共同研究成果登上Nature Communications,此突破性的研究採用零細節標註零分割方法,得以將未經切割的病理全玻片影像直接用於訓練AI模型,大幅節省醫師專業人力,技術超越美國知名病理AI公司Paige,成功解決數位病理AI開發痛點。


雲象科技執行長葉肇元醫師表示,從建立AI技術之初就已預料到,病理AI開發所需的大量標註工作將會成為此領域發展之重大瓶頸。因此雲象科技三年來持續探索直接以超高解析度影像,在不需細節標註的情況下即可訓練AI的解決方案。


此項發表的重要性在於,數位病理全玻片影像,不再需要影像切割及費時的細節標註,即可用來訓練AI做癌症辨識。這項突破性的技術,可以讓醫院充分應用過去所累積的大量玻片資料,進行AI研發。可以預見此技術將為醫療AI尖端技術帶來嶄新的局勢,奠定臺灣於全球AI發展上的領先地位。