基於貝氏曲線的果蠅腦模型三維特徵平均與變形研究

A Bezier curve based averaging and local deformation method for drosophila brain model

孫健袁(Chien-Yuan Sun)

指導教授 ： 陳永昌

國立清華大學/電機資訊學院/電機工程學系/碩士(2009年)

摘要

為了要解開人類學習和記憶的謎題，人腦的機制一直是科學家所致力的研究目標，但是人腦的結構頗為複雜龐大，研究其中神經網路的互相連結關係非常困難。果蠅的腦部被發現具有和人類類似的能力：記憶與學習。果蠅腦的獨立結構和人腦雖然有很多相異處，但是基因的控制卻和人類相當類似，果蠅大腦結構比人腦簡單許多，因此其結構和神經傳導的關係也成為目前腦科學研究團隊目前致力於研究的課題之一。
為了研究果蠅的腦部結構和機制，腦科學中心利用共軛焦顯微鏡與染色的方式得到腦殼的模型影像切片，配合影像切割技術，再將其重建回果蠅腦殼模型。但是在個體發育程度不同的情況下，每一隻果蠅的腦部形狀也會有所不同。得到許多個體的果蠅腦模型之後，我們必須建立出一個標準的平均果蠅腦模型，這個標準平均腦模型足以代表全體果蠅的標準形狀，腦科學研究團隊可以在其中放入果蠅腦內所屬的蕈狀體與其他神經網路等結構，以利研究。
在平均果蠅腦模型的過程中，我們發現腦模型中的形狀雖然可以直接平均取得，在凸體形狀中可以得到正確的結果，但在內凹形體的部份卻會因為資料的歧異以及廣義的平均方法，而產生一些錯誤的資訊：諸如內凹口角度錯誤亦或是被填平等問題出現，這些內凹的特徵是腦科學研究者所希望保留下來的結構；為了保留果蠅腦模型特徵，我們提出一個修正廣義平均的方法，先經過剛體對位之後，使用者可以透過貝氏曲線畫出希望保留的個體模型特徵，配合自動貼合模型的方法，自動追蹤出代表每一個個體模型該位置的特徵貝氏曲線。本篇論文著重於如何依據這些資訊，產生出平均的貝氏曲線，並且使用非剛體形變的方式將每個個體模型扭曲到理論上平均的位置和形狀。經由這些步驟處理之後，就可以回到廣義的切片平均方法並重建出標準的果蠅腦模型，且修正上述廣義平均所發生的問題。

關鍵字

貝氏曲線 ； 平均模型 ； 變形 ； 特徵

並列摘要

As a noted topic of life science, brain research is aimed to solve how people learn and memorize. In Drosophila brain, it is discovered that several brain controlling genes are very similar to human’s, and so as how they function. Thus, Drosophila brain plays an important role for studying simple neural networks and can help figure out such main functions as learning and memory.
In order to study the main structures and functions of Drosophila brain, confocal microscope is used to image fluorescent brain slices, and reconstruct individual 3-D brain models, which from the platform that biologists place neural networks and other structures into. Due to the objectiveness, rather than taking any one model as a standard, it’s better to average all individual models. The averaged brain model is a standard model for image database.
The average method works well on most convex regions but encounters some problems on concave regions. The averaged concave region is filled up because of feature location, orientation and size variations, which need to be adjusted before averaging process.
A curve averaging and local warping method to solve this problem is presented in this study. A Bezier curve is drawn manually near the featured spots. After auto fitting to each brain model, the individual Bezier curves is produced. In order to keep the desired feature during performing Bezier curve averaging, calculate the Bezier curve moving vectors, and apply brain surface model a local deformation according to the Bezier curve moving. The proposed algorithm can perform a warping on local region to the average shape. Furthermore, it works well on keeping the desired feature after overall averaging process.

並列關鍵字

Bezier Curve ； Modle Averaging ； Deformation ； Feature

參考文獻

[1] Chao-Yu. Chen, “A Framework for Averaging and Updating Surface-Model on Drosophila Brain from Confocal Microscopy Images,” M.S. thesis, National Tsing Hua University 2008.

[4] T.W. Sederberg and S.R. Parry,“Free-Form Deformation of Solid Geometric Models,” Computer Graphics, vol. 20, no. 4, 1986.

[2] Chih-Lun. Lin, “A Semi-automatic Multi-segment Cubic Bezier Tube Based Feature Extraction Method for Surface Models” M.S. thesis, National Tsing Hua University 2009.

[3] H.-B. Yan, S.-M. Hu, and R. Martin, “Shape Deformation Using a Skeleton to Drive Simplex Transformations” IEEE Transactions on Visualization and Computer Graphics, Vol. 14, no. 3, MAY/JUNE 2008

[5] G. Y. Chen, Y. C. Chen, C. F. Lin, A. C. Hu, C. C. Wu, and Y. C. Chen, “Template-based automatic segmentation of Drosophila mushroom bodies”, ICS 2006, Dec. 4-6, Taipei, Taiwan.