{ "schema_version": "0.1.0-research-proposals", "updated": "2026-10-28", "clinical_validation": false, "scored_new_cases": 0, "threshold_status": "not calibrated; report raw metrics and validity, not a composite leaderboard", "search_method": "Targeted search and primary-source verification using Ego Lite Browser; not systematic review", "sources": [ { "id": "R1", "authors_short": "Sheng et al.", "year": 2025, "title": "Using Artificial Intelligence to Predict Implantable Collamer Lens Vault: A Low Parameter-Dependent Model for Better Surgical Outcomes", "doi": "10.1167/tvst.14.9.32", "url": "https://pubmed.ncbi.nlm.nih.gov/40985791/", "verification_level": "abstract" }, { "id": "R2", "authors_short": "Yang et al.", "year": 2024, "title": "Clinical decision support system based on deep learning for evaluating implantable collamer lens size and vault after implantable collamer lens surgery: a retrospective study", "doi": "10.1136/bmjopen-2023-081050", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10875548/", "verification_level": "full_text_selected_sections" }, { "id": "R3", "authors_short": "Li et al.", "year": 2026, "title": "EVO-ICL Vault Prediction: A Data Wrangling Framework Integrating Multicenter Big Data and Machine Learning", "doi": "10.1007/s40123-025-01288-8", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC12882899/", "verification_level": "full_text_selected_sections" }, { "id": "R4", "authors_short": "Yang, Fu, Heidrich", "year": 2024, "title": "Curriculum learning for ab initio deep learned refractive optics", "doi": "10.1038/s41467-024-50835-7", "url": "https://www.nature.com/articles/s41467-024-50835-7", "verification_level": "publisher_abstract_introduction_and_author_method_summary" }, { "id": "R5", "authors_short": "Ntatsis et al.", "year": 2025, "title": "Deep learning ocular aberration retrieval from simulated retinal images under straylight", "doi": "10.1364/BOE.559749", "url": "https://pubmed.ncbi.nlm.nih.gov/40677828/", "verification_level": "abstract" }, { "id": "R6", "authors_short": "Chen et al.", "year": 2019, "title": "Evaluation of Disk Halo Size after Implantation of a Collamer Lens with a Central Hole (ICL V4c)", "doi": "10.1155/2019/7174913", "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC6710753/", "verification_level": "full_text_measurements_and_results" } ], "cases": [ { "id": "AI-01", "title": "跨中心拱高与不确定度", "track": "anatomy", "sources": [ "R1", "R2", "R3" ], "status": "requires_authorized_clinical_dataset", "input": [ "preoperative_biometry_with_units", "missingness_mask", "implanted_lens_identity", "followup_time" ], "output": [ "vault_um", "prediction_interval_90_um", "missing_input_policy" ], "metrics": [ "MAE_um", "RMSE_um", "interval_coverage", "interval_width_um", "per_center_errors" ], "split": "patient grouped, external center held out, preprocessing fitted on training only", "negative_controls": [ "training_median", "no_uncertainty", "patient_eye_leakage_audit" ] }, { "id": "AI-02", "title": "物理环境与隐藏干预", "track": "physics", "sources": [ "R4" ], "status": "public_synthetic_development_examples_only", "input": [ "explicit_synthetic_surfaces_and_media", "pupil", "source_angles_and_flux", "detector_and_units" ], "output": [ "executable_solver", "path_resolved_linear_fields", "energy_ledger", "sampling_and_error_budget" ], "metrics": [ "energy_relative_error", "fixed_coordinate_centroid_error_deg", "radial_energy_error", "intervention_difference_error", "invalid_rate" ], "split": "hidden parameter combinations and independent reference solver required", "negative_controls": [ "fixed_ring_overlay", "exposure_changes_physical_values", "per_condition_retina_refocus" ], "public_examples": [ "preop", "night", "small", "wide", "noport", "decenter" ] }, { "id": "AI-03", "title": "受约束逆向推理与梯度", "track": "physics_inverse", "sources": [ "R4" ], "status": "requires_gradient_validation_and_hidden_set", "input": [ "multi_condition_linear_psfs", "known_geometry", "unknown_parameter_bounds" ], "output": [ "candidate_parameters_with_units", "gradients", "forward_reprojection", "ambiguity_statement" ], "metrics": [ "smooth_region_gradient_error_vs_independent_finite_difference", "heldout_forward_error", "geometry_invalid_rate", "identifiable_parameter_error" ], "split": "hold out observations and geometry families; exclude path discontinuities from smooth derivative claim", "negative_controls": [ "surface_intersection", "unbounded_fit", "single_condition_overfit" ] }, { "id": "AI-04", "title": "像差与杂散光辨析", "track": "physics_inverse", "sources": [ "R5" ], "status": "requires_explicit_scatter_truth_generator", "input": [ "linear_psf", "noise_model", "dynamic_range", "capture_method", "pupil_geometry" ], "output": [ "wavefront_candidates", "scatter_candidates", "uncertainty", "forward_reprojection" ], "metrics": [ "wavefront_RMS_without_piston", "forward_psf_residual", "OOD_error", "equivalent_solution_coverage" ], "split": "separate scatter model families and aberration ranges", "negative_controls": [ "high_confidence_unique_cause_on_ambiguous_input", "symmetric_pupil_positive_negative_defocus_pair" ] }, { "id": "AI-05", "title": "光学代理模型外推", "track": "physics_surrogate", "sources": [ "R4" ], "status": "project_proposal_requires_hidden_geometry_and_reference", "input": [ "training_geometry_parameters", "training_path_resolved_linear_fields", "hidden_test_parameters" ], "output": [ "linear_field_prediction", "path_energy", "uncertainty", "training_cost_and_inference_time_and_hardware" ], "metrics": [ "raw_power_error", "centroid_error_deg", "radial_energy_error", "conservation_error", "worst_group_error", "runtime" ], "split": "geometry family and parameter combination split; exposures of same field stay together", "negative_controls": [ "nearest_neighbor", "analytic_fixed_ring", "tone_mapped_image_only" ] }, { "id": "AI-06", "title": "测量语义与患者盲预测", "track": "perception", "sources": [ "R6" ], "status": "synthetic_reports_only_requires_pilot_and_authorized_data", "input": [ "allowed_eye_measurements", "illumination_and_geometry", "postoperative_time", "measurement_endpoint" ], "output": [ "presence_probability", "halo_features_in_deg", "prediction_intervals", "endpoint_definition" ], "metrics": [ "existing_patient_layer_subscores", "test_retest_noise", "patient_macro_score", "heldout_patient_error" ], "split": "patient grouped across eyes, conditions and visits; no target drawings provided to agent", "negative_controls": [ "disk_halo_equals_bright_ring", "burden_equals_psf_energy", "always_one_fixed_ring" ] } ] }