| 摘要: | 文獻背景: 如何獲得可靠的食物份量評估一直具有相當的挑戰性。傳統飲食評估法，如秤重紀錄法，耗時費力，對受試者負擔大，24小時飲食回憶法也會受營養師訪問技巧以及受試者記憶力影響，因此，研究常會搭配塑膠食物模型、或食物圖像集等輔助工具。食物圖像集有著省時、便利、可涵蓋大量食物的優點，適合運用於大型研究。近年來科技進展飛速，在建立數位3-dimensional (D)模型的攝影測量法(Photogrammetry)上有所突破，展現出能將數位3D食物模型大量製作為食物影像圖集的潛力。目前臺灣並沒有已開發完成的數位食物影像圖集或相關線上平台。
目的: 本研究的目的為: 開發數位食物份量評估輔助工具：eDiet24–臺灣數位影像食物圖集，以圖像飲食評估法(image-based dietary assessment, IBDA) 驗證以eDiet24輔助之24小時飲食回憶的效度，為未來開發線上自動化食物影像圖集飲食回憶平台做準備。
方法: 透過於臺灣北區進行之營養調查資料庫，歸納出臺灣人常吃的食物。所有2D食物圖像皆由45°與85°角拍攝；3D食物圖像則以攝影測量技術進行拍攝。食物份量呈現之間隔，依據臺灣衛生福利部食品藥物管理署之食品營養成分資料庫進行設計。後續招募41位健康成年人，以實驗室自行開發的”Formosa FoodApp” App拍照紀錄一日飲食，進行效度研究。
結果: 我們總共攝影664種食物的2D及3D食物圖像集。以eDiet24之2D或3D圖像集輔助之24小時飲食回憶，此兩者產生之一日飲食評估皆與食物圖像飲食評估法無顯著差異，然而，2D與3D圖像集輔助之24小時飲食回憶彼此間具有顯著差異(p < 0.05)，其中2D發生高估的情形。皮爾森相關係數顯示三種飲食評估方法間皆具有中等至強相關性(r > 0.6; p < 0.001)，其中3D輔助之24小時飲食回憶與圖像飲食評估法相關性較高。在食物品項的子分析中顯示，2D與3D食物圖像集的差異主要來自於白飯、麵條、麵包、豆腐四種食物，統計上呈現顯著差異(p < 0.05)。
總結: 前導實驗顯示，以eDiet24之2D或3D圖像集輔助之24小時飲食回憶，其效度與圖像飲食評估法無統計差別。但eDiet24仍需進一步的擴大拍攝食物種類，並修正拍攝條件以增加某些特定食物的準確性。
Background and Aims: Obtaining reliable information on food portion size is crucial in dietary assessments. Digital photographic food atlases, as portion size estimation aids (PSEAs), have advantages in terms of affordability, accessibility, and representation of various portion sizes. Recent advancements in 3-dimensional (D) photogrammetry technology offer great potential for using digital 3D models as PSEAs in photographic food atlases. Currently, there is no digital photographic food atlas that has been developed in Taiwan. The aim of this study was to develop eDiet24, a digital photographic food atlas of commonly consumed Taiwanese foods, in both 2D (Atlas_2D) and 3D (Atlas_3D), and validate eDiet24 among healthy adults.
Methods: Study 1: Development of eDiet24: We identified commonly consumed Taiwanese foods through previous dietary surveys conducted in northern Taiwan. The interval of portion size interval varied between 0.2 – 1.0 exchange based on the Nutrient Composition Database for Foods in Taiwan. 2D images were taken at 45° and 85°. 3D food models were generated by 3D photogrammetry technic. Study 2: Validation: We validated eDiet24 as PSEAs in assisting 24-hour dietary recalls (24-HDR) among healthy adults against image-based dietary assessment (IBDA). Paired t-test, repeated One-way ANOVA and Pearson’s correlation coefficient were used to assess agreement between methods.
Results: Study 1 Development: In total, 664 food items of Atlas_2D and 490 food items of Atlas_3D were included in eDiet24. Food items were caterogized into 13 major food groups and 57 sub-groups. Study 2 Validation: A total of 41 healthy adults were included. Both nutrient intakes assessed from 24-HDR-assisted Atlas_2D and Atlas_3D were in agreement with IBDA. The Pearson correlation coefficient (r) confirmed moderate to strong correlations between all three methods (all r > 0.6; p < 0.001). However, a significant difference (p < 0.05) was observed between 24-HDR-assisted Atlas_2D and 24-HDR-assisted Atlas_3D, with Atlas_2D tend to overestimate nutrient intake compared to Atlas_3D. Certain foods, including rice, noodles, tofu, and bread, were found to significantly contribute to the differences observed between Atlas_2D and Atlas_3D when compared to the IBDA method.
Conclusion: eDiet24 as PSEAs in assisting 24-HDR demonstrated comparability with the validated method, IBDA. This study emphasizes the need for further development of the eDiet24 digital photographic food atlas to enhance accuracy, particularly in specific food items. |
