Graph-Based Multi-Surface Segmentation of OCT Data Using Trained Hard and Soft Constraints

Optical coherence tomography (OCT) is a well-established image modality in ophthalmology and used daily in the clinic. Automatic evaluation of such datasets requires an accurate segmentation of the retinal cell layers. However, due to the naturally low signal to noise ratio and the resulting bad image quality, this task remains challenging. We propose an automatic graph-based multi-surface segmentation algorithm that internally uses soft constraints to add prior information from a learned model. This improves the accuracy of the segmentation and increase the robustness to noise. Furthermore, we show that the graph size can be greatly reduced by applying a smart segmentation scheme. This allows the segmentation to be computed in seconds instead of minutes, without deteriorating the segmentation accuracy, making it ideal for a clinical setup. An extensive evaluation on 20 OCT datasets of healthy eyes was performed and showed a mean unsigned segmentation error of 3.05 ±0.54 μm over all datasets when compared to the average observer, which is lower than the inter-observer variability. Similar performance was measured for the task of drusen segmentation, demonstrating the usefulness of using soft constraints as a tool to deal with pathologies.

Morphometric assessment of normal, suspect and glaucomatous optic discs with Stratus OCT and HRT II

AIMS: To compare morphometric parameters and diagnostic performance of the new Stratus Optical Coherence Tomograph (OCT) Disc mode and the Heidelberg Retina Tomograph (HRT); to evaluate OCT's accuracy in determining optic nerve head (ONH) borders. METHODS: Controls and patients with ocular hypertension, glaucoma-like discs, and glaucoma were imaged with OCT Disc mode, HRT II, and colour disc photography (DISC-PHOT). In a separate session, automatically depicted ONH shape and size in OCT were compared with DISC-PHOT, and disc borders adjusted manually where required. In a masked fashion, all print-outs and photographs were studied and discs classified as normal, borderline, and abnormal. The Cohen kappa method was then applied to test for agreement of classification. Bland-Altman analysis was used for comparison of disc measures. RESULTS: In all, 49 eyes were evaluated. Automated disc margin recognition failed in 53%. Misplaced margin points were more frequently found in myopic eyes, but only 31/187 were located in an area of peripapillary atrophy. Agreement of OCT with photography-based diagnosis was excellent in normally looking ONHs, but moderate in discs with large cups, where HRT performed better. OCT values were consistently larger than HRT values for disc and cup area. Compared with HRT, small rim areas and volumes tended to be minimized by OCT, and larger ones to be magnified. CONCLUSIONS: Stratus OCT Disc protocol performed overall well in differentiating between normal and glaucomatous ONHs. However, failure of disc border recognition was frequently observed, making manual correction necessary. ONH measures cannot be directly compared between HRT and OCT.

Quantitative analysis of OCT characteristics in patients with achromatopsia and blue-cone monochromatism

PURPOSE: To quantify optical coherence tomography (OCT) images of the central retina in patients with blue-cone monochromatism (BCM) and achromatopsia (ACH) compared with healthy control individuals. METHODS: The study included 15 patients with ACH, 6 with BCM, and 20 control subjects. Diagnosis of BCM and ACH was established by visual acuity testing, morphologic examination, color vision testing, and Ganzfeld ERG recording. OCT images were acquired with the Stratus OCT 3 (Carl Zeiss Meditec AG, Oberkochen, Germany). Foveal OCT images were analyzed by calculating longitudinal reflectivity profiles (LRPs) from scan lines. Profiles were analyzed quantitatively to determine foveal thickness and distances between reflectivity layers. RESULTS: Patients with ACH and BCM had a mean visual acuity of 20/200 and 20/60, respectively. Color vision testing results were characteristic of the diseases. The LRPs of control subjects yielded four peaks (P1-P4), presumably representing the RPE (P1), the ovoid region of the photoreceptors (P2), the external limiting membrane (ELM) (P3), and the internal limiting membrane (P4). In patients with ACH, P2 was absent, but foveal thickness (P1-P4) did not differ significantly from that in the control subjects (187 +/- 20 vs. 192 +/- 14 microm, respectively). The distance from P1 to P3 did not differ significantly (78 +/- 10 vs. 82 +/- 5 microm) between ACH and controls subjects. In patients with BCM, P3 was lacking, and P2 advanced toward P1 compared with the control subjects (32 +/- 6 vs. 48 +/- 4 microm). Foveal thickness (153 +/- 16 microm) was significantly reduced compared with that in control subjects and patients with ACH. CONCLUSIONS: Quantitative OCT image analysis reveals distinct patterns for controls subjects and patients with ACH and BCM, respectively. Quantitative analysis of OCT imaging can be useful in differentiating retinal diseases affecting photoreceptors. Foveal thickness is similar in both normal subjects and patients with ACH but is decreased in patients with BCM.

Exposure dating of Late Glacial and pre-LGM moraines in the Cordon de Doña Rosa, Northern/Central Chile (~ 31°S)

Despite the important role of the Central Andes (15–30° S) for climate reconstruction, knowledge about the Quaternary glaciation is very limited due to the scarcity of organic material for radiocarbon dating. We applied 10Be surface exposure dating (SED) on 22 boulders from moraines in the Cordon de Doña Rosa, Northern/Central Chile (~31° S). The results show that several glacial advances in the southern Central Andes occurred during the Late Glacial between ~14.7±1.5 and 11.6±1.2 ka. A much more extensive glaciation is dated to ~32±3 ka, predating the temperature minimum of the global LGM (Last Glacial Maximum: ~20 ka). Reviewing these results in the paleoclimatic context, we conclude that the Late Glacial advances were most likely caused by an intensification of the tropical circulation and a corresponding increase in summer precipitation. High-latitude temperatures minima, e.g. the Younger Dryas (YD) and the Antarctic Cold Reversal (ACR) may have triggered individual advances, but current systematic exposure age uncertainties limit precise correlations. The absence of LGM moraines indicates that moisture advection was too limited to allow significant glacial advances at ~20 ka. The tropical circulation was less intensive despite the maximum in austral summer insolation. Winter precipitation was apparently also insufficient, although pollen and marine studies indicate a northward shift of the westerlies at that time. The dominant pre-LGM glacial advances in Northern/Central Chile at ~32 ka required lower temperatures and increased precipitation than today. We conclude that the westerlies were more intense and/or shifted equatorward, possibly due to increased snow and ice cover at higher southern latitudes coinciding with a minimum of insolation.