{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,7,20]],"date-time":"2026-07-20T11:36:54Z","timestamp":1784547414496,"version":"3.55.0"},"posted":{"date-parts":[[2025,11,17]]},"group-title":"Ophthalmology","reference-count":43,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2025,11,17]],"date-time":"2025-11-17T00:00:00Z","timestamp":1763337600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.medrxiv.org\/about\/FAQ#license"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2025,11,17]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Globally, cataract remains the leading cause of blindness, affecting over 100 million people, with a disproportionate burden in low- and middle-income countries where access to ophthalmologists is limited. Although cataract surgery can restore vision almost immediately, timely diagnosis and referral remains a major barrier to care. We developed and prospectively evaluated lightweight, multimodal machine learning models capable of classifying lens status on a smartphone, enabling accessible screening in low-resource environments. We trained and evaluated both early and late fusion model architectures to classify lens status as clear, immature cataract, mature cataract, or pseudophakia using 6,794 anterior segment eye images captured using Scout\u2122 smartphone-based diffuse illumination system paired with clinical data (age, visual acuity, and pinhole acuity) from 2,956 patients from three eye hospitals in India. The early fusion model, which jointly integrates image and clinical features via a learnable gating mechanism, achieved superior performance (AUROC=0.98) compared to late fusion. Model interpretation using feature importance and Grad-CAM revealed that early fusion effectively balanced visual and clinical parameters, mirroring ophthalmologist diagnostic reasoning. Prospective on-device evaluation in 210 patients at the Aravind Eye Hospital demonstrated equivalent performance, achieving an AUROC of 0.96, confirming robustness and real-time feasibility on mobile hardware. These results demonstrate the first prospectively validated, on-device, multimodal cataract machine learning model, demonstrating the feasibility of instant, offline cataract classification and referral in low-resource environments. This advance has potential to broaden cataract screening, allowing minimally trained workers to screen and refer patients, and enabling earlier diagnosis, referral, and treatment in underserved populations.<\/jats:p>","DOI":"10.1101\/2025.11.14.25340278","type":"posted-content","created":{"date-parts":[[2025,11,18]],"date-time":"2025-11-18T07:45:17Z","timestamp":1763451917000},"source":"Crossref","is-referenced-by-count":1,"title":["Development and Clinical Validation of Lightweight, Multimodal Machine Learning Models for Smartphone-Based Cataract Detection and Classification"],"prefix":"10.64898","author":[{"given":"Vishnu","family":"Kannan","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jordan","family":"Shuff","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Arushi","family":"Acharya","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rengaraj","family":"Venkatesh","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nakul S.","family":"Shekhawat","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3483-7429","authenticated-orcid":false,"given":"Kunal S.","family":"Parikh","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2025111917502800000_2025.11.14.25340278v1.1","doi-asserted-by":"publisher","DOI":"10.1016\/S2214-109X(20)30489-7"},{"key":"2025111917502800000_2025.11.14.25340278v1.2","doi-asserted-by":"publisher","DOI":"10.1097\/ICU.0b013e3283414fc8"},{"key":"2025111917502800000_2025.11.14.25340278v1.3","doi-asserted-by":"publisher","DOI":"10.1016\/j.ophtha.2014.05.014"},{"key":"2025111917502800000_2025.11.14.25340278v1.4","doi-asserted-by":"publisher","DOI":"10.1016\/S2214-109X(20)30488-5"},{"key":"2025111917502800000_2025.11.14.25340278v1.5","doi-asserted-by":"publisher","DOI":"10.1016\/j.ophtha.2010.05.020"},{"key":"2025111917502800000_2025.11.14.25340278v1.6","doi-asserted-by":"publisher","DOI":"10.4103\/ijo.IJO_1235_18"},{"key":"2025111917502800000_2025.11.14.25340278v1.7","doi-asserted-by":"publisher","DOI":"10.1136\/bjophthalmol-2011-301378"},{"key":"2025111917502800000_2025.11.14.25340278v1.8","doi-asserted-by":"publisher","DOI":"10.1001\/jamaophthalmol.2018.5449"},{"key":"2025111917502800000_2025.11.14.25340278v1.9","doi-asserted-by":"publisher","DOI":"10.1016\/j.amjmed.2015.06.004"},{"key":"2025111917502800000_2025.11.14.25340278v1.10","doi-asserted-by":"publisher","DOI":"10.1055\/a-2003-2369"},{"key":"2025111917502800000_2025.11.14.25340278v1.11","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pgph.0004619"},{"key":"2025111917502800000_2025.11.14.25340278v1.12","doi-asserted-by":"publisher","DOI":"10.1167\/tvst.10.8.29"},{"key":"2025111917502800000_2025.11.14.25340278v1.13","doi-asserted-by":"publisher","DOI":"10.1136\/bjo-2023-324488"},{"key":"2025111917502800000_2025.11.14.25340278v1.14","doi-asserted-by":"publisher","DOI":"10.1109\/TBME.2015.2444389"},{"key":"2025111917502800000_2025.11.14.25340278v1.15","doi-asserted-by":"publisher","DOI":"10.1136\/bjophthalmol-2019-315651"},{"key":"2025111917502800000_2025.11.14.25340278v1.16","doi-asserted-by":"publisher","DOI":"10.1097\/01.Apo.0000656988.16221.04"},{"key":"2025111917502800000_2025.11.14.25340278v1.17","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-021-24116-6"},{"key":"2025111917502800000_2025.11.14.25340278v1.18","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-022-30728-3"},{"key":"2025111917502800000_2025.11.14.25340278v1.19","article-title":"Challenges and obstacles towards deploying deep learning models on mobile devices","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.20","first-page":"493","article-title":"The physical basis for transparency of the crystalline lens","volume":"1","year":"1962","journal-title":"Invest Ophthalmol"},{"key":"2025111917502800000_2025.11.14.25340278v1.21","doi-asserted-by":"publisher","DOI":"10.1038\/nrdp.2015.14"},{"key":"2025111917502800000_2025.11.14.25340278v1.22","doi-asserted-by":"publisher","DOI":"10.18632\/aging.101208"},{"key":"2025111917502800000_2025.11.14.25340278v1.23","doi-asserted-by":"publisher","DOI":"10.1167\/iovs.05-0981"},{"key":"2025111917502800000_2025.11.14.25340278v1.24","article-title":"Comparative analysis of lightweight deep learning models for memory-constrained devices","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.25","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2019.04.009"},{"key":"2025111917502800000_2025.11.14.25340278v1.26","doi-asserted-by":"publisher","DOI":"10.1145\/3620665.3640366"},{"key":"2025111917502800000_2025.11.14.25340278v1.27","article-title":"Roboflow100-VL: a multi-domain object detection benchmark for vision-language models","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.28","article-title":"Comparative Study of Machine Learning Algorithms in Detecting Cardiovascular Diseases","year":"2024","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.29","doi-asserted-by":"publisher","DOI":"10.1007\/s10654-018-0390-z"},{"key":"2025111917502800000_2025.11.14.25340278v1.30","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2018.07.011"},{"key":"2025111917502800000_2025.11.14.25340278v1.31","article-title":"Keeping Deep Learning Models in Check: A History-Based Approach to Mitigate Overfitting","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.32","doi-asserted-by":"publisher","DOI":"10.1007\/s11222-012-9370-4"},{"key":"2025111917502800000_2025.11.14.25340278v1.33","doi-asserted-by":"publisher","DOI":"10.3390\/e24050687"},{"key":"2025111917502800000_2025.11.14.25340278v1.34","article-title":"Improving the gating mechanism of recurrent neural networks","journal-title":"arXiv preprint"},{"key":"2025111917502800000_2025.11.14.25340278v1.35","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-019-01228-7"},{"key":"2025111917502800000_2025.11.14.25340278v1.36","doi-asserted-by":"publisher","DOI":"10.1136\/bmjopen-2024-089741"},{"key":"2025111917502800000_2025.11.14.25340278v1.37","doi-asserted-by":"publisher","DOI":"10.1007\/s00521-019-04051-w"},{"key":"2025111917502800000_2025.11.14.25340278v1.38","doi-asserted-by":"publisher","DOI":"10.1001\/archophthalmol.2009.113"},{"key":"2025111917502800000_2025.11.14.25340278v1.39","doi-asserted-by":"publisher","DOI":"10.1097\/j.jcrs.0000000000001318"},{"key":"2025111917502800000_2025.11.14.25340278v1.40","doi-asserted-by":"publisher","DOI":"10.1097\/IIO.0b013e31821e58aa"},{"key":"2025111917502800000_2025.11.14.25340278v1.41","doi-asserted-by":"publisher","DOI":"10.1186\/s40662-021-00273-z"},{"key":"2025111917502800000_2025.11.14.25340278v1.42","doi-asserted-by":"publisher","DOI":"10.1186\/s40662-020-00183-6"},{"key":"2025111917502800000_2025.11.14.25340278v1.43","doi-asserted-by":"publisher","DOI":"10.1038\/s41433-024-03070-9"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1101\/2025.11.14.25340278","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,15]],"date-time":"2026-01-15T20:51:20Z","timestamp":1768510280000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.1101\/2025.11.14.25340278"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,17]]},"references-count":43,"URL":"https:\/\/doi.org\/10.1101\/2025.11.14.25340278","relation":{},"subject":[],"published":{"date-parts":[[2025,11,17]]},"subtype":"preprint"}}