1
|
Tonneau M, Phan K, Manem VSK, Low-Kam C, Dutil F, Kazandjian S, Vanderweyen D, Panasci J, Malo J, Coulombe F, Gagné A, Elkrief A, Belkaïd W, Di Jorio L, Orain M, Bouchard N, Muanza T, Rybicki FJ, Kafi K, Huntsman D, Joubert P, Chandelier F, Routy B. Generalization optimizing machine learning to improve CT scan radiomics and assess immune checkpoint inhibitors' response in non-small cell lung cancer: a multicenter cohort study. Front Oncol 2023; 13:1196414. [PMID: 37546399 PMCID: PMC10400292 DOI: 10.3389/fonc.2023.1196414] [Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Grants] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/29/2023] [Accepted: 06/28/2023] [Indexed: 08/08/2023] Open
Abstract
Background Recent developments in artificial intelligence suggest that radiomics may represent a promising non-invasive biomarker to predict response to immune checkpoint inhibitors (ICIs). Nevertheless, validation of radiomics algorithms in independent cohorts remains a challenge due to variations in image acquisition and reconstruction. Using radiomics, we investigated the importance of scan normalization as part of a broader machine learning framework to enable model external generalizability to predict ICI response in non-small cell lung cancer (NSCLC) patients across different centers. Methods Radiomics features were extracted and compared from 642 advanced NSCLC patients on pre-ICI scans using established open-source PyRadiomics and a proprietary DeepRadiomics deep learning technology. The population was separated into two groups: a discovery cohort of 512 NSCLC patients from three academic centers and a validation cohort that included 130 NSCLC patients from a fourth center. We harmonized images to account for variations in reconstruction kernel, slice thicknesses, and device manufacturers. Multivariable models, evaluated using cross-validation, were used to estimate the predictive value of clinical variables, PD-L1 expression, and PyRadiomics or DeepRadiomics for progression-free survival at 6 months (PFS-6). Results The best prognostic factor for PFS-6, excluding radiomics features, was obtained with the combination of Clinical + PD-L1 expression (AUC = 0.66 in the discovery and 0.62 in the validation cohort). Without image harmonization, combining Clinical + PyRadiomics or DeepRadiomics delivered an AUC = 0.69 and 0.69, respectively, in the discovery cohort, but dropped to 0.57 and 0.52, in the validation cohort. This lack of generalizability was consistent with observations in principal component analysis clustered by CT scan parameters. Subsequently, image harmonization eliminated these clusters. The combination of Clinical + DeepRadiomics reached an AUC = 0.67 and 0.63 in the discovery and validation cohort, respectively. Conversely, the combination of Clinical + PyRadiomics failed generalizability validations, with AUC = 0.66 and 0.59. Conclusion We demonstrated that a risk prediction model combining Clinical + DeepRadiomics was generalizable following CT scan harmonization and machine learning generalization methods. These results had similar performances to routine oncology practice using Clinical + PD-L1. This study supports the strong potential of radiomics as a future non-invasive strategy to predict ICI response in advanced NSCLC.
Collapse
Affiliation(s)
- Marion Tonneau
- Department of Cancer Research, Centre de Recherche du Centre Hospitalier Universitaire de Montréal (CRCHUM), Montreal, QC, Canada
- Université de Médecine, Lille, France
| | - Kim Phan
- Imagia Canexia Health, Montreal, QC, Canada
| | - Venkata S. K. Manem
- Institut Universitaire de Cardiologie et de Pneumologie de Quebec, Université Laval, Québec City, QC, Canada
- Department of Mathematics and Computer Science, University of Quebec at Trois-Rivières, Trois-Rivières, QC, Canada
| | | | | | - Suzanne Kazandjian
- Department of Medical Oncology, Jewish General Hospital, Montreal, QC, Canada
| | - Davy Vanderweyen
- Department of Radiology, Centre Hospitalier de Sherbrooke (CHUS), Sherbrooke, QC, Canada
| | - Justin Panasci
- Department of Medical Oncology, Jewish General Hospital, Montreal, QC, Canada
| | - Julie Malo
- Department of Cancer Research, Centre de Recherche du Centre Hospitalier Universitaire de Montréal (CRCHUM), Montreal, QC, Canada
| | - François Coulombe
- Institut Universitaire de Cardiologie et de Pneumologie de Quebec, Université Laval, Québec City, QC, Canada
| | - Andréanne Gagné
- Institut Universitaire de Cardiologie et de Pneumologie de Quebec, Université Laval, Québec City, QC, Canada
| | - Arielle Elkrief
- Department of Cancer Research, Centre de Recherche du Centre Hospitalier Universitaire de Montréal (CRCHUM), Montreal, QC, Canada
- Hemato-Oncology Division, Centre Hospitalier de l’université de Montreal, Montreal, QC, Canada
| | - Wiam Belkaïd
- Department of Cancer Research, Centre de Recherche du Centre Hospitalier Universitaire de Montréal (CRCHUM), Montreal, QC, Canada
| | | | - Michele Orain
- Institut Universitaire de Cardiologie et de Pneumologie de Quebec, Université Laval, Québec City, QC, Canada
| | - Nicole Bouchard
- Department of Oncology, Centre Hospitalier de Sherbrooke (CHUS), Sherbrooke, QC, Canada
| | - Thierry Muanza
- Department of Medical Oncology, Jewish General Hospital, Montreal, QC, Canada
- Department of Radiation Oncology, Lady Davis Institute of the Jewish General Hospital, Montreal, QC, Canada
| | | | - Kam Kafi
- Imagia Canexia Health, Montreal, QC, Canada
| | | | - Philippe Joubert
- Institut Universitaire de Cardiologie et de Pneumologie de Quebec, Université Laval, Québec City, QC, Canada
- Department of Pathology, Institut Universitaire de Cardiologie et de Pneumologie de Québec, Québec, QC, Canada
| | | | - Bertrand Routy
- Department of Cancer Research, Centre de Recherche du Centre Hospitalier Universitaire de Montréal (CRCHUM), Montreal, QC, Canada
- Hemato-Oncology Division, Centre Hospitalier de l’université de Montreal, Montreal, QC, Canada
| |
Collapse
|
2
|
Tonneau M, Phan K, Kazandjian S, Elkrief A, Panasci J, Richard C, Nolin-Lapalme A, El Sayed R, Ding L, Nair T, Malo J, Chandelier F, Kafi K, O'Brien J, Di Jorio L, Muanza T, Routy B. 1357P A deep radiomics approach to assess PD-L1 expression and clinical outcomes in patients with advanced non-small cell lung cancer treated with immune checkpoint inhibitors: A multicentric study. Ann Oncol 2021. [DOI: 10.1016/j.annonc.2021.08.1957] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/20/2022] Open
|
4
|
Byrne MF, Chapados N, Soudan F, Oertel C, Linares Pérez M, Kelly R, Iqbal N, Chandelier F, Rex DK. Real-time differentiation of adenomatous and hyperplastic diminutive colorectal polyps during analysis of unaltered videos of standard colonoscopy using a deep learning model. Gut 2019; 68:94-100. [PMID: 29066576 PMCID: PMC6839831 DOI: 10.1136/gutjnl-2017-314547] [Citation(s) in RCA: 364] [Impact Index Per Article: 72.8] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 05/23/2017] [Revised: 09/20/2017] [Accepted: 10/11/2017] [Indexed: 12/12/2022]
Abstract
BACKGROUND In general, academic but not community endoscopists have demonstrated adequate endoscopic differentiation accuracy to make the 'resect and discard' paradigm for diminutive colorectal polyps workable. Computer analysis of video could potentially eliminate the obstacle of interobserver variability in endoscopic polyp interpretation and enable widespread acceptance of 'resect and discard'. STUDY DESIGN AND METHODS We developed an artificial intelligence (AI) model for real-time assessment of endoscopic video images of colorectal polyps. A deep convolutional neural network model was used. Only narrow band imaging video frames were used, split equally between relevant multiclasses. Unaltered videos from routine exams not specifically designed or adapted for AI classification were used to train and validate the model. The model was tested on a separate series of 125 videos of consecutively encountered diminutive polyps that were proven to be adenomas or hyperplastic polyps. RESULTS The AI model works with a confidence mechanism and did not generate sufficient confidence to predict the histology of 19 polyps in the test set, representing 15% of the polyps. For the remaining 106 diminutive polyps, the accuracy of the model was 94% (95% CI 86% to 97%), the sensitivity for identification of adenomas was 98% (95% CI 92% to 100%), specificity was 83% (95% CI 67% to 93%), negative predictive value 97% and positive predictive value 90%. CONCLUSIONS An AI model trained on endoscopic video can differentiate diminutive adenomas from hyperplastic polyps with high accuracy. Additional study of this programme in a live patient clinical trial setting to address resect and discard is planned.
Collapse
Affiliation(s)
- Michael F Byrne
- Division of Gastroenterology, Vancouver General Hospital, Vancouver, British Columbia, Canada
| | - Nicolas Chapados
- Department of Technology, Imagia, Montreal, Quebec, Canada,Department of Applied Mathematics, Ecole Polytechnique de Montreal, Montreal, Quebec, Canada
| | - Florian Soudan
- Department of Technology, Imagia, Montreal, Quebec, Canada
| | - Clemens Oertel
- Department of Technology, Imagia, Montreal, Quebec, Canada
| | | | - Raymond Kelly
- Department of Anaesthetics, Beaumont Hospital, Dublin, Ireland
| | - Nadeem Iqbal
- Department of Gastroenterology, Saint Luke’s Hospital, Kilkenny, Ireland
| | | | - Douglas K Rex
- Division of Gastroenterology and Hepatology, Indiana University Medical Center, Indianapolis, Indiana, USA
| |
Collapse
|
5
|
Rabbani M, Kanevsky J, Kafi K, Chandelier F, Giles FJ. Role of artificial intelligence in the care of patients with nonsmall cell lung cancer. Eur J Clin Invest 2018; 48. [PMID: 29405289 DOI: 10.1111/eci.12901] [Citation(s) in RCA: 42] [Impact Index Per Article: 7.0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 12/31/2017] [Accepted: 01/28/2018] [Indexed: 12/27/2022]
Abstract
BACKGROUND Lung cancer is the leading cause of cancer death worldwide. In up to 57% of patients, it is diagnosed at an advanced stage and the 5-year survival rate ranges between 10%-16%. There has been a significant amount of research using machine learning to generate tools using patient data to improve outcomes. METHODS This narrative review is based on research material obtained from PubMed up to Nov 2017. The search terms include "artificial intelligence," "machine learning," "lung cancer," "Nonsmall Cell Lung Cancer (NSCLC)," "diagnosis" and "treatment." RESULTS Recent studies support the use of computer-aided systems and the use of radiomic features to help diagnose lung cancer earlier. Other studies have looked at machine learning (ML) methods that offer prognostic tools to doctors and help them in choosing personalized treatment options for their patients based on molecular, genetics and histological features. Combining artificial intelligence approaches into health care may serve as a beneficial tool for patients with NSCLC, and this review outlines these benefits and current shortcomings throughout the continuum of care. CONCLUSION We present a review of the various applications of ML methods in NSCLC as it relates to improving diagnosis, treatment and outcomes.
Collapse
Affiliation(s)
- Mohamad Rabbani
- McGill University Health Centre, McGill University, Montreal, QC, Canada
| | - Jonathan Kanevsky
- McGill University Health Centre, McGill University, Montreal, QC, Canada
| | - Kamran Kafi
- McGill University Health Centre, McGill University, Montreal, QC, Canada
| | | | | |
Collapse
|
7
|
Raum K, Leguerney I, Chandelier F, Talmant M, Saïed A, Peyrin F, Laugier P. Site-matched assessment of structural and tissue properties of cortical bone using scanning acoustic microscopy and synchrotron radiation μCT. Phys Med Biol 2006; 51:733-46. [PMID: 16424592 DOI: 10.1088/0031-9155/51/3/017] [Citation(s) in RCA: 64] [Impact Index Per Article: 3.6] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/12/2022]
Abstract
200 MHz scanning acoustic microscopy (SAM) and synchrotron radiation muCT (SR-muCT) were used to assess microstructural parameters and tissue properties in site-matched regions of interest in cortical bone. Anterior and postero-lateral regions of ten cross sections from human cortical radius were explored. Structural parameters, including diameter and number of Haversian canals per cortical area (Ca.Dm, N.Ca/Ar) and porosity Po were assessed with both methods using a custom-developed image fusion and analysis software. Acoustic impedance Z and degree of mineralization of bone DMB were extracted separately for osteonal and interstitial tissues from the fused images. Structural parameter estimations obtained from radiographic and acoustic images were almost identical. DMB and impedance values were in the range between 0.77 and 1.28 g cm(-3) and 5.13 and 12.1 Mrayl, respectively. Interindividual and regional variations were observed, whereas the strongest difference was found between osteonal and interstitial tissues (Z: 7.2 +/- 1.1 Mrayl versus 9.3 +/- 1.0 Mrayl, DMB: 1.06 +/- 0.07 g cm(-3) versus 1.16 +/- 0.05 g cm(-3), paired t-test, p < 0.05). Weak, but significant correlations between DMB and Z were obtained for the osteonal (R(2) = 0.174, p < 10(-4)) and for the pooled (osteonal and interstitial) data. The regression of the pooled osteonal and interstitial tissue data follows a second-order polynomial (R(2) = 0.39, p < 10(-4)). Both modalities fulfil the requirement for a simultaneous evaluation of cortical bone microstructure and material properties at the tissue level. While SAM inspection is limited to the evaluation of carefully prepared sample surfaces, SR-muCT provides volumetric information on the tissue without substantial preparation requirements. However, SAM provides a quantitative estimate of elastic properties at the tissue level that cannot be captured by SR-muCT.
Collapse
Affiliation(s)
- K Raum
- Laboratoire d'Imagerie Paramétrique, CNRS/Université Paris 6, UMR 7623, 15, rue de l'Ecole de Médecine, 75006 Paris, France.
| | | | | | | | | | | | | |
Collapse
|
9
|
Raum K, Leguerney I, Chandelier F, Bossy E, Talmant M, Saïed A, Peyrin F, Laugier P. Bone microstructure and elastic tissue properties are reflected in QUS axial transmission measurements. Ultrasound Med Biol 2005; 31:1225-35. [PMID: 16176789 DOI: 10.1016/j.ultrasmedbio.2005.05.002] [Citation(s) in RCA: 41] [Impact Index Per Article: 2.2] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Track Full Text] [Subscribe] [Scholar Register] [Received: 12/17/2004] [Revised: 04/25/2005] [Accepted: 05/11/2005] [Indexed: 05/04/2023]
Abstract
Accurate clinical interpretation of the sound velocity derived from axial transmission devices requires a detailed understanding of the propagation phenomena involved and of the bone factors that have an impact on measurements. In the low megahertz range, ultrasonic propagation in cortical bone depends on anisotropic elastic tissue properties, porosity and the cortical geometry (e.g., thickness). We investigated 10 human radius samples from a previous biaxial transmission study using a 50-MHz scanning acoustic microscope (SAM) and synchrotron radiation microcomputed tomography. The relationships between low-frequency axial transmission sound speed at 1 and 2 MHz, structural properties (cortical width Ct.Wi, porosity, Haversian canal density and material properties (acoustic impedance, mineral density) on site-matched cross-sections were investigated. Significant linear multivariate regression models (1 MHz: R(2) = 0.84, p < 10(-4), root-mean-square error (RMSE) = 38 m/s, 2 MHz: R(2) = 0.65, p < 10(-4), RMSE = 48 m/s) were found for the combination of Ct.Wi with porosity and impedance. A new model was derived that accounts for the nonlinear dispersion relation with Ct.Wi and predicts axial transmission velocities measured at different ultrasonic frequencies (R(2) = 0.69, p < 10(-4), RMSE = 52 m/s).
Collapse
Affiliation(s)
- Kay Raum
- Laboratoire d'Imagerie Paramétrique, CNRS/Université Paris 6, Paris, France.
| | | | | | | | | | | | | | | |
Collapse
|