| 摘要: | 在進行腦部手術之前，醫師依靠患者所拍攝的醫學解剖影像，如電腦斷層(CT)或磁振造影(MRI)，以及自身的知識與手術經驗進行手術路徑規劃，但2D影像能提供的資訊量有限，可能導致手術時對患者的腦部組織產生傷害。若是能在手術路徑規劃的時候提供腦部功能區與白質神經纖維的3D模型，將可以幫助醫師規劃出更安全的手術路徑，降低對患者可能產生的傷害。
本研究結合皮質層功能區模型與白質神經纖維模型，發展出一套可幫助醫師於術前了解皮質層功能區分布及腫瘤周圍神經纖維分布的系統。研究內容包含了皮質層功能區以及白質神經纖維的3D模型重建、影像對位以及腫瘤所在區域的神經模型感興趣區域(ROI)設定。使用FreeSurfer對MRI影像進行功能區的辨識而後重建出皮質層功能區模型。白質神經纖維則使用DTI影像針對不同的神經纖維追蹤演算法進行正確性比對；影像對位是將皮質層功能區與白質神經束纖維的座標空間轉換為MRI影像座標，之後醫師可依據對位好的融合影像及設定在腫瘤周遭的感興趣區,規畫出可避開重要組織的安全路徑。
所建構的皮質層功能區模型根據解剖特徵進行比對，比對結果顯示重要的功能區如初級體感皮層、初級運動皮層、初級視覺皮層以及初級聽覺皮層均有完整重建又與Desikan-Killiany Atlas進行比對，顯示出功能區都有成功識別出來，因此模型是具有可靠性的。白質神經纖維模型針對上、下縱束、鉤束、弓狀束、下枕額束、扣帶束、放射冠以及胼胝體的完整性進行確認，結果顯示以Tensorline神經纖維追蹤演算法較為完整可靠。皮質層功能區影像與MRI影像對位準確度，使用棋盤分類法可以觀察到大腦輪廓、白質、灰質以及腫瘤皆接合在一起，以此推論對位精準度是可靠的。
;Before undergoing brain surgery, surgeons can only rely on the patient’s medical anatomical images, such as computerized tomography (CT) or magnetic resonance imaging (MRI), and their knowledge to plan surgical paths. However, 2D images can provide limited information. If 3D models of brain functional areas and white matter nerve fibers can be provided, it can help surgeons plan safer surgical paths.
This research includes 3D model reconstruction of cortical functional areas and white matter nerve fibers, image alignment, and region of interest (ROI) setting of neural model around the tumor. MRI images are used to identify functional areas and then reconstruct cortical functional area models. Different nerve fiber tracking algorithms are applied to generate white matter nerve fibers using DTI images and their correctness of reconstructed nerve fibers is compared. The coordinate frames of cortical functional areas and white matter nerve fibers are converted and aligned with MRI images coordinate frame. Then the surgeon can plan a safe surgical path based on the fused images.
By comparing with anatomical features, the constructed cortex functional area model shows that important functional areas have been completely reconstructed and identified. The white matter nerve fiber model is confirmed for the integrity of the superior, inferior longitudinal fasciculus, the uncinate fasciculus, the arcuate fasciculus, the inferior occipitofrontal fasciculus, the cingulate fasciculus, the corona radiata, and the corpus callosum. The results show that the Tensorline algorithm is much better. The alignment accuracy of cortical functional area images and MRI images is evaluated by using the checkerboard classification method. The results show that the brain outline, white matter, gray matter, and tumors are all connected smoothly, so it can be inferred that the alignment accuracy is reliable. |
