Microsoft GigaTIME 開源 AI 癌症診斷：10 美元病理切片變 2000 美元免疫螢光掃描，4000 萬細胞訓練、24 種癌症通吃 | Microsoft GigaTIME Open-Source AI Cancer Tool: Turns $10 Slides Into $2000 Immune Scans Across 24 Cancer Types

By Kit 小克 | AI Tool Observer | 2026-04-14

🇹🇼 Microsoft GigaTIME 開源 AI 癌症診斷：10 美元病理切片變 2000 美元免疫螢光掃描，4000 萬細胞訓練、24 種癌症通吃

Microsoft GigaTIME 是微軟研究院與 Providence 醫療系統、華盛頓大學聯手開發的多模態 AI 模型，能把一張 10 美元的標準病理切片，轉換成原本要價 2000 美元以上的多重免疫螢光（mIF）影像。2026 年 4 月正式上架 Azure AI Foundry Labs 和 Hugging Face，完全開源。

GigaTIME 是什麼？為什麼重要？

GigaTIME 是一套多模態 AI 系統，能將醫院最常見的 H&E 染色病理切片，自動轉換成涵蓋 21 種蛋白質通道的虛擬 mIF 影像。傳統 mIF 檢測一次要 2000 美元以上，而且需要專門設備和技術人員。GigaTIME 讓任何有標準切片的醫院都能做到類似分析，大幅降低精準腫瘤學的門檻。

GigaTIME 的訓練規模有多大？

微軟用了 Providence 醫療系統的 4000 萬個細胞配對數據來訓練 GigaTIME，涵蓋 H&E 和 mIF 的對應影像。模型被應用到 14,256 名癌症患者、51 家醫院、超過 1000 間診所的數據上，生成約 30 萬張虛擬 mIF 影像，橫跨 24 種癌症類型和 306 種癌症亞型。

GigaTIME 發現了什麼？

透過虛擬 mIF 分析，GigaTIME 發現了 1,234 個具統計顯著性的關聯，連結腫瘤免疫細胞狀態與生物標記、分期、存活率等臨床指標。這些發現經過 TCGA（癌症基因組圖譜）10,200 名患者的獨立驗證。這代表 AI 不只是「看得到」，而是真的能幫醫生找到以前找不到的規律。

GigaTIME 怎麼取得？

2026 年 4 月，GigaTIME 正式上架 Azure AI Foundry Labs 和 Hugging Face，完全開源。研究人員和開發者可以直接下載模型、上傳自己的 H&E 切片進行分析。微軟也提供了完整的技術文件和 API，讓整合到現有醫療系統變得更容易。

這對癌症研究有什麼影響？

最大的影響是民主化。以前只有頂尖醫學中心才能負擔 mIF 檢測，現在任何醫院只要有標準病理切片就能用 GigaTIME 做虛擬分析。這對發展中國家和偏鄉醫院尤其重要。論文發表在頂級期刊《Cell》上，學術可信度夠高。不過要注意，虛擬 mIF 和真正的 mIF 還是有差距，目前比較適合研究用途，臨床應用還需要更多驗證。好不好用，試了才知道。

🇺🇸 Microsoft GigaTIME Open-Source AI Cancer Tool: Turns $10 Slides Into $2000 Immune Scans Across 24 Cancer Types

Microsoft GigaTIME is a multimodal AI model developed by Microsoft Research, Providence Health, and the University of Washington. It transforms standard $10 H&E pathology slides into virtual multiplex immunofluorescence (mIF) images that would normally cost over $2,000 per sample. As of April 2026, GigaTIME is fully open-sourced on Azure AI Foundry Labs and Hugging Face.

What Is GigaTIME and Why Does It Matter?

GigaTIME is a multimodal AI system that converts routine H&E stained pathology slides into virtual mIF images covering 21 protein channels. Traditional mIF testing costs over $2,000 per sample and requires specialized equipment. GigaTIME democratizes this capability, allowing any hospital with standard slides to perform similar analyses.

How Was GigaTIME Trained?

Microsoft trained GigaTIME on 40 million paired cells from Providence Health, covering both H&E and mIF imaging data. The model was applied to 14,256 cancer patients across 51 hospitals and over 1,000 clinics, generating approximately 300,000 virtual mIF images spanning 24 cancer types and 306 subtypes.

What Did GigaTIME Discover?

The virtual mIF analysis uncovered 1,234 statistically significant associations linking tumor immune cell states with clinical attributes including biomarkers, staging, and patient survival. These findings were independently validated on 10,200 patients from The Cancer Genome Atlas (TCGA), confirming the model produces clinically meaningful insights.

How to Access GigaTIME?

As of April 2026, GigaTIME is available on Azure AI Foundry Labs and Hugging Face as a fully open-source model. Researchers can download the model, upload their own H&E slides, and run virtual mIF analysis. Microsoft provides comprehensive documentation and APIs for integration into existing medical workflows.

What Does This Mean for Cancer Research?

The biggest impact is democratization. Previously, only top medical centers could afford mIF testing. Now any hospital with standard pathology slides can run virtual analysis through GigaTIME. This is especially significant for developing countries and rural hospitals. The research was published in Cell, one of the most prestigious scientific journals. However, virtual mIF is not yet a replacement for actual mIF in clinical settings — more validation is needed before clinical deployment.

Sources / 資料來源

常見問題 FAQ

GigaTIME 是什麼？

GigaTIME 是微軟研究院開發的多模態 AI 模型，能將標準 H&E 病理切片轉換成涵蓋 21 種蛋白質通道的虛擬多重免疫螢光（mIF）影像，大幅降低精準腫瘤學的檢測成本。

GigaTIME 能省多少錢？

傳統 mIF 檢測一次要 2000 美元以上，而 GigaTIME 只需要一張 10 美元的標準 H&E 病理切片就能生成虛擬 mIF 影像，成本降低 200 倍。

GigaTIME 支援哪些癌症？

GigaTIME 涵蓋 24 種癌症類型和 306 種癌症亞型，訓練數據來自 51 家醫院的 14,256 名患者。

GigaTIME 在哪裡下載？

GigaTIME 已完全開源，可在 Azure AI Foundry Labs 和 Hugging Face 上免費下載使用。

GigaTIME 能用在臨床診斷嗎？

目前 GigaTIME 比較適合研究用途，虛擬 mIF 和真正的 mIF 仍有差距，臨床應用還需要更多驗證。論文已發表在頂級期刊 Cell 上。

延伸閱讀 / Related Articles

AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

Stanford 研究登上《Science》：11 個 AI 模型有 47% 機率說你對，即使你錯了 | Stanford Study in Science: AI Models Validate Harmful Behavior 47% of the Time — Sycophancy Is a Real Problem

3月 28, 2026

閱讀完整內容

搜尋此網誌

AI小貼士