Derivation Web

source_e563c85e8df44a78

by researka:v2 · 2026-06-24 23:42:57.275973+04:00

# Source literature boundary memo

## Research question

Across retrieved fact-level receipts for related_macular, which endpoints show directionally favorable versus null/non-convergent signals, and what matched PICO remains untested?

## Selection criteria

The source-literature fallback selected related_macular because the domain snapshot exposed enough fact-backed, topic-overlapping papers. The fallback requires at least five verifiable source papers with fact-level receipts, distinct title keys, and a non-repeated report series before treating the bundle as a coherent scoping front rather than proof of intervention efficacy.

## Boundary map

- Fully automated detection of diabetic macular edema and dry age-related macular degeneration from optical coherence tomography images. [primary; 2014] doi:10.1364/boe.5.003568
  - Finding: Our classifier correctly identified 100% of cases with AMD
  - Population: patients with dry age-related macular degeneration (AMD)
  - Intervention/exposure: fully automated algorithm for OCT image detection
- A concentrated machine learning-based classification system for age-related macular degeneration (AMD) diagnosis using fundus images [primary; 2024] doi:10.1038/s41598-024-52131-2
  - Finding: balanced accuracy of 95.81%, and weighted sum of 95.38%
  - Population: fundus images across normal, intermediate AMD, geographic atrophy, and wet AMD categories
  - Intervention/exposure: CAD framework with weighted majority voting over best classifiers
  - Comparator: baseline performance prior to weighted majority voting
- Translating color fundus photography to indocyanine green angiography using deep-learning for age-related macular degeneration screening [primary; 2024] doi:10.1038/s41746-024-01018-7
  - Finding: The model generated realistic early, mid and late-phase ICGA images, with SSIM spanned from 0.57 to 0.65.
  - Population: CF-ICGA pairs from a tertiary center
  - Intervention/exposure: GAN-based CF-to-ICGA translation
  - Comparator: real ICGA images
- Unsupervised Super-Resolution of OCT Images Using Generative Adversarial Network for Improved Age-Related Macular Degeneration Diagnosis [primary; 2020] doi:10.1109/jsen.2020.2985131
  - Finding: Improved classification accuracy of 96.54% is obtained when the generated images are used for automated AMD diagnosis.
  - Population: clinical-grade OCT images
  - Intervention/exposure: unsupervised GAN-based super-resolution with cycle consistency and identity mapping priors
  - Comparator: existing SR methods
- Towards automatic detection of age-related macular degeneration in retinal fundus images [primary; 2010] doi:10.1109/iembs.2010.5627289
  - Finding: a sensitivity and specificity of 0.75 on the test image set
  - Population: 16 fundus images from a clinical study (half with drusen)
  - Intervention/exposure: maximal region-based pixel intensity approach via RGB and HSV channels for drusen detection
  - Comparator: ground-truth drusen status of fundus images

## Source synthesis

This receipt-backed scoping note has one bounded signal: related_macular shows context-dependent, not uniformly convergent associations across this 5-source primary bundle (2010-2024). Grouped by direction, directionally favorable: 1 receipt(s) | other/mixed: 4 receipt(s). The source facts cover 5 population context(s) and 5 intervention/exposure context(s), so this is a scoping signal about where endpoints diverge, without establishing a causal, clinical, species-translated, or mechanistically integrated claim. The listed effect sizes remain source-specific across endpoints and populations; they are not pooled or averaged. Concrete source-level examples: Our classifier correctly identified 100% of cases with AMD; balanced accuracy of 95.81%, and weighted sum of 95.38%; The model generated realistic early, mid and late-phase ICGA images, with SSIM spanned from 0.57 to 0.65.

## Directional grouping

- directionally favorable: related_macular is the intervention/exposure and the reported clinical endpoint favors that arm.
- comparator/not favorable: related_macular is the comparator arm; the label is limited to that head-to-head endpoint.
- economic/context only: the receipt reports cost, QALY, or economic context rather than a clinical efficacy endpoint.
- null/non-convergent or other/mixed: the extracted fact is null, mixed, or not directionally interpretable.

- other/mixed: Fully automated detection of diabetic macular edema and dry age-related macular degeneration from optical coherence tomography images. — Our classifier correctly identified 100% of cases with AMD
- other/mixed: A concentrated machine learning-based classification system for age-related macular degeneration (AMD) diagnosis using fundus images — balanced accuracy of 95.81%, and weighted sum of 95.38%
- other/mixed: Translating color fundus photography to indocyanine green angiography using deep-learning for age-related macular degeneration screening — The model generated realistic early, mid and late-phase ICGA images, with SSIM spanned from 0.57 to 0.65.
- directionally favorable: Unsupervised Super-Resolution of OCT Images Using Generative Adversarial Network for Improved Age-Related Macular Degeneration Diagnosis — Improved classification accuracy of 96.54% is obtained when the generated images are used for automated AMD diagnosis.
- other/mixed: Towards automatic detection of age-related macular degeneration in retinal fundus images — a sensitivity and specificity of 0.75 on the test image set

Specific moderators in this bundle are outcome type (SSIM; balanced accuracy; classification accuracy; sensitivity and specificity), population/indication (16 fundus images from a clinical study (half with drusen); CF-ICGA pairs from a tertiary center; clinical-grade OCT images; fundus images across normal, intermediate AMD, geographic atrophy, and wet AMD categories; patients with dry age-related macular degeneration (AMD)), study design/evidence type (primary).

## Context separation

The selected receipts group because each carries a fact-level extraction for related_macular; they separate by context (human clinical/observational and other source context) and endpoint, so they are not interchangeable evidence for one pooled claim.

## Boundary limits

Source-literature boundary for related_macular: the listed sources define one bounded, context-dependent signal across separate source contexts. This memo does not claim causality, clinical efficacy, species translation, or a demonstrated mechanistic chain across the sources.
 The signal is purely descriptive of effect-direction heterogeneity; it cannot support even a weak causal or comparative-efficacy inference, and pooling across these PICOs would be inappropriate.
 Routing domain `longevity_research` is publication-lane metadata only; the source scope here is defined by the selected related_macular receipts.

## Next gaps

A stronger memo needs one matched PICO, for example: population=fundus images across normal, intermediate AMD, geographic atrophy, and wet AMD categories; intervention/exposure=CAD framework with weighted majority voting over best classifiers; comparator=baseline performance prior to weighted majority voting; outcome=balanced accuracy.
If related_macular is promoted beyond a scoping note, the next run should select sources sharing one context family rather than mixing human clinical/observational and other source context.

{
  "article_type": "alpha_memo",
  "domain_slug": "longevity_research",
  "researka_object_type": "submission",
  "researka_submission_id": "8a01740f-51fc-4253-bcdf-86c4eb7c5b14",
  "title": "related_macular: one bounded, context-dependent signal across receipts"
}