RESUMEN
Gestational age estimation at time of birth is critical for determining the degree of prematurity of the infant and for administering appropriate postnatal treatment. We present a fully automated algorithm for estimating gestational age of premature infants through smartphone lens imaging of the anterior lens capsule vasculature (ALCV). Our algorithm uses a fully convolutional network and blind image quality analyzers to segment usable anterior capsule regions. Then, it extracts ALCV features using a residual neural network architecture and trains on these features using a support vector machine-based classifier. The classification algorithm is validated using leave-one-out cross-validation on videos captured from 124 neonates. The algorithm is expected to be an influential tool for remote and point-of-care gestational age estimation of premature neonates in low-income countries. To this end, we have made the software open source.
RESUMEN
Purpose: To correlate ellipsoid zone (EZ) defects on spectral-domain optical coherence tomography (SD-OCT) with retinal sensitivity loss on macular integrity assessment (MAIA) microperimetry in macular telangiectasia type 2 (MacTel). Methods: Macular SD-OCT volumes and microperimetry maps were obtained during the international, multicenter, randomized phase 2 trial of ciliary neurotrophic factor for type 2 MacTel on two visits within 5 days of one another. Software was developed to register SD-OCT to MAIA scanning laser ophthalmoscopy images and to overlay EZ defect areas on the microperimetry maps generated from microperimetry sensitivity values at specific points and from interpolated sensitivity values. A total of 134 eyes of 67 patients were investigated. Results: The semiautomated registration algorithm was found to be accurate, both qualitatively by visual inspection of the nearly perfect overlap of the retinal vessels and quantitatively as assessed by interobserver reliability metrics performed in 98 eyes of 49 patients (intraclass correlation of aggregate retinal sensitivity loss >0.99). Aggregate retinal sensitivity loss within the EZ defect area was highly correlated with EZ defect area (Pearson correlation coefficient 0.93 and 0.92 at screening and baseline for noninterpolated maps; both were 0.94 for interpolated maps; P values <0.001). Conclusions: With our software and image processing algorithms, there is nearly perfect correlation between retinal sensitivity on microperimetry and EZ defect area on SD-OCT. Our software allows determination of functional and structural changes with increasing disease severity and demonstrates that functional loss on microperimetry may be used as a surrogate marker of EZ loss on SD-OCT in type 2 MacTel.