Dataset Information

DeepPDT-Net: predicting the outcome of photodynamic therapy for chronic central serous chorioretinopathy using two-stage multimodal transfer learning.

ABSTRACT: Central serous chorioretinopathy (CSC), characterized by serous detachment of the macular retina, can cause permanent vision loss in the chronic course. Chronic CSC is generally treated with photodynamic therapy (PDT), which is costly and quite invasive, and the results are unpredictable. In a retrospective case-control study design, we developed a two-stage deep learning model to predict 1-year outcome of PDT using initial multimodal clinical data. The training dataset included 166 eyes with chronic CSC and an additional learning dataset containing 745 healthy control eyes. A pre-trained ResNet50-based convolutional neural network was first trained with normal fundus photographs (FPs) to detect CSC and then adapted to predict CSC treatability through transfer learning. The domain-specific ResNet50 successfully predicted treatable and refractory CSC (accuracy, 83.9%). Then other multimodal clinical data were integrated with the FP deep features using XGBoost.The final combined model (DeepPDT-Net) outperformed the domain-specific ResNet50 (accuracy, 88.0%). The FP deep features had the greatest impact on DeepPDT-Net performance, followed by central foveal thickness and age. In conclusion, DeepPDT-Net could solve the PDT outcome prediction task challenging even to retinal specialists. This two-stage strategy, adopting transfer learning and concatenating multimodal data, can overcome the clinical prediction obstacles arising from insufficient datasets.

SUBMITTER: Yoo TK

PROVIDER: S-EPMC9636239 | biostudies-literature | 2022 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

DeepPDT-Net: predicting the outcome of photodynamic therapy for chronic central serous chorioretinopathy using two-stage multimodal transfer learning.

Yoo Tae Keun TK Kim Seo Hee SH Kim Min M Lee Christopher Seungkyu CS Byeon Suk Ho SH Kim Sung Soo SS Yeo Jinyoung J Choi Eun Young EY

Scientific reports 20221104 1

Central serous chorioretinopathy (CSC), characterized by serous detachment of the macular retina, can cause permanent vision loss in the chronic course. Chronic CSC is generally treated with photodynamic therapy (PDT), which is costly and quite invasive, and the results are unpredictable. In a retrospective case-control study design, we developed a two-stage deep learning model to predict 1-year outcome of PDT using initial multimodal clinical data. The training dataset included 166 eyes with ch ...[more]

PMID: 36333442

Similar Datasets

Project description:ObjectiveTo evaluate visual acuity (VA) and factors influencing VA using new multimodal imaging-based classification of central serous chorioretinopathy (CSCR).MethodsRetrospective, observational and cross-sectional study on 229 naïve eyes diagnosed as CSCR with available baseline data and multimodal imaging. Each case was classified into (i) simple/complex/atypical; (ii) primary/recurrent/resolved; (iii) persistent or not; (iv) outer retinal atrophy(ORA) present/absent; (v) foveal involvement present/absent; and (vi) macular neovascularization(MNV) present/absent. Best corrected visual acuity (BCVA) was correlated to the classification as well as every parameter of the classification.ResultsMedian BCVA was 0.18 logMAR [95% Confidence Interval (CI)0.16-0.18] with median duration of complaints of one month (95% CI,6.14-13.0 months). Age of the patient (r = -0.24, p = 0.002) and duration of the disease (r = -0.32, p < 0.001) correlated significantly with BCVA. Logistic regression model showed that older age [odds ratio (OR) = 0.96, p = 0.05], female gender (OR = 2.45, p = 0.046), presence of ORA(OR = 0.34, p = 0.012),and foveal involvement(OR = 0.18, p = 0.007) were statistically significantly associated with poorer BCVA. Eyes classified as complex, persistent CSCR, with ORA or foveal involvement demonstrated lower BCVA compared to those with simple, non-persistent CSCR, without ORA or without foveal involvement (p < 0.05). Eyes with complex CSCR (p < 0.001), atypical CSCR(p = 0.025), persistent subretinal fluid (SRF) (p = 0.001) and those with ORA (p < 0.001) demonstrated a trend towards severe visual loss. Prevalence of persistent SRF, recurrent episodes and ORA was significantly higher among eyes with complex CSCR (p < 0.001) while there was no difference in prevalence of resolved cases (p = 0.07), foveal involvement (p = 0.28) and MNV (p = 0.45) between simple and complex cases.ConclusionThere is a strong correlation between VA and foveal involvement and ORA using the new classification. Thus, the objective parameters of the classification can be incorporated in establishing the treatment guidelines for CSCR.

Project description:PurposeThis study evaluated changes in choroidal vascularity after half-fluence photodynamic therapy (HF-PDT) in patients with central serous chorioretinopathy (CSC) using swept-source optical coherence tomography (SS-OCT) en face imaging.MethodsThis retrospective comparative case series included 50 eyes of 25 patients with unilateral CSC who underwent HF-PDT and 50 age-and sex-matched normal healthy control eyes. En face SS-OCT images of the choriocapillaris, Sattler's layer, and Haller's layer were converted into binary images. The vascular proportions were defined as the percentage of the area of vascular lumen against the area of the 3.0-mm-diameter circular area. The main outcome measures were the vascular proportions before HF-PDT and at 6 weeks, 6 months, and 12 months after HF-PDT.ResultsAt baseline, the vascular proportions in the CSC eyes were significantly greater than those in the control eyes in all layers (choriocapillaris: 51.8% ± 15.5% vs. 41.3 ± 18.7%, P = 0.018; Sattler's: 58.6% ± 13.4% vs. 49.7% ± 15.7%, P = 0.017; Haller's: 65.3% ± 15.3% vs. 53.0% ± 13.4%, P = 0.001). In the CSC eyes, the vascular proportion in the choriocapillaris significantly decreased at 6 weeks (36.6% ± 16.9%, P < 0.001), 6 months (34.0% ± 12.3%, P < 0.001), and 12 months (34.8% ± 17.6%, P < 0.001) after HF-PDT compared with baseline. The vascular proportions in Sattler's and Haller's layers did not show a significant decrease at 6 weeks (Sattler's: 49.7% ± 17.3%, P = 0.052 and Haller's: 58.3% ± 12.9%, P = 0.558) but decreased significantly at 6 months (Sattler's: 48.9% ± 12.4%, P < 0.001 and Haller's: 57.7% ± 15.7%, P = 0.027) and 12 months after HF-PDT from the baseline values (Sattler's: 45.8% ± 10.4%, P < 0.001 and Haller's: 56.8% ± 15.7%, P < 0.001).ConclusionAfter HF-PDT, the choriocapillaris showed the earliest decrease in vascular proportion of en face images, Sattler's and Haller's layers showed later decreases. The temporal differences in the response of each layer may reflect the pathophysiology of CSC and the therapeutic mechanism of HF-PDT.

Dataset Information

DeepPDT-Net: predicting the outcome of photodynamic therapy for chronic central serous chorioretinopathy using two-stage multimodal transfer learning.

Publications

DeepPDT-Net: predicting the outcome of photodynamic therapy for chronic central serous chorioretinopathy using two-stage multimodal transfer learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets