AI and Cancer Research ── Lung ── 2026-10-06

Overview of the data and code supporting a systematic review and meta-analysis of AI that integrates histological imaging with omics data in non-small-cell lung cancer. Four parts: search and inclusion, with 35 of 5504 records included; the primary outcome, a within-study difference in AUC or C-index on the same patients, with 16 studies pooled; REML random-effects pooling with HKSJ intervals; and the released extraction data, risk-of-bias assessments and code. Abstract only was checked. — Image abstract — the whole article on one page (click to enlarge)

1 Lung papers. Lead: Extraction dataset, risk-of-bias assessments and analysis code for Incremental value of integrating histopathology with omics for prediction in non-small-cell lung cancer: a systematic review and meta-analysis

Journals covered and how papers are chosen: see the index

Lung cancer: treatment, trials and AI

Extraction dataset, risk-of-bias assessments and analysis code for Incremental value of integrating histopathology with omics for prediction in non-small-cell lung cancer: a systematic review and meta-analysis
Zenodo (CERN European Organization for Nuclear Research) · 2026-10-04 · Evidence: Systematic review and meta-analysis (supporting data and code)
Abstract
This record contains the data and code supporting the systematic review and meta-analysis "Incremental value of integrating histopathology with omics for prediction in non-small-cell lung cancer: a systematic review and meta-analysis" (submitted to Briefings in Bioinformatics). The review followed PRISMA 2020. PubMed, Scopus and IEEE Xplore were searched; of 5504 records, 35 studies using artificial intelligence to integrate histological imaging with genomic, transcriptomic, proteomic or epigenomic data in non-small-cell lung cancer (NSCLC) were included, and 16 contributed to meta-analyses. The primary outcome was the within-study difference in discrimination (AUC or C-index) between multimodal models and the strongest histology-only comparator evaluated on the same patients, pooled with restricted maximum-likelihood random-effects models and Hartung–Knapp–Sidik–Jonkman confidence intervals. Contents:1. Extraction dataset: 35 included studies × 32 fields (bibliographic data, cohorts, modalities, architectures, fusion strategy, validation design, endpoints and all reported performance estimates).2. Analysis dataset: every value entering a meta-analysis, with its source location (table, figure or page) in the original article and the derived standard errors.3. Risk-of-bias assessments: PROBAST with PROBAST-AI signalling questions for all 35 studies.4. Analysis code (Python 3.11, NumPy, SciPy): REML random-effects meta-analysis with HKSJ confidence intervals, prediction intervals, subgroup analyses, meta-regression, sensitivity and leave-one-out analyses, and Egger's test. The code reproduces the reference outputs of the R metafor package on the BCG benchmark dataset.5.
Results: (JSON) and figure-generation scripts. The included articles themselves are not redistributed; they remain subject to their original licences.
Journal IF: unknown (could not be matched)
Reference: [No authors listed]. Extraction dataset, risk-of-bias assessments and analysis code for Incremental value of integrating histopathology with omics for prediction in non-small-cell lung cancer: a systematic review and meta-analysis. Zenodo (CERN European Organization for Nuclear Research). [Epub ahead of print]. doi:10.5281/zenodo.23132027.
Checked: Abstract only

← Back to this issue