| Purpose: A combined management of Traditional Chinese Medicine(TCM)and modern medicine methods was employed and discussed to treat refractory glaucoma after vitrectomy due to diabetic retinopathy(DR),and an approach to predict the incidence risk of DR for diabetes mellitus(DM)patients based on artificial intelligence(AI)was also explored,so as to propose some new ideas for the prevention and management of DR.Materials and methods: This study included 2 parts: Part 1.From January 2015 through December 2018,29 qi stagnation and blood stasis patients(29 eyes)of refractory glaucoma that had underwent previous pars plana vitrectomy due to DR were recruited in the department of ophthalmology,No.4 People's Hospital of Shenyang.The patients were then devided into 2 groups.The experimental group was treated with trabeculectomy with 23 G vitrectomy techniques combining oral administration of Modified Decoction of Taohong Siwu and Wuling Powder,and the control group was treated with trabeculectomy with 23 G vitrectomy techniques only.All the patients were followed-up for at least 12 months post-operatively.Pre-operative and post-operative intraocular pressure(IOP),best corrected visual accuity(BCVA),and complications were recorded.Part 2.Inspired by the traditional Chinese medical idea of "Zhi Wei Bing(to treat the disease before it develops)",researchers explored a possibility to evaluate the risk of diabetic retinopathy in the condition of community screening.Clinical data(excluding fundus images)of diabetic patients(including those who had retinopathy and those who didn't)were collected to train an artificial neural network,so as to establish an incidence risk prediction system of diabetic retinopathy based on artificial intelligence.From January 2014 to December 2017,86 diabetic patients were recruited in Shenyang No.4 people's Hospital,including 58 patients with various levels of DR and 32 patients without DR.The patients' information were recorded,including medical history,main vital signs,relevant biochemical indicators and DR diagnosis.An artificial neural network was built.69 patients were randomly chosen to train the network,and the remaining 17 patients' information was used for testing.Results:1.Part 1:(1)In all of the 29 patients,post-operative IOP decreased and finally remained in normal range in 18 eyes(62.1%),decreased and then rerised in 9 eyes(31.0%),and remained above normal in 2 eye(6.9%).The mean intraocular pressure at the end point was 17.9 ±4.9mm Hg,while the mean baseline intraocular pressure 41.9 ± 5.6mm Hg,with significantly statistic difference in between(p=0.000).In the trial group,the mean IOP at the end point was18.6 ± 2.9mmhg,which was significantly different from that at the baseline(43.9 ±4.8mmhg)(P = 0.000).In the control group,the mean IOP at the end point was 16.9 ±2.9mmhg,which was significantly different from that at the baseline(39.2 ± 2.9mmhg)(P= 0.000).(2)In all of the 29 patients,post-operative visual functions improved to some extent in 15 eyes(51.7%),declined in 9 eyes(31.0%),and remained unchanged in the other 5 eyes(17.2%).After evaluation,the mean BCVA at the end point was 0.120 ± 0.094,while the mean baseline BCVA 0.086 ± 0.117,with no statistic difference in between(p=0.097).The average BCVA at the end point was 0.108 ± 0.130,with no significant difference compared with the baseline BCVA(0.086 ± 0.117).(3)Hyphema occurred in 6 eyes(20.7%)during operation.Intraocular hypotension occurred in 13 eyes(44.8%)due to excessive filtration.Post-operative intraocular hemorrhage occurred in 5 eyes(17.2%).No severe complications,such as expeditious choroidal hemorrhage,retinal detachment,or Infectious endophthalmitis,were observed.2.Part 2:(1)The artificial neural network was trained in this study with the data of general information of patients,medical history information,vision,diabetes routine test indicator,etc.,and it was found that the most significant impact on the risk of diabetic retinopathy came from indicators such as age,length of DM history,vision,creatinine,DM management strategy,etc.,while family history,bad hobbies,glycosylated hemoglobin,etc.,seemed to have made less influences.(2)The accuracy of the artificial neural network trained in this study to predict diabetic retinopathy reached 85.5%.Conclusions:1.It is safe and effective to control intraocular pressure and protect visual function to treat refractory glaucoma after vitrectomy due to DR with the combination of "Modified Decoction of Taohong Siwu and Wuling Powder" and trabeculectomy utilizing 23 G vitrectomytechniques.2.Artificial intelligence technology has the possibility to overcome the limitation of fundus imaging acquisition and to realize DR incidence risk prediction.3.In the aging society,in comprehensive consideration of patients' quality of life,emotional and economic burden of the family and the society,and other aspects,it may be more important and reasonable to pay some more attention to prevention and control before or in the early stage of the diseases compared with the treatment of patients with severe blinding fundus diseases.Artificial intelligence based on big data shows an obvious advantage in this field. |
PDF Full Text Request