在現代醫學診斷領域中，內視鏡技術的創新發展一直是研究重點。近期，國立台灣大學醫學院駱遠教授研究團隊在這一領域取得重大突破，成功開發出結合人工智慧與奈米光學的革命性內視鏡顯微系統，為醫學影像技術開創了新紀元。

超穎介面微型光學元件的發展

超穎透鏡是一種由奈米尺度的人工結構陣列組成的光學元件，能夠精確操控入射光的相位、振幅和偏振態。這種特殊的光學結構使其能夠突破傳統光學元件的衍射極限，實現超高解析度成像。超穎透鏡的厚度通常只有波長量級，遠小於傳統光學透鏡，這種超薄特性使其特別適合應用於微型化醫療器材。在本研究中，團隊開發的超穎透鏡不僅保持了超薄特性，更通過特殊的結構設計實現了可變焦功能，為內視鏡系統的革新提供了關鍵技術支撐。

突破傳統限制的創新設計

傳統內視鏡技術雖在醫療診斷中扮演重要角色，但在影像品質、操作便利性及診斷效率方面仍面臨諸多挑戰。特別是在調整焦距時，常需依賴複雜的光學結構和機械調整，不僅限制了設備的微型化發展，也增加了操作的複雜度。此外，獲取高解析度影像時所需的較長曝光時間，更為活體組織的快速成像帶來挑戰。

研究團隊透過開發超穎介面光學元件，創造出多項創新技術，其中最具代表性的是可變焦光學超穎透鏡。這項技術通過精確控制光的相位和振幅，實現了動態焦距調整，且無需複雜的機械結構。團隊更進一步利用光學擴散片結合客製化的內視鏡設計，使系統能夠從原始影像中快速重建出高品質的光學切片影像。超穎透鏡的另一個重要特點是其可以在單一平面上實現複雜的光場調控，這使得系統能夠在不增加體積的情況下，實現多功能光學成像。

AI賦能的新世代顯微系統

在人工智慧的加持下，研究團隊的系統展現出前所未有的突破性能力。新開發的智慧可變焦技術結合超穎介面與ResNet技術，實現了高解析度即時成像，特別適用於深層組織觀察。系統採用深度學習U-net架構的量化相位對比技術，提供了精確的細胞形態與病理分析能力。超穎透鏡的奈米結構設計與AI演算法的結合，不僅提升了系統的成像質量，還大幅提高了圖像處理的效率。

技術創新與臨床應用

這項創新系統的核心優勢在於成功整合了奈米光學與人工智慧技術。超穎透鏡的可調控性與AI技術的智慧化完美結合，創造出一個高效能、微型化的診斷平台。系統能即時辨識惡性組織細胞，協助醫師及早發現病兆，其應用範圍涵蓋從胃腸道檢查到神經外科手術等多個領域。

研究成果還包括突聚焦光束設計的超穎介面雷射手術系統，這項技術特別適用於需要高精度操作的皮膚科、眼科及神經外科手術，在癌症診斷與腦部研究中展現出巨大潛力。通過超穎透鏡的精確光場調控，系統可以實現對特定深度組織的選擇性照明與成像，大大提升了診斷的準確性。

未來發展與挑戰

駱遠教授指出，未來研發將著重於進一步優化超穎透鏡的設計與製程，並加強AI演算法的即時處理能力。研究團隊也將探索新型奈米材料與結構，以提升超穎透鏡的光學性能。同時，團隊將致力於解決人工智慧訓練與即時重建對計算資源的高需求問題，期望通過整合專用AI晶片，實現系統的進一步微型化與效能提升。

這項結合超穎介面技術、人工智慧與生物醫學的創新研究，展現了奈米光學與AI技術在醫學診斷領域的巨大潛力。超穎透鏡的獨特優勢與AI技術的智能化處理相得益彰，為醫學影像技術開創了新的發展方向。隨著技術持續優化與臨床應用的推進，這項突破性技術必將為醫療診斷帶來革命性的改變，造福更多患者。

（114年度TBF學術講座、台大醫學院醫療器材與醫學影像研究所　駱遠教授）

【Vari-Focal Metalens Integrated with AI for HiLo Structure Illumination Endoscopic Microscopy】

In the field of modern medical diagnostics, innovative development in endoscopic technology has consistently been a research priority. Professor Yuan Luo's research team at the National Taiwan University College of Medicine has recently achieved a significant breakthrough in this field, successfully developing a revolutionary endoscopic microscopy system that combines artificial intelligence and nano-optics, pioneering a new era in medical imaging technology.

 

Development of Metalens Miniature Optical Components

Metalens is an optical component composed of nanoscale artificial structure arrays, capable of precisely manipulating the phase, amplitude, and polarization of incident light. This special optical structure enables it to overcome the diffraction limit of traditional optical components, achieving super-resolution imaging. The thickness of metalens is typically only at the wavelength scale, significantly thinner than traditional optical lenses, making it particularly suitable for miniaturized medical devices. In this research, the team's developed metalens not only maintain their ultra-thin characteristics but also achieve variable focus functionality through special structural design, providing crucial technical support for endoscopic system innovation.

 

Innovative Design Breaking Traditional Limitations

While traditional endoscopic technology plays a crucial role in medical diagnosis, it still faces numerous challenges in image quality, operational convenience, and diagnostic efficiency. Particularly when adjusting focal length, it often relies on complex optical structures and mechanical adjustments, which not only limits equipment miniaturization but also increases operational complexity. Additionally, the longer exposure time required for high-resolution imaging presents challenges for rapid imaging of living tissues.

The research team has created multiple innovative technologies through the development of metasurface optical components, with the most representative being the variable-focus optical metalens. This technology achieves dynamic focal length adjustment through precise control of light phase and amplitude, without requiring complex mechanical structures. The team further utilizes optical diffusers combined with customized endoscope design, enabling the system to quickly reconstruct high-quality optical sectioning images from raw images. Another important feature of metalens is its ability to achieve complex light field control on a single plane, allowing the system to achieve multifunctional optical imaging without increasing volume.

 

AI-Empowered New Generation Microscopy System

Enhanced by artificial intelligence, the team's system demonstrates unprecedented breakthrough capabilities. The newly developed intelligent variable-focus technology combines metasurface interface with ResNet technology, achieving high-resolution real-time imaging particularly suitable for deep tissue observation. The system employs quantitative phase contrast technology based on deep learning U-net architecture, providing precise cellular morphology and pathological analysis capabilities. The combination of metalens nanostructure design and AI algorithms not only enhances system imaging quality but also significantly improves image processing efficiency.

 

Technological Innovation and Clinical Applications

The core advantage of this innovative system lies in its successful integration of nano-optics and artificial intelligence technology. The controllability of metalens perfectly combines with AI technology's intelligence, creating a high-performance, miniaturized diagnostic platform. The system can identify malignant tissue cells in real-time, helping physicians detect early signs of disease, with applications ranging from gastrointestinal examinations to neurosurgery.

Research outcomes also include a metasurface laser surgical system with a focused beam design, particularly suitable for high-precision operations in dermatology, ophthalmology, and neurosurgery, showing great potential in cancer diagnosis and brain research. Through precise light field control of metalens, the system can achieve selective illumination and imaging of specific depth tissues, significantly improving diagnostic accuracy.

 

Future Development and Challenges

Professor Luo points out that future research and development will focus on further optimizing metalens design and manufacturing processes while enhancing AI algorithm real-time processing capabilities. The research team will also explore new nanomaterials and structures to improve metalens optical performance. Meanwhile, the team will work on resolving the high computational resource demands of AI training and real-time reconstruction, hoping to achieve further system miniaturization and performance enhancement by integrating specialized AI chips.

This innovative research combining metasurface technology, artificial intelligence, and biomedicine demonstrates the enormous potential of nano-optics and AI technology in medical diagnostics. The unique advantages of metalens complement AI technology's intelligent processing, pioneering new directions for medical imaging technology. As technology continues to optimize and clinical applications advance, this breakthrough technology will bring revolutionary changes to medical diagnosis, benefiting more patients.

(2025 TBF Chair in Biotechnology, Professor Yuan Luo)