Instituição onde foi realizado o trabalho
- Principal: Duke University
- Secundaria: Universidade Estadual de Campinas (UNICAMP)
- ALESSANDRO ADAD JAMMAL (Interesse Comercial: NÃO)
- Atalie C Thompson (Interesse Comercial: NÃO)
- Nara G Ogata (Interesse Comercial: NÃO)
- Carla N Urata (Interesse Comercial: NÃO)
- Eduardo B Mariottoni (Interesse Comercial: NÃO)
- Vital P Costa (Interesse Comercial: NÃO)
- Felipe A Medeiros (Interesse Comercial: NÃO)
DETECTING RETINAL NERVE FIBER LAYER SEGMENTATION ERRORS ON SPECTRAL DOMAIN OPTICAL COHERENCE TOMOGRAPHY WITH A DEEP LEARNING ALGORITHM
To propose a deep learning (DL) algorithm that detects errors in retinal never fiber layer (RNFL) segmentation from spectral-domain optical coherence tomography (SDOCT) B-scans.
A cross-sectional study with 25,250 OCT B-scans reviewed for segmentation errors from 1,363 eyes of 709 subjects. The sample was randomly divided into validation plus training (80%) and test (20%) sets. SDOCT B-scans that had been labeled for quality by human graders were used as the reference standard to train a DL convolutional neural network to detect RNFL segmentation errors. The performance of the DL algorithm was evaluated in the test sample by outputting a probability of having a segmentation error for each B-scan. The accuracy and the ability of the algorithm to detect segmentation errors was evaluated with the area under the receiver operating characteristic curve (ROC).
Mean DL probabilities of segmentation error in the test sample were 0.91 0.17 vs. 0.12 0.21 (P<0.001) for scans with and without segmentation errors, respectively. The DL algorithm had an area under the ROC curve of 0.981 (95% CI: 0.975 to 0.986) and an overall accuracy of 92.8%. Figure 1 shows SDOCT B-scans with segmentation errors correctly detected by both the reading center and the DL algorithm. Class activation maps (heatmaps) on the right show the regions of the B-scans that had greatest weight in the DL algorithm classification. It is possible to see that the DL algorithm identified (A) errors of segmentation involving both the delineation of the internal limiting membrane, (B) as well as the posterior boundary of the RNFL, (C) multiple errors in the same scan, and (D) even very small segmentation errors.
We introduced a novel DL approach to assess RNFL segmentation errors in SDOCT B-scans. This algorithm can help clinicians and researchers review images for artifacts in SDOCT tests in a timely manner and avoid inaccurate diagnostic interpretations.