Intelligent Diagnosis And Clinical Outcomes Of Different Optical Coherence Tomography Patterns Of Diabetic Macular Edema

Chapter1 Intelligent diagnosis of different optical coherence tomography patterns of diabetic macular edemaPurpose:To develop a deep learning(DL)model for detection of morphologic patterns of diabetic macular edema(DME)based on optical coherence tomography(OCT)images.Methods:In the training set,12,365 OCT images extracted from a public dataset and an ophthalmic center were used.A total of 656 OCT images from another ophthalmic center were used as the external validation set.Three OCT patterns of DME,including diffused retinal thickening(DRT),cystoid macular edema(CME),and serous retinal detachment(SRD)were defined and multi-labeled.A DL model was trained to detect three OCT patterns of DME using Visual Geometry Group(VGG)16 convolutional neural networks.Five-fold cross-validation method was applied for internal validation.Occlusion test was applied to demonstrate whether the critical components in OCT images were highly correlated with accurate detection of DME patterns.The accuracy,sensitivity,specificity,and area under the receiver operating characteristic curve(AUC)were calculated.Results:In internal validation,the mean area under the receiver operating characteristic curve(AUC)for detection of three OCT patterns(i.e.,DRT,CME,and SRD)were 0.971,0.974,and 0.994,respectively,with mean accuracy of 93.0%,95.1%,and 98.8%,respectively,mean sensitivity of 93.5%,94.5%,and 96.7%,respectively,and mean specificity of 92.3%,95.6%,and 99.3%,respectively.In external validation,the AUC were 0.970,0.997,and 0.997,respectively,with accuracy of 90.2%,95.4%,and 95.9%,respectively,sensitivity of 80.1%,93.4%,and 94.9%,respectively,and specificity of 97.6%,97.2%,and 96.5%,respectively.Occlusion test showed that the DL model could successfully identify the pathologic regions most critical for accurate detection.Conclusions:Our DL model demonstrated high accuracy and transparency in detection of morphologic patterns of DME based on OCT images.These results emphasized the potential of artificial intelligence in assisting clinical decision-making processes in DME patients.Chapter2 Comparison of clinical outcomes of different optical coherence tomography patterns of diabetic macular edemaPurpose:To evaluate the edema reduction after intravitreal injection of ranibizumab(IVR)in two DME components in the same eye using OCT.Methods:A total of 113 eyes with mixed OCT pattern of DME were included.All the eyes underwent best-corrected visual acuity(BCVA)examination and OCT scanning at baseline and follow-up visits(1,3,and 6 months after 3 monthly consecutive IVR).The mixed OCT pattern of DME was classified into 2 OCT components:SRD component and non-SRD component.Central foveal thickness(CFT),foveal thickness of the SRD component(SRDFT)and the non-SRD component(NSRDFT)was compared between baseline and follow-up visits.Reduction and reduction ratio of the SRDFT and the NSRDFT at each follow-up were also compared.When calculating the reduction ratio,the commonly used formula was innovatively optimized by subtracting the normal CFT to avoid calculation errors caused by the normal CFT.Results:At 1,3,and 6 months after IVR,CFT,SRDFT,and NSRDFT were significantly reduced compared to baseline foveal thickness(all p<0.001).There was no significant difference in reduction or reduction ratio between NSRDFT and SRDFT during each follow-up(all p>0.05).The correlation between BCVA and SRDFT was most significant at baseline(r=0.366,p<0.001)and the correlation between BCVA and NSRDFT was most significant at 6 months(r=0.426,p<0.001).BCVA improvement was more significantly correlated with reduction or reduction ratio of SRDFT at each follow-up(r=0.271-0.426,all p<0.01).Conclusions:IVR could effectively reduce both the SRD and non-SRD components according to our optimized formula,indicating that IVR was effective for repairing the breakdown of the inner and outer blood-retinal barriers.The association between BCVA improvement and edema reduction was more significant in the SRD component.Chapter3 Prognostic factors of diabetic macular edema with coexisting serous retinal detachmentPurpose:To explore the correlation between foveal bulge(FB)and visual prognosis after resolution of diabetic macular edema with coexisting serous retinal detachment(SRD-DME)and the influencing factors of restoration of FB.Methods:A total of 52 eyes with resolved SRD-DME after IVR treatment and an intact ellipsoid zone at the central fovea were included.All eyes underwent BCVA examination and OCT scanning at baseline and follow-up visits(1,3 and 6 months after 3 monthly consecutive IVR).The eyes were divided into two groups according to the presence of FB at 6 months.BCVA,CFT,height of serous retinal detachment(SRDH),outer nuclear layer(ONL)thickness,photoreceptor inner segment(PIS)and outer segment(POS)length was compared between the two groups.Results:A FB was found in 25 of 52(48.1%)eyes at 6 months.The FB(+)group had lower SRDH at baseline,and better BCVA,longer POS length at 6 months(all p<0.05).There was no significant difference in the CFT,ONL thickness and PIS length at 6 months between the two groups(all p>0.05).More eyes in the FB(+)group had complete SRD resolution at 1 month(84.0%vs.48.1%,p=0.009)and 3 months(96.0%vs.66.7%,p=0.012).Eyes with complete SRD resolution were more likely to have a FB at 6 months compared with eyes without SRD resolution at 1 month(61.8%vs.22.2%,p=0.009)or 3 months(51.7%vs.10.0%,p=0.012).Conclusions:Restoration of the FB is associated with better BCVA after resolution of SRD-DME.SRD-DME eyes with lower baseline SRDH or faster SRD resolution are more likely to have a FB at 6 months.