New AI Smartphone App May be Useful in the Detection of Ocular Surface Tumors

CaptureTumor is a smartphone application currently being developed for at-home detection of malignant ocular surface tumors. The researchers and developers present this model as “a potentially replicable and scalable strategy to advance resource allocation and equity in the management of rare diseases.” Click image to enlarge. Eyelid and ocular surface lumps and bumps are a perennial source of worry for patients, even though most turn out to be benign. Given the importance of early detection of true malignancies, developing broadly accessible screening methods is a priority. Researchers from China conducted a study to validate a machine-learning model that determines malignant or benign ocular surface lesions.1In a paper published in JAMA Ophthalmology, the researchers presented their findings on a smartphone app called CaptureTumor (CaT). This program uses a deep-learning model trained and validated with 12 years of slit lamp images. The app was shared nationwide in China to conduct a real-world implementation trial.A total of 614 subjects completed self-screening through the app. During the course of the study, 20 malignant tumors were identified by the model. Nineteen out of 20 of these individuals were newly diagnosed.Compared to slit lamp evaluation, the area under the receiver operating characteristic curve (AUC)—a measure of predictive ability—for CaT was similar. Under controlled conditions, the model boasted an AUC of 0.905, slightly lower than slit lamp’s AUC of 0.945. Once the model was publicly available, the AUC of CaT increased to 0.977, with a sensitivity of 89.3% and a specificity of 95.9%.“Conventional hospital-based care models struggle to implement effective large-scale screening for rare ocular surface malignancies, constrained by low disease prevalence and limited subspecialty capacity,” explained the authors in their paper. “This mobile health model offers a potentially scalable, accessible and affordable strategy for early detection of rare vision- and life-threatening diseases.”However, CaT is not without its limitations. Several were described in this paper, then restated in an invited commentary also published in JAMA Ophthalmology.2According to the commentary, three development concerns should be addressed to move forward with a successful AI screening tool. First, developers need to address the concern over representation, as the study was limited to a Chinese population. Second, developers should consider who within a given population will engage in this tool, as not all individuals can withstand the focus, framing and exposure required for smartphone imaging. And third, since this app targeted patients screened for ocular surface lesions, developers should figure out how to expand the AI’s intuition for the masses to mitigate false-positive referrals.Besides the development of the application for real-world application, the commentary’s authors had a major concern regarding one of the study’s results. CaT had a sensitivity of 89.3%, which sounds high. However, the commentary’s authors pointed out that this figure suggests that one in 10 malignant cases was misclassified as normal.“None of this diminishes the impressive achievement of [the study’s authors],” stated the commentary’s authors. “Finding 20 confirmed malignancies through a mobile app study is a meaningful step. The challenges of race, representation and real-world deployment are tractable ones, and this study provides the foundation from which to address them.”Both the study’s and commentary’s authors agree that further validation and long-term assessment of CaT are needed before it can influence patient outcomes and health systems worldwide.Click here for the journal source and here for the commentary. 1. Wang R, Bi S, Lin D, et al. Smartphone-based proactive self-screening for ocular surface malignancies: A nonrandomized clinical trial. JAMA Ophthalmol. June 4, 2026. [Epub ahead of print].2. Manjaly C, Lee AY. Mobile-based artificial intelligence and ocular surface malignancies. JAMA Ophthalmol. June 4, 2026. [Epub ahead of print].This article was developed by the editorial staff in conjunction with experts in the field. In the process, AI may have been among the editorial tools used to meet the goals of human editors, who approved all content.