Carolyn Horst

Carolyn Horst

Auto-generated profile, based on publicly available information

Carolyn Horst

Research Department of Cancer Imaging School of Biomedical Engineering and Imaging Sciences, King's College London London UK

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
23 April 2025 in Nature Medicine
Verified
Journal Article
Closed access
PDF file
MTMickael Tordjman
Mickael Tordjman
ZLZelong Liu
Zelong Liu
MYMurat Yuce
Murat Yuce
VFValentin Fauveau
Valentin Fauveau
YMYunhao Mei
Yunhao Mei
XMXueyan Mei
Xueyan Mei

13 August 2020 in Thorax
Verified
Journal Article
Closed access
PDF file
CHCarolyn Horst
Carolyn Horst
JDJennifer L Dickson
Jennifer L Dickson
STSophie Tisi
Sophie Tisi
MRMamta Ruparel
Mamta Ruparel
ANArjun Nair
Arjun Nair
SJSam M Janes
Sam M Janes
The SUMMIT study (ClinicalTrials.gov NCT03934866) is an LCS study recruiting individuals 55–77 years old at high risk of lung and other smoking-related cancers to LDCT screening. Its twin aims are to examine the performance of delivering an LDCT screening service for lung cancer to a high-risk population and to validate a cell-free nucleic acid blood test for detection of multiple cancers. The study began enrolment in April 2019 after the development of protocols for the management of pulmonary nodules and incidental findings that enabled a consistent approach to management across the entirety of the study (target recruitment of 25 000). The study aims to deliver a programme of LCS that is pragmatic, evidence-based, and practically deliverable by primary and secondary care, importantly avoiding overzealous investigation of all findings (and therefore potentially increasing harms). Examination of the evidence that medical intervention of incidental findings makes a difference to participants turns out to be sparse, making detailed radiological reporting probably unnecessary. The reader will see here that we provide only limited and highly specific information beyond the presence of lung cancer or pulmonary nodules. It is our hope that this balanced approach will be borne out in the data we collect, bolstering a safe, effective and efficient implementation of LCS. Studies on whether a future health service could manage a more holistic approach, aligning the reporting of incidental findings such as coronary artery calcification (CAC), early emphysema and other findings to a more personalised health intervention with intensive smoking advice, cardiovascular disease (CVD) prevention and the like, are urgently needed.

26 July 2022 in European Respiratory Journal
Verified
Journal Article
Open access
PDF file
STSophie Tisi
Sophie Tisi
JDJennifer L. Dickson
Jennifer L. Dickson
CHCarolyn Horst
Carolyn Horst
SQSamantha L. Quaife
Samantha L. Quaife
HHHelen Hall
Helen Hall
SJSam M. Janes
Sam M. Janes
Background: COPD is a major comorbidity in lung cancer screening (LCS) cohorts, with a high prevalence of undiagnosed COPD. Combining symptom assessment with spirometry in this setting may enable earlier diagnosis of clinically significant COPD and facilitate increased understanding of lung cancer risk in COPD. In this study, we wished to understand the prevalence, severity, clinical phenotype and lung cancer risk of individuals with symptomatic undiagnosed COPD in a LCS cohort.Methods: 16 010 current or former smokers aged 55–77 years attended a lung health check as part of the SUMMIT Study. A respiratory consultation and spirometry were performed alongside LCS eligibility assessment. Those with symptoms, no previous COPD diagnosis and airflow obstruction were labelled as undiagnosed COPD. Baseline low-dose computed tomography (LDCT) was performed in those at high risk of lung cancer (PLCOm2012score ≥1.3% and/or meeting USPSTF 2013 criteria).Results: Nearly one in five (19.7%) met criteria for undiagnosed COPD. Compared with those previously diagnosed, those undiagnosed were more likely to be male (59.1%versus53.2%; p<0.001), currently smoking (54.9%versus47.6%; p<0.001) and from an ethnic minority group (p<0.001). Undiagnosed COPD was associated with less forced expiratory volume in 1 s impairment (Global Initiative for Chronic Obstructive Lung Disease (GOLD) grades 1 and 2: 85.3%versus68.4%; p<0.001) and lower symptom/exacerbation burden (GOLD A and B groups: 95.6%versus77.9%; p<0.001) than those with known COPD. Multivariate analysis demonstrated that airflow obstruction was an independent risk factor for lung cancer risk on baseline LDCT (adjusted OR 2.74, 95% CI 1.73–4.34; p<0.001), with a high risk seen in those with undiagnosed COPD (adjusted OR 2.79, 95% CI 1.67–4.64; p<0.001).Conclusions: Targeted case-finding within LCS detects high rates of undiagnosed symptomatic COPD in those most at risk. Individuals with undiagnosed COPD are at high risk for lung cancer.

5 August 2020 in Thorax
Verified
Journal Article
Open access
PDF file
MRMamta Ruparel
Mamta Ruparel
SQSamantha L Quaife
Samantha L Quaife
JDJennifer L Dickson
Jennifer L Dickson
CHCarolyn Horst
Carolyn Horst
STSophie Tisi
Sophie Tisi
SJSam M Janes
Sam M Janes
The Lung Screen Uptake Trial tested a novel invitation strategy to improve uptake and reduce socioeconomic and smoking-related inequalities in lung cancer screening (LCS) participation. It provides one of the first UK-based ‘real-world’ LCS cohorts. Of 2012 invited, 1058 (52.6%) attended a ‘lung health check’. 768/996 (77.1%) in the present analysis underwent a low-dose CT scan. 92 (11.9%) and 33 (4.3%) participants had indeterminate pulmonary nodules requiring 3-month and 12-month surveillance, respectively; 36 lung cancers (4.7%) were diagnosed (median follow-up: 1044 days). 72.2% of lung cancers were stage I/II and 79.4% of non-small cell lung cancer had curative-intent treatment.

1 March 2026 in Radiology
Verified
Journal Article
Closed access
MTMickael TORDJMAN
Mickael TORDJMAN
MYMurat Yuce
Murat Yuce
AAAmine Ammar
Amine Ammar
MHMingqian Huang
Mingqian Huang
FBFadila Mihoubi Bouvier
Fadila Mihoubi Bouvier
BTBachir Taouli
Bachir Taouli
Synthetic radiographs, including OpenAI’s ChatGPT-generated ones, are not easily distinguishable from authentic radiographs by either radiologists or large language models; physician training and dedicated tools are crucial to mitigate risks.

1 November 2023 in Radiology: Artificial Intelligence
Verified
Journal Article
Closed access
MMMatthew S. Macpherson
Matthew S. Macpherson
CHCharles E. Hutchinson
Charles E. Hutchinson
CHCarolyn Horst
Carolyn Horst
VGVicky Goh
Vicky Goh
Giovanni MontanaGiovanni Montana
Giovanni Montana
To train an explainable deep learning model for patient reidentification in chest radiograph datasets and assess changes in model-perceived patient identity as a marker for emerging radiologic abnormalities in longitudinal image sets. This retrospective study used a set of 1 094 537 frontal chest radiographs and free-text reports from 259 152 patients obtained from six hospitals between 2006 and 2019, with validation on the public ChestX-ray14, CheXpert, and MIMIC-CXR datasets. A deep learning model was trained for patient reidentification and assessed on patient identity confirmation, retrieval of patient images from a database based on a query image, and radiologic abnormality prediction in longitudinal image sets. The representation learned was incorporated into a generative adversarial network, allowing visual explanations of the relevant features. Performance was evaluated with sensitivity, specificity, F1 score, Precision at 1, R-Precision, and area under the receiver operating characteristic curve (AUC) for normal and abnormal prediction. Patient reidentification was achieved with a mean F1 score of 0.996 ± 0.001 (2 SD) on the internal test set (26 152 patients) and F1 scores of 0.947–0.993 on the external test data. Database retrieval yielded a mean Precision at 1 score of 0.976 ± 0.005 at 299 × 299 resolution on the internal test set and Precision at 1 scores between 0.868 and 0.950 on the external datasets. Patient sex, age, weight, and other factors were identified as key model features. The model achieved an AUC of 0.73 ± 0.01 for abnormality prediction versus an AUC of 0.58 ± 0.01 for age prediction error. The image features used by a deep learning patient reidentification model for chest radiographs corresponded to intuitive human-interpretable characteristics, and changes in these identifying features over time may act as markers for an emerging abnormality. Keywords: Conventional Radiography, Thorax, Feature Detection, Supervised Learning, Convolutional Neural Network, Principal Component Analysis Supplemental material is available for this article. © RSNA, 2023 See also the commentary by Raghu and Lu in this issue.