Virtual Library
Start Your Search
Jerome Declerck
Author of
-
+
OA06 - Refining Lung Cancer Screening (ID 131)
- Event: WCLC 2019
- Type: Oral Session
- Track: Screening and Early Detection
- Presentations: 1
- Now Available
- Moderators:Tomasz Grodzki, Lluis Esteban Tejero
- Coordinates: 9/09/2019, 11:00 - 12:30, Hilton Head (1978)
-
+
OA06.05 - Evaluation of a Deep Learning-Based Automatic Classifier for the Classification of Perifissural Nodules (Now Available) (ID 1928)
11:00 - 12:30 | Author(s): Jerome Declerck
- Abstract
- Presentation
Background
Perifissural nodules (PFNs) comprise approximately 20% of screening-detected nodules and are almost certainly benign. Automatic PFN classification could therefore reduce the number of follow-up procedures required for nodule work-up. Prior work has shown some success in AI classification with limited datasets. Here we evaluate the performance of a new deep convolutional neural network (CNN) for PFN classification, trained on a dataset of nodules retrospectively collected from multiple European centers, including validation on an independent reader-study dataset.
Method
Data (1103 Patients, 1557 unique nodules and 3320 nodule images) were collected from three centers in the UK and the Netherlands. Each nodule was categorized into subtypes, including “PFN”, by on-site radiologists. Labels were reviewed centrally, overseen by a single clinician to ensure consistency between sites.
A CNN classifier was trained to produce a score that classifies nodules as (typical) PFN or not, using five-fold cross validation. The PFN classifier was developed by “transfer learning” from an existing benign-vs-malignant AI trained on the US National Lung Screening Trial.
To compare the CNN with human performance, independent validation was performed on a separate dataset of 158 benign patients (196 nodules/nodule images) from two of the sites. Three readers (two radiologists and a radiology resident) were asked to label each nodule as typical PFN, atypical PFN, or non-PFN. To match the AI training procedure, only the typical-PFN labels were used in the reader study, and compared to atypical/non-PFN classified nodules.
Model performance was evaluated by area under the ROC curve (AUC). For the independent validation, Cohen’s kappa was used to measure both the model’s agreement with reader consensus (at least 2 in agreement) and inter-reader agreement. For Cohen’s kappa calculations the CNN score was binarized using a threshold determined from the internal validation data.
Result
The mean cross-validated AUC on the internal dataset was 92% (95% CI = 90.6–92.9). For the independent dataset, the classifier labelled 61/196 (31%) as typical PFNs, and reader consensus gave 45/196 (23%). Versus reader consensus, the AUC of the CNN on the reader-study dataset was 96% (95% CI 93.3–98.4). Both the classifier–reader agreement [(k=0.74) 90%] and the inter-reader agreement [(k=0.64–0.79) 88%-92%] were substantial.
Conclusion
The performance of the PFN classifier is similar to that of radiologists and is within the inter-reader variability of radiologists. This demonstrates the potential utility of CNN-based systems for automatic PFN classification.
Only Members that have purchased this event or have registered via an access code will be able to view this content. To view this presentation, please login, select "Add to Cart" and proceed to checkout. If you would like to become a member of IASLC, please click here.
-
+
P1.11 - Screening and Early Detection (ID 177)
- Event: WCLC 2019
- Type: Poster Viewing in the Exhibit Hall
- Track: Screening and Early Detection
- Presentations: 1
- Moderators:
- Coordinates: 9/08/2019, 09:45 - 18:00, Exhibit Hall
-
+
P1.11-02 - Acceleration of Lung Cancer Diagnosis: Utility Study for AI-Based Stratification of Pulmonary Nodules (ID 2752)
09:45 - 18:00 | Author(s): Jerome Declerck
- Abstract
Background
Lung cancer diagnostic pathway guidelines promote the use of risk stratification models. Artificial Intelligence (AI)-based risk models have been shown to achieve better diagnostic accuracy than clinical models like Mayo Clinic (Mayo) for particular clinical populations. The aim of this study is to examine whether this could translate into faster diagnosis for high-risk cancer patients.
Method
116 patients (116 nodules) have been collected from a retrospective consecutive cohort acquired at Vanderbilt University Hospital. Time to diagnosis (TTD) was defined as the number of days between the CT scan and diagnosis date. Mean TTD was calculated on the cohort on which TTD could be defined, and on a reduced group comprising of TTD >31 days only.
Risk scores for each nodule were found using the Mayo model and an AI-based Lung Cancer Prediction model (LCP) based on CT images alone. A 65% risk of cancer was taken to be the threshold at which surgical intervention is indicated (according to ACCP guidelines).
Result
Seven patients were dropped due to negative TTD, and six for having no definitive diagnosis date. The eventual cohort contained 61 cancer patients and 42 controls. Mean TTD is 140 days (Interquartile Range – IQR 1-77 days). 25 patients have TTD=0, 60 are within 31 days (28 cancers, 32 controls) and 43 (33 cancers, 10 controls) are above 32 days.
On the full cohort: Mayo risk score is >=65% for 15 cancers and 4 controls (sensitivity, 24.6%, specificity 90.5%), with a mean cancer TTD of 75 days. The LCP scores >=65% in 43 cancers and 10 controls (sensitivity, 70.5%, specificity 76.2%), mean cancer TTD 81 days.
On the reduced group: Mayo >=65% for 7 cancers and 2 controls (sensitivity, 21.2%, specificity 80.0%) with mean cancer TTD 150 days. The LCP scores >=65% in 21 cancers and 4 controls (sensitivity, 63.6%, specificity 60.0%), with mean cancer TTD 156 days.
Conclusion
The LCP risk model could potentially accelerate the diagnosis in 40% more cancer patients who were not worked up fully in the month following a scan (the jump in sensitivity going from Mayo to LCP risk >=65% is 42.4%). For these patients, time to a cancer diagnosis and treatment could be shortened by up to 156 days compared to recommendations if applying the Mayo risk model.
-
+
P2.11 - Screening and Early Detection (ID 178)
- Event: WCLC 2019
- Type: Poster Viewing in the Exhibit Hall
- Track: Screening and Early Detection
- Presentations: 2
- Now Available
- Moderators:
- Coordinates: 9/09/2019, 10:15 - 18:15, Exhibit Hall
-
+
P2.11-13 - What Is the Impact of Localised Data When Training Deep Neural Networks for Lung Cancer Prediction? (Now Available) (ID 2485)
10:15 - 18:15 | Author(s): Jerome Declerck
- Abstract
Background
Deep neural networks (DNN) have been shown to offer a viable alternative for risk cancer prediction of indeterminate pulmonary nodules (IPNs). While the type of data used for training is known to impact performance, this issue has not been extensively studied. We present, for the first time, a study of the effect of including training data that matches the clinical pathway of the independent validation dataset, a nodule clinic of incidental findings.
Method
Two identical DNNs were trained on the task of diagnosis prediction of pulmonary nodules from CT images. The first one (DNNnlst) used purely screening data from the US National Lung Screening Trial (922 cancer and 14733 benign nodules), while the second one (DNNnlst+incidental) included data of incidentally detected nodules from European hospitals (1064 cancer and 7207 benign nodules). Both models were evaluated in an independent validation set of nodules coming from a referral center in the UK (Royal Brompton and Harefield Hospital, London) consisting of baseline scans of 406 cancer and 325 benign nodules. The models were compared in terms of AUC, as well as their ability to reclassify cancer patients with intermediate risk nodules. The Intermediate risk sub-population was defined by selecting patients with nodules in the size range of 8 to 15mm, and who were followed-up within a year with CT, referred to PET-CT, or referred to biopsy. Within this sub-population, a cancer prevalence of 30% was assumed. The operating points of the cancer prediction models were chosen by setting a cancer risk of 70%, corresponding to high-risk nodules in the guidelines of the British Thoracic Society.
Result
The DNNnlst and DNNnlst+incidental models achieved an AUC of 84.33 (95%CI: 81.49, 87.15) and 87.43 (95%CI: 84.79, 89.82) respectively on the entire validation set, showing an improvement in the discrimination capabilities (p < 0.01). For reference, using the nodule’s maximum axial diameter as a predictor led to an AUC of 79.07 (95%CI: 75.73, 82.73). Additionally, considering only the intermediate risk population of the data, all of which would require workup according to guidelines, the DNNnlst+incidental model correctly classified as high risk 34.34% more cases than the DNNnlst model (sensitivity 59.39% (95%CI: 47.68, 75.17) vs. 44.21% (95%CI: 20.98, 64.42)), an improvement significant at p < 0.05.
Conclusion
Although a DNN trained only on the US lung cancer screening data could have clinical utility in an incidental setting, exposing it to further incidental data can not only increase its discriminability, as expected, but also make it a potentially more effective tool for speeding up the diagnosis of cancer patients with intermediate risk nodules and reducing unnecessary workups.
-
+
P2.11-46 - Could Malignancies Diagnosed During Round 2 of Screening Be Worked Up Sooner? (ID 2488)
10:15 - 18:15 | Author(s): Jerome Declerck
- Abstract
Abstract not provided
