本計畫目的是想以深度學習的方式改進生物支架的外形品質。生物支架材料大致可分為天然材料跟人工合成兩大類。目前最為廣泛使用的天然材料有膠原蛋白、海藻酸鈉、明膠及殼聚醣等。由於上述材料係從動、植物取得，每批量配置出的生物墨水總有些許差距。生物列印之製造參數大多彼此相依，因此很難藉由理論推導出有效的預測公式。往往只能使用試誤法，導致需耗費大量時間在調整製造參數來達到預設的尺寸目標。多層感知器為深度學習的一支，其可透過大量過往實驗數據訓練出一個有效預測模型，在新的數據輸入此一模型後進而可預測結果，而這個能力可幫助改進生物支架的外形品質。本計畫規劃三年。第一年著重在建構深度學習之軟、硬體及將生物列印系統的各式不同感測器訊號與深度學習系統連接、分類、正規化，最後再萃取成供深度學習訓練的特徵。第二年著重發展線上自動量測支架外形資料，用以作為深度學習訓練的標記。之後整合第一年成果成為可自動收集訓練資料、自動訓練之智慧型生物列印系統。本計畫擬以殼聚醣、明膠及海藻酸鈉為支架材料，擬於第三年開始進行大量支架製作。所得的實際支架外形尺寸將與預測結果比對，並進行模型修正與改良。期望透過此三年計畫，能發展出一套能使用天然材料製作高品質支架的生物列印系統。 ;The purpose of this project is to improve the appearance quality of the bio-scaffold by utilizing deep learning. The materials of bio-scaffolds can be roughly divided into two categories: natural and synthetic materials. The most widely used natural materials are collagen, sodium alginate, gelatin and chitosan. As these natural materials are obtained from animals and plants, there is always a slight difference between the bio-inks in each batch. Most of the fabrication parameters have dependence with each other, it is difficult to derive effective predictive formulas for manufacturing scaffolds. Often can only use the trial and error method, resulting in the need to spend a lot of time to adjust the fabrication parameters for achieving the preset target. Multilayer perceptron is a kind of Deep learning, and it can train a model through a large amount of historical data, and predictable results obtain after inputting new data into the model. The capability can help to improve the appearance quality of the bio-scaffold.The project is planned for a three-year period, with the first year focusing on the construction of deep learning software and hardware, as well as, various types of sensor signals from bio-printing system will connect to deep learning system, and then classify and regularize. These regularized data will extract into feature for deep learning. The second year will focus on the development for online automatic measurement of scaffold shape, as the labels in depth learning. This work will take together with the results from first year, and bring an intelligent bio-printing system that automatically collects training data and automates training. The project plans to use chitosan, gelatin and sodium alginate as scaffold materials, and to produce a large number of scaffolds in the third year. The actual dimensions of the scaffold fabricated will compare with the predicted results, and the model of the deep learning will revise and improve. Through this three year project, this project would develop a bio-printing system capable of producing high quality scaffolds using natural materials.
