Skip to content

Multi-modal Deep Learning Based Intelligent Assisted Diagnosis and Portable Adaptation for Common Blinding Corneal Diseases

Multi-modal Deep Learning Based Intelligent Assisted Diagnosis and Portable Adaptation for Common Blinding Corneal Diseases

Status
Active, not recruiting
Phases
Unknown
Study type
Observational
Source
ChiCTR
Registry ID
ChiCTR2500115689
Enrollment
Unknown
Registered
2025-12-30
Start date
2026-01-01
Completion date
Unknown
Last updated
2026-01-05

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

Seven Common Causes of Blindness-Causing Corneal Diseases

Interventions

Gold Standard:Infectious keratitis is diagnosed using pathogen examination as the gold standard, combined with the comprehensive clinical diagnosis based on typical symptoms and signs assessed by expe
Index test:A deep learning-based auxiliary diagnostic algorithm for blinding corneal diseases

Sponsors

Eye Hospital, Wenzhou Medical University
Lead Sponsor

Eligibility

Sex/Gender
All

Inclusion criteria

Inclusion criteria: 1. Infectious keratitis (meeting any of the following diagnostic criteria): (1) Bacterial keratitis: Positive corneal scrapings smear or culture, or clinical manifestations consistent with bacterial keratitis characteristics (acute onset, severe pain, relatively well-defined stromal infiltrate, significant anterior chamber reaction, etc.), as comprehensively determined by a corneal specialist; (2) Fungal keratitis: Filaments/spores observed in scrapings smears or positive culture, or typical slit-lamp signs such as feathered infiltrates, satellite lesions, immune ring, etc., as determined by a specialist; (3) Acanthamoeba keratitis: Typical Acanthamoeba cysts/trophozoites found by confocal microscopy, or positive corneal scrapings or culture; (4) Viral keratitis (HSV/VZV): HSV: Typical dendritic/map-like epithelial defects or stromal keratitis, combined with PCR results if necessary; VZV: Corneal involvement in combination with a history of facial/eyelid herpes; 2. Non-infectious keratitis: (1) Immune-related keratitis (such as Mooren’s ulcer, marginal keratitis, etc.), diagnosed comprehensively by a corneal specialist based on clinical manifestations and systemic conditions, excluding evidence of infection; exposure keratitis, neurotrophic/paralytic keratitis, filamentary keratitis, etc.; (2) Non-infectious keratitis following chemical/thermal burns (with secondary infection clearly excluded); 3. Corneal dystrophy: Epithelial or stromal dystrophy with typical family history or bilateral symmetrical clinical manifestations, with slit-lamp diffuse illumination clearly showing the lesion; 4. Corneal degeneration: Primary cases of degenerative corneal diseases with typical slit-lamp features, such as Terrien’s marginal degeneration. Other general conditions: Complete slit-lamp photograph images of the initial visit are obtainable, and the image quality meets the analysis requirements; clinical history records are relatively complete, allowing extraction of core consultation information.

Exclusion criteria

Exclusion criteria: 1. Corneal lesions with unknown causes or insufficient clinical information to make a clear diagnosis; 2. Confirmed mixed infections (e.g., fungal and bacterial) where it is difficult to distinguish the primary cause; 3. History of significant ocular surface surgery causing notable local anatomical changes (such as penetrating keratoplasty, extensive conjunctival flap coverage, or scarring after amniotic membrane transplantation that makes the lesion unassessable); 4. Corneal perforation, prolapse of intraocular contents, or secondary endophthalmitis at the time of visit; 5. Poor-quality slit-lamp photographs (evident out-of-focus, severe exposure issues, eyelid obstructing the main lesion) affecting assessment; 6. Missing key medical history information making it impossible to construct a structured history questionnaire; 7. Non-initial cases (to avoid lesions that have been significantly altered by treatment, only images and medical history from the first visit are included).

Design outcomes

Primary

MeasureTime frame
Area Under the ROC Curve;Accuracy;Sensitivity;Specificity;

Countries

China

Contacts

Public ContactWei Chen

Eye Hospital, Wenzhou Medical University

chenweimd@wmu.edu.cn+86 577 8839 6800

Outcome results

None listed

Source: ChiCTR (via WHO ICTRP) · Data processed: Feb 4, 2026