Retinal Lesion Detection via Machine Teaching for Enhanced OCT Analysis

NU 2025-111

INVENTORS

• Rukhsana Mirza* (No affiliation provided)
• Donna Hooshmand
• Emma Alexander
• Hayden Sikora
• Jay Bisen
• Kristian Hammond
• Laura Machlab
• Michael Drakopoulos
• Paul Bryar

SHORT DESCRIPTION

This machine teaching algorithm automatically identifies retinal ischemic perivascular lesions on OCT images. It reduces manual workload and improves consistency in ophthalmic diagnostics.

BACKGROUND

Manual counting of retinal lesions is labor-intensive and prone to variability. Clinicians and researchers struggle with inconsistent assessments when identifying potential ocular biomarkers. This unreliability hampers timely diagnosis and slows research into systemic diseases linked to retinal changes.

ABSTRACT

The invention employs machine teaching techniques to detect retinal ischemic perivascular lesions (RIPLs) on OCT scans. A trained algorithm scans images to identify subtle retinal features with improved accuracy. It tackles the challenges of low-prevalence datasets and reduces manual error. Preliminary laboratory results confirm consistent lesion detection compared to conventional manual methods.

MARKET OPPORTUNITY

The global market for AI in ophthalmology is experiencing explosive growth, valued at $209.2 million in 2024 and projected to surge to $1.36 billion by 2030 at an exceptional CAGR of 36.79% (Source: Research and Markets, 2025). This growth is driven by the rising prevalence of chronic eye diseases and the urgent need for more efficient diagnostic solutions.

DEVELOPMENT STAGE

TRL-4 - Prototype Validated in Lab: The integrated system has demonstrated key functions by analyzing OCT images in a controlled laboratory setting.

APPLICATIONS

• Ophthalmic clinics: Rapidly identify retinal lesions to guide clinical decision-making.
• Research laboratories: Automate lesion detection to facilitate analysis of large imaging datasets.

ADVANTAGES

• Eliminates manual counting: Automates detection to reduce clinician workload.
• Reduces variability: Delivers consistent and objective analysis versus manual methods.
• Accelerates research: Enables rapid processing of large datasets.
• Enhances scalability: Adapts seamlessly to both clinical and research imaging systems.

PUBLICATIONS

• Rukhsana Mirza et al, Machine Teaching Allows for Rapid Development of Automated Systems for Retinal Lesion Detection From Small Image Datasets, Ophthalmic Surgery, Lasers and Imaging Retina, May 01, 2024

IP STATUS
US Patent Pending