電腦模擬顏面整形手術術後臉型之準確性及在不修改模擬軟體情況下增加模擬準確度
中文摘要
隨著資訊科技的進步，視訊影像的模擬逐漸被應用在顏面整形手術的 術後模擬預估上。這些模擬影像會大大影響患者決定是否接受此一手 術而且也可提供醫師訂定治療計畫的資訊；因此，這些模擬是否真實 可信是極需要探討的。本論文的目的在評估此類影像模擬的可信度及 真實性並提出一種改善模擬準確度的方法。所使用研究對象為三十位 因上下顎前凸而接受雙顎正顎手術的成人患者，將這些患者的術後模 擬影像和實際手術後的影像作一比較，而類神經網路則用來改進模擬 的準確度。其結果發現鼻尖點的模擬誤差最小，平均誤差小於 1 mm；模擬較可信的點在鼻尖及軟組織 A 點。平均每次模擬出現誤差小 於 2 mm 的機會約為 50% ；在應用類神經網路改善後，模擬誤差大 幅減少， X 軸平均誤差改善率為 43.9% ， Y 軸平均誤差改善率為 -6 .6% ；平均每次模擬出現誤差小於 2 mm 的機會約為 84.5% 。整體來 說，利用電腦影像來模擬顏面正顎整形手術出來的模擬是值得作為參 考的，然而其準確度還需要加以改進才能作為臨床訂定醫療計畫的依 據，而類神經網路則提供了一個不錯的改良方法。
Evaluation and improvement of computer simulation of the results in plastic surgery
英文摘要
As the advancement of computer technology, video image simulation became more and more frequently used in the simulation of craniofacial surgery. The simulation will greatly affect the decision making of the patients and also provide information to the surgeon and orthodontics.
Thus, how real would the simulation present was primary concern. The purpose of this thesis was to evaluate the accuracy and reliability of the post-surgical simulation and try to find out a method to improve the simulation. Thirty bimaxillary protrusion patients who under went two jaw surgery were taken into consideration. The simulation was compared with the real post-su rgical facial profile. Artificial neural network was used to make the simulation more accurate.
The results showed the most accurate area located at tip of nose with the average errors were l ess than 1 mm. The more reliable area located at tip of nose, and soft tissue A point. The avera ge probability of every simulation that the errors were smaller than 2 mm was about 50%. Afte r applying the artificial neural network to the input data, the simulation errors were reduced.Th e improvement rate of average simulation errors in X-axis and Y-axis were 43.9% and -6.6%.
The average probability of every simulation that the errors were smaller than 2 mm was 84.5%
. In general, the present computer simulation of the craniofacial plastic surgery is good for refe rence but efforts still need to be done before it can provide enough information for surgical tre atment plan. Artificial neural network may provide a good way to achieve this goal.
