Clinical Research On Automatic Acquisition And Registration Algorithm Of Reference Data For Mid-facial Bone Defect Reconstruction

Part1:Establishment of Clinical Protocol of Computer-assisted Post-traumatic Midfacial Reconstruction with Vascularized flapsPurpose:The aim of this study was to present a treatment protocol for the individual repair of post-traumatic midfacial bone defects with vascularized flaps assisted by digital techniques.Methods:This study reviewed 11 patients with post-traumatic maxillofacial bone defects who underwent reconstruction with composite vascularized bone flaps assisted by digital techniques between May 2009 and December 2018.Preoperative computed tomography(CT)data were imported into ProPlan CMF software to complete virtual fracture reduction and reconstruction.Surgical navigation,three dimensionally(3D)printed surgical plates,and prefabricated titanium mesh/plates were used to guide the actual surgery.All patients underwent open reduction,internal fixation and reconstruction surgery in one stage.CT data obtained at 1 week postoperatively were imported into Geomagic Control software to evaluate the accuracy of the virtual surgical plan.The mean follow-up interval was 23 months(range 3-60 months).Donor and recipient site morbidity and second-stage procedures to rehabilitate the dentition and cosmetic organs were recorded.Results:The flap success rate was 100%.Five patients had deep circumflex iliac artery flaps and six patients had fibula flaps.The accuracy of computer-assisted surgery was 4.4±0.8 mm.No patients suffered recipient site complications such as infection,iatrogenic facial nerve damage,flap resorption,or donor site complications such as limp.Six patients underwent dental implantation and seven patients completed sequential plastic surgeries rehabilitating oral function and cosmetic organs.All patients were satisfied with their postoperative appearanceConclusions:This study is novel in presenting a treatment protocol for post-traumatic midfacial bone defect reconstruction with vascularized flaps assisted by digital techniques.This is a feasible method that enables individualized,minimally invasive,and functional reconstructions.Part2:Iterative Closest Point Algorithm Assisted by Three-Dimensional Craniomaxillofacial Database of Normal Chinese People to Predict Missing Midfacial BonePurpose:The aim of this study was to present an iterative closest point algorithm assisted by three-dimensional craniomaxillofacial database of normal Chinese People for the automatic acquirement of reference data for midfacial bone defects.Methods:This study combined iterative closest point algorithm with a 3-Dimensional craniomaxillofacial database of normal Chinese people including 500 skull models.A total of 30 intact normal skull models(15 males,15 females)(S0)not included in the database and 15 patients with unilateral midfacial bone defect(Sx)were served as samples of model experiment.Virtual bony defects of the bilateral zygomatic(S1)and naso-orbital-ethmoid(NOE)(S2)region were created.For unilateral midfacial bone defects,the healthy side was mirrored and merged as the intact skull model(Sx').For each defected skull model,the algorithm was applied to select the most similar skull model from the database with the same gender.The deviation of each Euclidean distance between each paired point cloud present in the defected skull and the normal skull from the database was calculated.The mean of the deviation was considered as the similarity value of each model in the database.The model with the lowest similarity value in the database was exported as the reference data(S1r,S2r,Sxr).3Dimensional and 2-Dimensional comparison were conducted to evaluate the error between reference skull model(S1r,S2r,Sxr)with original intact model(S0,Sx')which was considered as ground truth.Root-mean-square deviation was considered as error of the algorithm.Results:The algorithm was tested on an CPU with 1.80 GHz and average processing time for NOE,bilateral zygomatic and unilateral midfacial bone defects was 495 ±106s,492±95s and 620±95s respectively.The average root-mean-square deviation of defect area was less than 2mm.Conclusions:It's feasible using iterative closest point algorithm based on 3D craniomaxillofacial database of normal Chinese people to automatically predict the reference data of missing midfacial bone.Part3:Non-rigid Registration Algorithm Assisted by Three-Dimensional Craniomaxillofacial Database of Normal Chinese People to Predict Missing Midfacial BonePurpose:The aim of this study was to present a non-rigid registration algorithm assisted by three-dimensional craniomaxillofacial database of normal Chinese People for the automatic acquirement of reference data for midfacial bone defects.Methods:This study combined non-rigid registration algorithm with a 3-Dimensional craniomaxillofacial database of normal Chinese people including 500 skull models.100 skull models were randomly selected from the database to creat SSM by ICP method.A total of 30 intact normal skull models(15 males,15 females)(SO)not included in the database and 15 patients with unilateral midfacial bone defect(Sx)were served as samples of model experiment.Virtual bony defects of the bilateral zygomatic(S1)and naso-orbital-ethmoid(NOE)(S2)region were created.For unilateral midfacial bone defects,the healthy side was mirrored and merged as the intact skull model(Sx').Non-rigid registration was applied between each skull defect model and SSM.A total of 40 skull models were generated.CNN network was then conducted to automatically detect 22 landmark points and measure distances between corresponding landmarks.Comparing with normal cephalometric values,the model with the largest similarity was selected as the reference data(S1r,S2r,Sxr).3Dimensional and 2-Dimensional comparison were conducted to evaluate the error between reference skull model(S1r,S2r,Sxr)with original intact model(S0,Sx')which was considered as ground truth.Root-mean-square deviation was considered as error of the algorithm.Results:The algorithm was tested on an CPU with 1.80 GHz and average non-rigid regiatration processing time for NOE,bilateral zygomatic,and unilateral midfacacial bone defects was 1594±94s,1636±115s and 1630±75s respectively.The average model selection processing time for NOE,bilateral zygomatic,and unilateral midfacacial bone defects was 846±13s,849±27s and 845± 15s respectively.The average root-mean-square deviation of defect area was less than 2mm.Conclusions:It's feasible using non-rigid registration algorithm based on 3D craniomaxillofacial database of normal Chinese people to automatically predict the reference data of missing midfacial bone.