EUVIMEDEUVIMED entdecken
Systematische Übersicht

Artificial Intelligence for Nodule Detection in Thoracic Computed Tomography: A Systematic review and Structural Analysis Based on the MI-CLAIM Model

Frederik Duedahl

British Journal of Cancer Research · 18.03.2026 · 9 (1)

DOI: 10.31488/bjcr.214 ↗

Kurz gefasst

Systematische Übersicht in der Zeitschrift British Journal of Cancer Research, online erschienen am 18.03.2026. Verfasst von Frederik Duedahl. Die Arbeit nennt 35 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Lung cancer remains a leading cause of cancer-related mortality, predominantly due to late-stage diagnosis. Lowdose Computed Tomography (LDCT) screening can reduce mortality but implementation is often challenged by already limited radiology resources. Artificial intelligence (AI) has the potential to improve CT-based lung nodule detection, but systematic and unified reporting on data sources, development, and model performance remains limited. This review applies MI-CLAIM and QUADAS-2 to assess methodological quality and reporting, following PRISMA-P guidelines1.

Objectives

1) Identify AI applications for thoracic CT nodule detection and their diagnostic accuracy. 2) Evaluate clinical implications, validation setup and data foundation. 3) Assess adherence to the MI-CLAIM checklist.

Methods

Databases (Medline, IEEE Xplore, Web of Science, Embase, Cinahl, Scopus, Cochrane, Google Scholar) were systematically searched using an adapted search string for the respective databases, developed with a trained librarian. Two named reviewers1,2 independently screened, and a third³ resolved disagreements. Data extraction focused on PICO elements, diagnostic performance, and reporting and bias with MI-CLAIM and QUADAS-2 adherence. The search concluded on the 27th of December 2024 Results Of 11,875 records identified, 8,729 remained after removal of duplicates. After title and abstract screening, 93 full-text articles were assessed, and 23 met the inclusion criteria. Most included studies were published after 2019 and were based on retrospective hospital CT datasets. According to QUADAS-2, 11 studies had low risk of bias. Mean MI-CLAIM adherence was 65%, and 3 of 23 studies achieved high reporting completeness. Reported AI sensitivities ranged from 59.2% to 98.8%, and false-positive rates from 0.067 to 2.4 per scan. Validation was mainly retrospective, with only 3 studies using prospective data.

Conclusion

AI for thoracic CT nodule detection shows promising but heterogeneous diagnostic performance, mostly based on retrospective, single-centre data with incomplete reporting. Before routine clinical use, robust external validation studies with prospective design and transparent reporting are required.

Zitieren

APA

Duedahl, F. (2026). Artificial Intelligence for Nodule Detection in Thoracic Computed Tomography: A Systematic review and Structural Analysis Based on the MI-CLAIM Model. British Journal of Cancer Research, 9(1). https://doi.org/10.31488/bjcr.214

Vancouver

Duedahl F. Artificial Intelligence for Nodule Detection in Thoracic Computed Tomography: A Systematic review and Structural Analysis Based on the MI-CLAIM Model. British Journal of Cancer Research. 2026;9(1). doi:10.31488/bjcr.214

BibTeX
@article{Duedahl2026,
  author = {Duedahl, Frederik},
  title = {Artificial Intelligence for Nodule Detection in Thoracic Computed Tomography: A Systematic review and Structural Analysis Based on the MI-CLAIM Model},
  journal = {British Journal of Cancer Research},
  year = {2026},
  volume = {9},
  number = {1},
  doi = {10.31488/bjcr.214}
}
Kostenlos starten →