Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lirot.ai: A Novel Platform for Crowd-Sourcing Retinal Image Segmentations

Jonathan Fhima, Jan Van Eijgen, Moti Freiman, Ingeborg Stalmans, Joachim A. Behar

TL;DR

The paper addresses the bottleneck of obtaining large annotated datasets for supervised deep learning in medical imaging by introducing Lirot.ai, a three-component platform (iPad app, Firebase backend, and Python API) designed for simultaneous remote annotation and centralized data management. Key contributions include an Apple Pencil–friendly iPadOS app, pre-segmentation, automated image quality assessment via FundusQ-Net, senior-annotator quality control with version history, and an integrated active-learning pipeline to optimize image selection. The authors demonstrate feasibility by building a retinal Fundus dataset of vasculature segmentations with multiple annotators, illustrating faster, scalable data collection and streamlined data transfer. This platform has practical impact in accelerating dataset generation for medical image segmentation and enabling iterative improvements through active learning.

Abstract

Introduction: For supervised deep learning (DL) tasks, researchers need a large annotated dataset. In medical data science, one of the major limitations to develop DL models is the lack of annotated examples in large quantity. This is most often due to the time and expertise required to annotate. We introduce Lirot. ai, a novel platform for facilitating and crowd-sourcing image segmentations. Methods: Lirot. ai is composed of three components; an iPadOS client application named Lirot. ai-app, a backend server named Lirot. ai-server and a python API name Lirot. ai-API. Lirot. ai-app was developed in Swift 5.6 and Lirot. ai-server is a firebase backend. Lirot. ai-API allows the management of the database. Lirot. ai-app can be installed on as many iPadOS devices as needed so that annotators may be able to perform their segmentation simultaneously and remotely. We incorporate Apple Pencil compatibility, making the segmentation faster, more accurate, and more intuitive for the expert than any other computer-based alternative. Results: We demonstrate the usage of Lirot. ai for the creation of a retinal fundus dataset with reference vasculature segmentations. Discussion and future work: We will use active learning strategies to continue enlarging our retinal fundus dataset by including a more efficient process to select the images to be annotated and distribute them to annotators.

Lirot.ai: A Novel Platform for Crowd-Sourcing Retinal Image Segmentations

TL;DR

The paper addresses the bottleneck of obtaining large annotated datasets for supervised deep learning in medical imaging by introducing Lirot.ai, a three-component platform (iPad app, Firebase backend, and Python API) designed for simultaneous remote annotation and centralized data management. Key contributions include an Apple Pencil–friendly iPadOS app, pre-segmentation, automated image quality assessment via FundusQ-Net, senior-annotator quality control with version history, and an integrated active-learning pipeline to optimize image selection. The authors demonstrate feasibility by building a retinal Fundus dataset of vasculature segmentations with multiple annotators, illustrating faster, scalable data collection and streamlined data transfer. This platform has practical impact in accelerating dataset generation for medical image segmentation and enabling iterative improvements through active learning.

Abstract

Introduction: For supervised deep learning (DL) tasks, researchers need a large annotated dataset. In medical data science, one of the major limitations to develop DL models is the lack of annotated examples in large quantity. This is most often due to the time and expertise required to annotate. We introduce Lirot. ai, a novel platform for facilitating and crowd-sourcing image segmentations. Methods: Lirot. ai is composed of three components; an iPadOS client application named Lirot. ai-app, a backend server named Lirot. ai-server and a python API name Lirot. ai-API. Lirot. ai-app was developed in Swift 5.6 and Lirot. ai-server is a firebase backend. Lirot. ai-API allows the management of the database. Lirot. ai-app can be installed on as many iPadOS devices as needed so that annotators may be able to perform their segmentation simultaneously and remotely. We incorporate Apple Pencil compatibility, making the segmentation faster, more accurate, and more intuitive for the expert than any other computer-based alternative. Results: We demonstrate the usage of Lirot. ai for the creation of a retinal fundus dataset with reference vasculature segmentations. Discussion and future work: We will use active learning strategies to continue enlarging our retinal fundus dataset by including a more efficient process to select the images to be annotated and distribute them to annotators.
Paper Structure (12 sections, 4 figures)

This paper contains 12 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Methods
  3. Lirot.ai-app
  4. Active Learning
  5. Pre-segmentation
  6. Image quality assessment
  7. Quality control
  8. Lirot.ai-server
  9. Lirot.ai-API
  10. Vasculature segmentation DFI
  11. Results
  12. Discussion and conclusion

Figures (4)

  • Figure 1: Pipeline of the Lirot.ai platform: a dataset to annotate is defined by the research team and uploaded to the Lirot.ai-server. The annotators can download and segment images on their iPad. When finished, they upload the images to the Lirot.ai-server.
  • Figure 2: (A): User interface of Lirot.ai-app during the segmentation of the arterioles and venules of DFI. (B): Annotator segmenting a DFI with Lirot.ai-app in the UZ Leuven hospital.
  • Figure 3: User interface of Lirot.ai-app during the segmentation of the arterioles and venules of DFI. (A): Zoomed DFI, (B): Pre-segmentation generated by our baseline model
  • Figure 4: Example of generated DFIs segmentation using Lirot.ai