台灣、日本、法國不同創投機構對AI醫學影像軟體醫療器材新創公司的投資評估偏好分析

Abstract

研究背景：人工智慧（Artificial Intelligence, AI）在醫學影像領域已展現出顯著的應用潛力，有望提升診斷效率、緩解全球醫療人力短缺與資源分配不均等問題。醫學影像的軟體醫療器材（Software as Medical Devices, SaMD）是 AI 醫療保健中最活躍的子領域，截至 2025 年占美國食品藥物管理局（U.S. Food and Drug Administration, FDA）SaMD核准項目裡的77%。然而自 2021年數位健康(Digital Health) 募資高峰過後，全球數位醫療創投投資大幅萎縮，台灣、日本、法國各自面臨不同卻相關的挑戰，然而至今尚無研究系統性比較三國創投對 AI 醫療影像新創公司（Startups）的評估偏好。

研究目的：本研究旨在比較台灣、日本與法國的創投機構（Venture Capital, VC）在評估 AI 醫學影像 SaMD 新創公司 Series A 投資時的偏好差異，並將差異轉化為新創跨境募資的具體策略指引。

研究方法：採用混合研究設計，結合結構式量化問卷調查（三國共 41 名專業投資人）與半結構式質化深度訪談（兩位台灣資深醫療器材投資人）。問卷涵蓋團隊能力、市場吸引力、產品與技術、財務可行性及策略契合度的五大評估維度，共 31項評估指標，並以 7 點李克特量表（7-point Likert-scale survey）衡量。質化部分則訪談兩位資深台灣醫材投資人，以脈絡方式呈現。

研究結果：研究獲得三項主要發現。第一，台灣投資人高度重視團隊誠信與商業模式擴張性；日本投資人強調市場規模與法規進入門檻；法國投資人則聚焦於產品成熟度與 AI 可解釋性。第二，研究發現明顯的亞洲與歐洲評估典範分歧：台灣與日本在 31 項標準中僅 3 項存在顯著差異，而台灣與法國、日本與法國之間分別有 14 項與 17 項顯著分歧，呈現出亞洲內部高度收斂、亞洲與歐洲系統性背離的格局。第三，質性訪談揭示市場評估邏輯已從「技術優先」轉向「商業模式」優先，醫院市場的實際滲透能力已成為決定競爭成敗的關鍵因素，而非演算法的技術優勢。

研究討論： 亞洲內部的收斂與亞歐之間的分歧反映了結構性現況：台灣與日本作為島國開放經濟體，將國際擴張視為達成規模的必要條件；法國則身處歐盟單一市場的調和環境中，可將鄰國視為國內市場的延伸。同時，實務工作者所觀察到的商業模式導向轉變顯示，取得法規許可與技術創新已不再構成可持續的競爭優勢，與策略性收購者的需求對接以及醫院臨床通路的掌握，正逐步取代演算法本身成為核心競爭力來源。 基於上述發現，本研究進一步提出三項假說作為市場策略建議：(1) 從靜態 2D 影像 AI 轉向3D動態、即時影像 AI；(2) 從醫療器材定位重新調整為非器材類臨床流程改良；(3) 從純軟體產品轉向軟硬整合方案。此外，本研究亦指出現行問卷工具的兩個範圍邊界，可作為未來改進方向：其一是需將醫院整合能力與保險給付途徑可行性納入為獨立評估項目；其二是基於代理式人工智慧（Agentic AI）的快速崛起，未來需將受訪者範圍由投資人擴展至醫院端與終端使用者。

研究結論：本研究為 AI 醫學影像新創公司提供跨國募資策略的實證依據，AI 醫療影像新創須依各市場評估典範調整募資策略，跨越亞歐界線時需根本性地重新定位。具體建議包括善用台日相似性以利區域擴張、針對法國市場強化技術驗證與臨床實證、面對 AI 影像市場成熟現況轉向商業模式創新。

Background. Artificial intelligence (AI) in medical imaging shows significant potential to improve diagnostic efficiency and address global workforce shortages. Medical imaging software as a medical device (SaMD) is the most active subsegment of AI healthcare, accounting for 77% of all FDA-approved SaMD products as of 2025. However, following the 2021 peak, global digital health venture capital investment has contracted sharply. Despite Taiwan, Japan, and France each facing distinct yet related challenges, no prior study has systematically compared how venture capital firms across these three countries evaluate AI medical imaging startups.

Objective. This study aims to compare the evaluation preferences of venture capital (VC) firms in Taiwan, Japan, and France when assessing AI medical imaging SaMD startups for Series A investment, and to translate these differences into concrete strategic guidance for cross-border fundraising.

Methods. A mixed-method research design was adopted, combining a structured quantitative questionnaire survey (41 professional investors across the three countries) with semi-structured qualitative in-depth interviews (two senior Taiwanese medical device investors). The questionnaire covered five evaluation dimensions: Team Capability, Market Attractiveness, Product and Technology, Financial Viability, and Strategic Fit, comprising 31 evaluation criteria in total, measured on a 7-point Likert scale. The qualitative component involved interviews with two senior Taiwanese medical device investors to provide contextual interpretation.

Results. Three principal findings emerged. First, Taiwanese investors place strong emphasis on team integrity and business model scalability; Japanese investors emphasize market size and regulatory barriers to entry; and French investors focus on product maturity and AI explainability. Second, a clear Asia and Europe paradigm divide was identified: Taiwan and Japan differed significantly on only 3 of the 31 criteria, whereas Taiwan vs. France and Japan vs. France pairs diverged on 14 and 17 criteria respectively, revealing a pattern of intra-Asian convergence and systematic Asia and Europe divergence. Third, qualitative interviews revealed that market evaluation logic has shifted from "technology-first" to "business-model-first", with hospital market penetration capability rather than algorithmic technical superiority emerging as the critical determinant of competitive success.

Discussion. The intra-Asian convergence and Asia–Europe divergence reflect underlying structural realities: Taiwan and Japan, as island open economies, treat international expansion as a near-prerequisite for achieving meaningful scale, whereas France operates within the harmonized environment of the EU single market and can treat neighboring countries as an extension of its domestic market. Meanwhile, the practitioner-observed shift toward business-model-led competition suggests that regulatory clearance and technological innovation alone no longer constitute sustainable competitive advantages, with alignment to strategic acquirer needs and access to hospital clinical channels progressively replacing the algorithm itself as the core source of competitive advantage. Building on these findings, three practitioner-derived strategic hypotheses are proposed for future investigation: (1) a shift from static 2D imaging AI toward dynamic, motion-based AI imaging; (2) repositioning from medical-device classification toward non-device clinical workflow optimization; and (3) re-orientation from pure software toward software-hardware integration. In the future, we should capture hospital integration capability and reimbursement-pathway feasibility as discrete evaluation dimensions, and the need to extend the respondent frame beyond investors to include hospital and end-user perspectives, particularly in light of the rapid emergence of agentic AI.

Conclusion. AI medical imaging startups must tailor their fundraising strategies to each market's evaluation paradigm, with fundamental repositioning required when crossing the Asia and Europe boundary. Specific recommendations include leveraging Taiwan and Japan similarities to facilitate regional expansion, strengthening technological validation and clinical evidence when targeting the French market, and pivoting toward business model innovation in response to the maturation of the AI imaging market.