Self-Supervised Learning for Improved Optical Coherence Tomography Detection of Macular Telangiectasia Type 2.

Importance: Deep learning image analysis often depends on large, labeled datasets, which are difficult to obtain for rare diseases. Objective: To develop a self-supervised approach for automated classification of macular telangiectasia type 2 (MacTel) on optical coherence tomography (OCT) with limited labeled data. Design, Setting, and Participants: This was a retrospective comparative study. OCT images from May 2014 to May 2019 were collected by the Lowy Medical Research Institute, La Jolla, California, and the University of Washington, Seattle, from January 2016 to October 2022. Clinical diagnoses of patients with and without MacTel were confirmed by retina specialists. Data were analyzed from January to September 2023. Exposures: Two convolutional neural networks were pretrained using the Bootstrap Your Own Latent algorithm on unlabeled training data and fine-tuned with labeled training data to predict MacTel (self-supervised method). ResNet18 and ResNet50 models were also trained using all labeled data (supervised method). Main Outcomes and Measures: The ground truth yes vs no MacTel diagnosis is determined by retinal specialists based on spectral-domain OCT. The models' predictions were compared against human graders using accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), area under precision recall curve (AUPRC), and area under the receiver operating characteristic curve (AUROC). Uniform manifold approximation and projection was performed for dimension reduction and GradCAM visualizations for supervised and self-supervised methods. Results: A total of 2636 OCT scans from 780 patients with MacTel and 131 patients without MacTel were included from the MacTel Project (mean [SD] age, 60.8 [11.7] years; 63.8% female), and another 2564 from 1769 patients without MacTel from the University of Washington (mean [SD] age, 61.2 [18.1] years; 53.4% female). The self-supervised approach fine-tuned on 100% of the labeled training data with ResNet50 as the feature extractor performed the best, achieving an AUPRC of 0.971 (95% CI, 0.969-0.972), an AUROC of 0.970 (95% CI, 0.970-0.973), accuracy of 0.898%, sensitivity of 0.898, specificity of 0.949, PPV of 0.935, and NPV of 0.919. With only 419 OCT volumes (185 MacTel patients in 10% of labeled training dataset), the ResNet18 self-supervised model achieved comparable performance, with an AUPRC of 0.958 (95% CI, 0.957-0.960), an AUROC of 0.966 (95% CI, 0.964-0.967), and accuracy, sensitivity, specificity, PPV, and NPV of 90.2%, 0.884, 0.916, 0.896, and 0.906, respectively. The self-supervised models showed better agreement with the more experienced human expert graders. Conclusions and Relevance: The findings suggest that self-supervised learning may improve the accuracy of automated MacTel vs non-MacTel binary classification on OCT with limited labeled training data, and these approaches may be applicable to other rare diseases, although further research is warranted.

Clinical features and predictive biomarkers for bladder cancer in patients with type 2 diabetes presenting with haematuria.

AIMS: To identify clinical features and protein biomarkers associated with bladder cancer (BC) in individuals with type 2 diabetes mellitus presenting with haematuria. MATERIALS AND METHODS: Data collected from the Haematuria Biomarker (HaBio) study was used in this analysis. A matched sub-cohort of patients with type 2 diabetes and patients without diabetes was created based on age, sex, and BC diagnosis, using approximately a 1:2 fixed ratio. Randox Biochip Array Technology and ELISA were applied for measurement of 66 candidate serum and urine protein biomarkers. Hazard ratios and 95% confidence intervals were estimated by chi-squared and Wilcoxon rank sum test for clinical features and candidate protein biomarkers. Diagnostic protein biomarker models were identified using Lasso-based binominal regression analysis. RESULTS: There was no difference in BC grade, stage, and severity between individuals with type 2 diabetes and matched controls. Incidence of chronic kidney disease (CKD) was significantly higher in patients with type 2 diabetes (p = 0.008), and CKD was significantly associated with BC in patients with type 2 diabetes (p = 0.032). A biomarker model, incorporating two serum (monocyte chemoattractant protein 1 and vascular endothelial growth factor) and three urine (interleukin 6, cytokeratin 18, and cytokeratin 8) proteins, predicted incidence of BC with an Area Under the Curve (AUC) of 0.84 in individuals with type 2 diabetes. In people without diabetes, the AUC was 0.66. CONCLUSIONS: We demonstrate the potential clinical utility of a biomarker panel, which includes proteins related to BC pathogenesis and type 2 diabetes, for monitoring risk of BC in patients with type 2 diabetes. Earlier urology referral of patients with type 2 diabetes will improve outcomes for these patients. TRIAL REGISTRATION: http://www.isrctn.com/ISRCTN25823942.

Screening attendance, prevalence and severity of diabetic retinopathy (DR) in a cohort of patients with diabetes mellitus secondary to chronic pancreatitis (DMsCP) in Northern Ireland.

INTRODUCTION: This study investigated Northern Ireland Diabetic Eye Screening Programme (NIDESP) attendance and diabetic retinopathy (DR) prevalence/severity in patients with diabetes mellitus secondary to chronic pancreatitis (PwDMsCP). RESEARCH DESIGN AND METHODS: Medical/NIDESP records for all PwDMsCP attending the pancreatic diabetes clinic were analyzed in 2017 (n=78) and 2019 (n=94). RESULTS: Between 2017 and 2019, those without DR decreased (76% to 63%); mild non-proliferative DR (NPDR), severe NPDR and PDR were found in 30%, 2% and 5%, respectively (previously 18%, 4%, 2%); diabetic maculopathy (DMac) was present in 12% (previously 10%). There was no significant difference between worst-eye DR/DMac grade and HbA1c, gender, body mass index, pancreatitis etiology and screening attendance (p>0.05). Patients with proliferative DR had longer diabetes and pancreatitis duration than DR-free patients (both p=0.001). CONCLUSIONS: DR prevalence was similar in PwDMsCP and patients with type 2 diabetes of similar disease duration. This work demonstrates the importance of reaching all patients for establishing DR severity reliably and to provide accessible, equitable care to PwDMsCP.