Deep learning for the detection of benign and malignant pulmonary nodules in non-screening chest CT scans

Sep 24, 2026

Uploaded by Elsa

Outside lung cancer screening, early-stage disease is often diagnosed only after incidental pulmonary nodules are found on clinically ordered chest CT. This work evaluates a deep learning AI system’s clinical validation in a non-screening setting, focusing on actionable benign nodules, small primary lung cancers, and pulmonary metastases. The study uses internal and external validation data from two Dutch hospitals, with radiologist labeling and sensitivity measured at predefined false-positive rates, benchmarked against a radiology panel.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Background

Method

Results

Conclusions

Plain language summary

ARTICLE [https://doi.org/10.1038/s43856-023-00388-5](https://doi.org/10.1038/s43856-023-00388-5) OPEN Deep learning for the detection of benign and malignant pulmonary nodules in non-screening chest CT scans Ward Hendrix 1,2, Nils Hendrix1,2,3, Ernst T. Scholten1, Mariëlle Mourits 4, Joline Trap-de Jong5, Steven Schalekamp1, Mike Korst2, Maarten van Leuken4, Bram van Ginneken1, Mathias Prokop1,6, Matthieu Rutten 1,2 & Colin Jacobs 1 ✉ Abstract Background Outside a screening program, early-stage lung cancer is generally diagnosed after the detection of incidental nodules in clinically ordered chest CT scans. Despite the advances in artiﬁcial intelligence (AI) systems for lung cancer detection, clinical validation of these systems is lacking in a non-screening setting. Method We developed a deep learning-based AI system and assessed its performance for the detection of actionable benign nodules (requiring follow-up), small lung cancers, and pulmonary metastases in CT scans acquired in two Dutch hospitals (internal and external validation) . A panel of ﬁve thoracic radiologists labeled all nodules, and two additional radiologists veriﬁed the nodule malignancy status and searched for any missed cancers using data from the national Netherlands Cancer Registry. The detection performance was evaluated by measuring the sensitivity at predeﬁned false positive rates on a free receiver operating characteristic curve and was compared with the panel of radiologists. Results On the external test set (100 scans from 100 patients), the sensitivity of the AI system for detecting benign nodules, primary lung cancers, and metastases is respectively 94.3%(82/87, 95% CI: 88.1–98.8%), 96.9%(31/32, 95% CI: 91.7–100%), and 92.0%(104/ 113, 95% CI: 88.5–95.5%) at a clinically acceptable operating point of 1 false positive per scan (FP/s) . These sensitivities are comparable to or higher than the radiologists, albeit with a slightly higher FP/s (average difference of 0 . 6) . Conclusions The AI system reliably detects benign and malignant pulmonary nodules in clinically indicated CT scans and can potentially assist radiologists in this setting. Plain language summary Early-stage lung cancer can be diagnosed after identifying an abnormal spot on a chest CT scan ordered for other medical reasons. These spots or lung nodules can be overlooked by radiologists, as they are not necessarily the focus of an examination and can be as small as a few millimeters. Software using Artiﬁcial Intelligence (AI) technology has proven to be successful for aiding radiologists in this task, but its performance is understudied outside a lung cancer screening setting. We therefore developed and validated AI software for the detection of cancerous nodules or non-cancerous nodules that would need attention. We show that the software can reliably detect these nodules in a non-screening setting and could potentially aid radiologists in daily clinical practice. 1 Diagnostic Imaging Analysis Group, Radiology and Nuclear Medicine Department, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 GA Nijmegen, The Netherlands. 2 Radiology Department, Jeroen Bosch Hospital, Henri Dunantstraat 1, 5223 GZ ‘s-Hertogenbosch, The Netherlands. 3 Jheronimus Academy of Data Science, Sint Janssingel 92, 5211 DA ‘s-Hertogenbosch, The Netherlands. 4 Radiology Department, Canisius Wilhelmina Hospital, Weg door Jonkerbos 100, 6532 SZ Nijmegen, The Netherlands. 5 Radiology Department, St. Antonius Hospital, Koekoekslaan 1, 3435 CM Nieuwegein, The Netherlands. 6 Radiology Department, University Medical Center Groningen, Hanzeplein 1, 9713 GZ Groningen, The Netherlands.✉ email: [colin.jacobs@radboudumc.nl](colin.jacobs@radboudumc.nl) COMMUNICATIONS MEDICINE | (2023)3:156| [https://doi.org/10.1038/s43856-023-00388-5|www.nature.com/commsmed](https://doi.org/10.1038/s43856-023-00388-5|www.nature.com/commsmed) 1 Lung cancer is one of the most frequent cancers

FAQs

What problem does the document address?

It addresses how early-stage lung cancer is often diagnosed in non-screening chest CT scans after incidental nodules are detected, even though clinical validation of AI systems in this setting remains limited.

How was the deep learning AI system evaluated?

The system was developed and assessed using CT scans from two Dutch hospitals with internal and external validation. A panel of five thoracic radiologists labeled nodules, and additional radiologists verified malignancy and searched for missed cancers using national cancer registry data.

What were the main detection results and how were they measured?

Detection performance was evaluated by sensitivity at predefined false-positive rates on a free receiver operating characteristic curve and compared with the radiologists. On the external test set, sensitivities for benign nodules, primary lung cancers, and metastases were reported at a clinically acceptable operating point of 1 false positive per scan.

Sep 24, 2026

Uploaded by Elsa