基於擠出生物列印(EBB)是生物列印領域中有效率、簡單且具成本效益的技術，其能夠製造多孔且複雜的三維支架結構，廣泛應用於組織修復與再生。然而，由於生物材料墨水大多屬於非牛頓流體，且不同墨水的流變性質及熱性質存在差異。這些流體的多樣性及高複雜性使支架製造過程需依賴試誤法以找尋合適的列印參數配置，使支架列印品質及重現性存在疑慮。 本研究將流量感測技術整合於EBB系統中，並基於捕捉所有列印參數的時間序列資料，取代傳統的固定列印參數設置方式，以提升列印過程的可監控性。為了EBB的適應性，我們在不同的參數配置下，實驗了三種不同性質的生物材料墨水。透過發展不同的深度學習模型，以評估模型在列印過程中對於多種墨水的適應能力。此方法係實現了對墨水不敏感之支架線徑預測。 為了提升黑盒子模型的透明度與其預測結果的可信度。我們使用兩種可解釋人工智慧方法分析深度學習模型的決策過程，以識別支架列印過程中關鍵的列印參數及時間序列特徵。這一分析不僅提高了模型的可靠性，關鍵特徵的識別也為生物列印過程帶來新的見解，為未來提升生物列印效率及品質奠定了基礎。 鑑於流量被識別為生物列印過程中影響支架品質關鍵的參數。我們評估了流量計在生物列印中的量測規範，並基於粒子影像測速技術(PIV)發展了新的流量應用，以量測生物材料墨水的擠出流量。透過數學理論模型、精密天秤與PIV量測結果的比較，我們驗證了PIV在量測生物材料墨水中的可靠性，為提高生物列印的品質及重複性提供新的方法及應用。 ;Extrusion-based bioprinting is recognized as an efficient, simple, and cost-effective technique in bioprinting. It is capable of fabricating porous and complex three-dimensional scaffold structures widely utilized in tissue repair and regeneration. However, most biomaterial inks are non-Newtonian fluids, with significant variations in their rheological and thermal properties. The inherent diversity and complexity of these inks often result in the scaffold manufacturing process relying on several trial-and-error approaches to determine suitable printing parameter configurations, raising concerns about the quality and reproducibility of the printed scaffolds. This study integrates flow sensing into an extrusion-based bioprinting (EBB) system, replacing the conventional approach with fixed printing parameters by capturing time-series data of all printing parameters, enhancing the monitorability of the printing process. To improve the adaptability of EBB, experiments were conducted using three different biomaterial inks with various parameter configurations. Multiple deep-learning models were developed to evaluate their adaptability to different inks during the printing process. This approach enabled ink-insensitive predictions of the scaffold linewidth. To improve the transparency of black box models and enhance the credibility of their prediction results, two explainable artificial intelligence methods were employed to analyze the decision-making process of the deep learning model. This analysis aimed to identify key features of the printing parameters and time series during the scaffold printing process. The results enhanced the reliability of the model, and the identification of these key printing parameters also provides new insights into the bioprinting process, laying the foundation for improving the efficiency and quality of future bioprinting applications. Flow rate was identified as a critical parameter that significantly influenced the quality of scaffolds during the bioprinting process. To assess the specifications of flow meters used in bioprinting, a novel flow measurement application based on Particle Image Velocimetry (PIV) technology was developed to measure the flow rate of biomaterial ink during printing. The reliability of PIV in measuring biomaterial ink was validated through comparisons with mathematical theoretical models, precision scales, and PIV measurement results. This study provided novel approaches and applications for enhancing bioprinting quality and reproducibility.
