Data-Centric Label Smoothing for Explainable Glaucoma Screening from Eye Fundus Images

As current computing capabilities increase, modern machine learning and computer vision system tend to increase in complexity, mostly by means of larger models and advanced optimization strategies. Although often neglected, in many problems there is also much to be gained by considering potential improvements in understanding and better leveraging already-available training data, including annotations. This so-called data-centric approach can lead to substantial performance increases, sometimes beyond what can be achieved by larger models. In this paper we adopt such an approach for the task of justifiable glaucoma screening from retinal images. In particular, we focus on how to combine information from multiple annotators of different skills into a tailored label smoothing scheme that allows us to better employ a large collection of fundus images, instead of discarding samples suffering from inter-rater variability. Internal validation results indicate that our bespoke label smoothing approach surpasses the performance of a standard resnet50 model and also the same model trained with conventional label smoothing techniques, in particular for the multi-label scenario of predicting clinical reasons of glaucoma likelihood in a highly imbalanced screening context. Our code is made available at github.com/agaldran/justraigs .

[1]  H. A. Serhan,et al.  Domain Adaptation-Based Deep Learning Model for Forecasting and Diagnosis of Glaucoma Disease. , 2024, Biomedical signal processing and control.

[2]  Rashmiranjan Nayak,et al.  A comprehensive review of datasets for detection and localization of video anomalies: a step towards data-centric artificial intelligence-based video anomaly detection , 2023, Multim. Tools Appl..

[3]  Aditya Thyagarajan,et al.  ObjectLab: Automated Diagnosis of Mislabeled Images in Object Detection Data , 2023, ArXiv.

[4]  Niccolò McConnell,et al.  Exploring advanced architectural variations of nnUNet , 2023, Neurocomputing.

[5]  Matthew B. Blaschko,et al.  A generalizable deep learning regression model for automated glaucoma screening from fundus images , 2023, npj Digit. Medicine.

[6]  Ismail Ben Ayed,et al.  Calibrating segmentation networks with margin-based label smoothing , 2023, Medical Image Anal..

[7]  Zaid Pervaiz Bhat,et al.  Data-centric Artificial Intelligence: A Survey , 2023, ArXiv.

[8]  G. Carneiro,et al.  Multi-Head Multi-Loss Model Calibration , 2023, MICCAI.

[9]  H. Lemij,et al.  Characteristics of a Large, Labeled Data Set for the Training of Artificial Intelligence for Glaucoma Screening with Fundus Photographs , 2023, Ophthalmology science.

[10]  F. Isensee,et al.  Extending nnU-Net is all you need , 2022, Bildverarbeitung für die Medizin.

[11]  A. Sauter,et al.  TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images. , 2022, Radiology. Artificial intelligence.

[12]  S. Cherry,et al.  Fully Automated, Semantic Segmentation of Whole-Body 18F-FDG PET/CT Images Based on Data-Centric Artificial Intelligence , 2022, The Journal of Nuclear Medicine.

[13]  R. Weinreb,et al.  Performances of machine learning in detecting glaucoma using fundus and retinal optical coherence tomography images: A meta-analysis. , 2021, American journal of ophthalmology.

[14]  Ross Wightman,et al.  ResNet strikes back: An improved training procedure in timm , 2021, ArXiv.

[15]  Bjoern H Menze,et al.  The Medical Segmentation Decathlon , 2021, Nature Communications.

[16]  Matthew B. Blaschko,et al.  Deep learning on fundus images detects glaucoma beyond the optic disc , 2021, Scientific Reports.

[17]  Jens Petersen,et al.  nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation , 2020, Nature Methods.

[18]  Siamak Yousefi,et al.  Predicting glaucoma prior to its onset using deep learning , 2019, bioRxiv.

[19]  Isaac L. Chuang,et al.  Confident Learning: Estimating Uncertainty in Dataset Labels , 2019, J. Artif. Intell. Res..

[20]  Pedro Costa,et al.  A No-Reference Quality Metric for Retinal Vessel Tree Segmentation , 2018, MICCAI.

[21]  Adrian V. Dalca,et al.  Medical Image Imputation From Image Collections , 2018, IEEE Transactions on Medical Imaging.

[22]  Lukasz Kaiser,et al.  Attention is All you Need , 2017, NIPS.

[23]  Xiaoxiao Li,et al.  REFUGE Challenge: A Unified Framework for Evaluating Automated Methods for Glaucoma Assessment from Fundus Photographs , 2019, Medical Image Anal..