AI Smartphone App Detects Eye Cancer with Near-Specialist Accuracy, Study Finds

AI Smartphone App Detects Eye Cancer with Near-Specialist Accuracy, Study Finds

A smartphone-based Artificial Intelligence (AI) application has demonstrated the ability to detect cancers on the surface of the eye with near-specialist accuracy, offering hope for faster diagnosis and improved access to care for patients with potentially sight-threatening and life-threatening conditions.

The application, known as Capture-Tumor, uses advanced deep-learning technology to analyze photographs of the eye taken with a smartphone and identify signs of ocular surface malignancies. Researchers say the innovation could transform early cancer detection by enabling users to perform initial screenings from home before being referred to specialist care.

The findings were published in JAMA Ophthalmology following a non-randomized clinical trial led by researchers at Sun Yat-sen University, China, who evaluated the system’s effectiveness in real-world settings.

According to the researchers, the AI model was trained using more than 12 years of specialist ophthalmic images collected by eye care professionals. The system was initially developed using slit-lamp photographs taken in hospitals before being adapted to work with standard smartphone images captured by patients themselves.

The app includes built-in image-quality assessment tools and provides real-time instructions to help users take suitable photographs. Once captured, images are uploaded to a cloud-based platform where the AI analyzes them and flags suspicious lesions that may require specialist review.

The study involved 614 participants aged between four and 87 years, with a median age of 46. Participants were recruited through television campaigns, social media platforms and online hospital portals. Researchers analyzed 805 eye images from 535 participants included in the final assessment.

READ ALSO:

• NAJA Holds 3rd Auto Industry Summit July 30 on Nigeria’s EV, CNG Future
• EFCC to Arraign US-Blacklisted Lagos BDC Operator Over Alleged ISIS Financing
• US-Iran Ceasefire: Why Petrol Still Costs N1,200/Litre Despite Crude Crash to $70

To ensure diagnostic accuracy, images were matched against histopathological findings where available. In cases where tissue diagnosis was not possible, clinical examinations and telemedicine reviews were used to establish final diagnoses.

The primary objective was to determine how effectively the AI system could distinguish between malignant and benign eye surface lesions.

Results showed impressive performance. During controlled testing using specialist clinical images, the model achieved an Area Under the Curve (AUC) score of 0.945, indicating a high level of diagnostic accuracy. When deployed in real-world conditions using smartphone photographs and in-app guidance, performance improved further, reaching an AUC of 0.977.

The system recorded a sensitivity rate of 89.3 percent, meaning it correctly identified most cancer cases, while its specificity reached 95.9 percent, demonstrating a strong ability to rule out non-cancerous conditions.

One of the study’s most significant findings was the app’s ability to identify previously undiagnosed cancers.

The AI platform generated 58 referrals to specialist centers, resulting in the confirmation of 20 malignant cases through histopathological examination. Remarkably, 19 of those cancers had not been previously diagnosed, highlighting the technology’s potential role in detecting disease earlier than traditional referral pathways.

Researchers noted that none of the newly diagnosed patients required removal of the eye or surrounding orbital tissue, suggesting that earlier intervention may have contributed to more favorable treatment outcomes.

The study also found that the app significantly streamlined access to specialist care. Before using the AI system, patients required an average of 3.69 referrals before reaching definitive treatment. Following implementation of the technology, that figure dropped dramatically to just 1.02 referrals, representing a major improvement in efficiency and patient access.

READ ALSO:

• Police Dismiss Viral Katsina Video, Say Armed Men Are Vigilantes, Hunters
• Jetour Puts Customer Peace of Mind at the Centre of X70 PHEV Rollout
• Mike Ozekhome Barred from Using SAN Title as LPPC Orders Suspension

Experts believe this could reduce delays that often occur when patients move between multiple healthcare providers before receiving an accurate diagnosis.

The researchers further suggested that widespread adoption of smartphone-based screening could substantially increase the number of eye cancer cases detected and treated at specialist centers. However, they cautioned that these projections require additional validation through larger studies.

An accompanying editorial described Capture-Tumor as a pioneering “closed-loop” healthcare model that combines public awareness, AI-assisted screening, specialist triage and targeted referrals within a single platform.

Experts said the study provides an important proof of concept for using AI and mobile technology to decentralize screening for rare diseases, particularly in regions where access to specialist ophthalmologists remains limited.

Eye surface cancers, collectively known as ocular surface squamous neoplasia (OSSN) and related malignancies, can be difficult to diagnose in their early stages because symptoms often resemble less serious eye conditions. Delayed diagnosis can lead to vision impairment, extensive surgery and, in severe cases, cancer spread.

The emergence of smartphone-based AI screening tools could therefore play a critical role in reducing diagnostic delays and improving patient outcomes, especially in low-resource settings where specialist care is scarce.

Despite the promising findings, researchers acknowledged several limitations. Most participants were of Chinese origin, meaning further studies are needed to determine how well the technology performs across diverse ethnic and geographic populations. They also noted that older users may face challenges using smartphone-based screening tools and that the study primarily assessed short-term diagnostic outcomes rather than long-term clinical benefits.

The team emphasized that the application is intended to support—not replace—medical professionals. Any suspicious findings identified by the app would still require confirmation through clinical examination and specialist evaluation.

As healthcare systems increasingly adopt AI-powered diagnostic tools, experts say technologies such as Capture-Tumor could help bridge gaps in access to specialist care, improve early cancer detection and potentially save sight and lives through timely intervention.

AI Smartphone App Detects Eye Cancer with Near-Specialist Accuracy, Study Finds