Radiologists’ diagnostic confidence in surgically referred pulmonary nodules: a routine computed tomography report retrospective analysis

Read full text
Authors: Marina Higashi, Rintaro Ito, Shingo Iwano, Shinichiro Kamiya, Taketo Kato, Shota Nakamura, Tetsuya Mizuno, Toyofumi Fengshi Chen-Yoshikawa, Shinji Naganawa
Cited by: 1
References: 25

Abstract

BACKGROUND: Imaging plays a critical role in the preoperative evaluation of pulmonary nodules, especially when obtaining a definitive biopsy is not feasible. This study evaluated the reliability of routine computed tomography (CT) reports in differentiating benign from malignant nodules and to identify imaging features that influenced radiologists' diagnostic confidence. METHODS: We retrospectively analyzed 823 patients who underwent preoperative CT for pulmonary nodules at a single institution between 2017 and 2023. Radiologists' diagnostic confidence levels, recorded on a five-point scale (1 = benign, 5 = malignant), were extracted from reports and compared with pathological diagnoses. Diagnostic performance was assessed using the area under the curve (AUC) and risk of malignancy (ROM). RESULTS: A total of 823 pulmonary lesions were evaluated, including 701 malignant and 122 benign nodules. Radiologists' interpretations demonstrated excellent diagnostic performance, with an AUC of 0.961; 95%CI: 0.94-0.98, sensitivity of 0.95, and specificity of 0.86. Notably, 99.4% of nodules rated "5" were confirmed malignant. Diagnostic accuracy was significantly higher for ground-glass nodules compared with solid nodules (AUC 0.995 vs. 0.938, p < 0.001) and when prior CT scans were available (AUC 0.983 vs. 0.951, p = 0.036). The presence of coexisting chronic obstructive pulmonary disease or interstitial lung disease modestly reduced accuracy (AUC 0.973 vs. 0.919, p = 0.089). CONCLUSION: Radiologists' diagnostic confidence levels recorded in routine preoperative CT reports demonstrated high concordance with pathological outcomes in pulmonary nodules referred for diagnostic surgical resection after nondiagnostic biopsy, suggesting that clinical decision-making in this highly selected population is generally reliable.

Fields of Study & Subjects

Lung Cancer Diagnosis and Treatment Radiology practices and education Radiomics and Machine Learning in Medical Imaging

Publication details

Journal / source
BMC Pulmonary Medicine
Publisher
Springer Science and Business Media LLC
Volume
26
Issue
1
Publication date
18 May 2026
Language
EN