TY - JOUR
T1 - Prognostic Value of Deep Learning-Mediated Treatment Monitoring in Lung Cancer Patients Receiving Immunotherapy
AU - Trebeschi, Stefano
AU - Bodalal, Zuhir
AU - Boellaard, Thierry N.
AU - Tareco Bucho, Teresa M.
AU - Drago, Silvia G.
AU - Kurilova, Ieva
AU - Calin-Vainak, Adriana M.
AU - Delli Pizzi, Andrea
AU - Muller, Mirte
AU - Hummelink, Karlijn
AU - Hartemink, Koen J.
AU - Nguyen-Kim, Thi Dan Linh
AU - Smit, Egbert F.
AU - Aerts, Hugo J. W. L.
AU - Beets-Tan, Regina G. H.
N1 - Funding Information:
This work was also carried out on the Dutch national e-infrastructure with the support of SURF Cooperative. The authors would also like to thank NVIDIA for their kind donation via the Academic GPU Grant Program as well as the Maurits en Anna de Kock Stichting for its financial support. TN-K was funded by the Oncologic Imaging Fellowship Grant from the European Society of Radiology.
Publisher Copyright:
© Copyright © 2021 Trebeschi, Bodalal, Boellaard, Tareco Bucho, Drago, Kurilova, Calin-Vainak, Delli Pizzi, Muller, Hummelink, Hartemink, Nguyen-Kim, Smit, Aerts and Beets-Tan.
PY - 2021/3/2
Y1 - 2021/3/2
N2 - BackgroundCheckpoint inhibitors provided sustained clinical benefit to metastatic lung cancer patients. Nonetheless, prognostic markers in metastatic settings are still under research. Imaging offers distinctive advantages, providing whole-body information non-invasively, while routinely available in most clinics. We hypothesized that more prognostic information can be extracted by employing artificial intelligence (AI) for treatment monitoring, superior to 2D tumor growth criteria.MethodsA cohort of 152 stage-IV non-small-cell lung cancer patients (NSCLC) (73 discovery, 79 test, 903CTs), who received nivolumab were retrospectively collected. We trained a neural network to identify morphological changes on chest CT acquired during patients' follow-ups. A classifier was employed to link imaging features learned by the network with overall survival.ResultsOur results showed significant performance in the independent test set to predict 1-year overall survival from the date of image acquisition, with an average area under the curve (AUC) of 0.69 (p < 0.01), up to AUC 0.75 (p < 0.01) in the first 3 to 5 months of treatment, and 0.67 AUC (p = 0.01) for durable clinical benefit (6 months progression-free survival). We found the AI-derived survival score to be independent of clinical, radiological, PDL1, and histopathological factors. Visual analysis of AI-generated prognostic heatmaps revealed relative prognostic importance of morphological nodal changes in the mediastinum, supraclavicular, and hilar regions, lung and bone metastases, as well as pleural effusions, atelectasis, and consolidations.ConclusionsOur results demonstrate that deep learning can quantify tumor- and non-tumor-related morphological changes important for prognostication on serial imaging. Further investigation should focus on the implementation of this technique beyond thoracic imaging.
AB - BackgroundCheckpoint inhibitors provided sustained clinical benefit to metastatic lung cancer patients. Nonetheless, prognostic markers in metastatic settings are still under research. Imaging offers distinctive advantages, providing whole-body information non-invasively, while routinely available in most clinics. We hypothesized that more prognostic information can be extracted by employing artificial intelligence (AI) for treatment monitoring, superior to 2D tumor growth criteria.MethodsA cohort of 152 stage-IV non-small-cell lung cancer patients (NSCLC) (73 discovery, 79 test, 903CTs), who received nivolumab were retrospectively collected. We trained a neural network to identify morphological changes on chest CT acquired during patients' follow-ups. A classifier was employed to link imaging features learned by the network with overall survival.ResultsOur results showed significant performance in the independent test set to predict 1-year overall survival from the date of image acquisition, with an average area under the curve (AUC) of 0.69 (p < 0.01), up to AUC 0.75 (p < 0.01) in the first 3 to 5 months of treatment, and 0.67 AUC (p = 0.01) for durable clinical benefit (6 months progression-free survival). We found the AI-derived survival score to be independent of clinical, radiological, PDL1, and histopathological factors. Visual analysis of AI-generated prognostic heatmaps revealed relative prognostic importance of morphological nodal changes in the mediastinum, supraclavicular, and hilar regions, lung and bone metastases, as well as pleural effusions, atelectasis, and consolidations.ConclusionsOur results demonstrate that deep learning can quantify tumor- and non-tumor-related morphological changes important for prognostication on serial imaging. Further investigation should focus on the implementation of this technique beyond thoracic imaging.
KW - artificial intelligence
KW - checkpoint inhibitors
KW - immunotherapy
KW - non small cell lung cancer
KW - treatment monitoring
KW - LYMPH-NODE METASTASIS
KW - PREOPERATIVE PREDICTION
KW - MODEL
U2 - 10.3389/fonc.2021.609054
DO - 10.3389/fonc.2021.609054
M3 - Article
C2 - 33738253
SN - 2234-943X
VL - 11
JO - Frontiers in Oncology
JF - Frontiers in Oncology
M1 - 609054
ER -