1
|
Ma B, Guo J, Chu H, van Dijk LV, van Ooijen PM, Langendijk JA, Both S, Sijtsema NM. Comparison of computed tomography image features extracted by radiomics, self-supervised learning and end-to-end deep learning for outcome prediction of oropharyngeal cancer. Phys Imaging Radiat Oncol 2023; 28:100502. [PMID: 38026084 PMCID: PMC10663809 DOI: 10.1016/j.phro.2023.100502] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/03/2023] [Revised: 10/02/2023] [Accepted: 10/17/2023] [Indexed: 12/01/2023] Open
Abstract
Background and purpose To compare the prediction performance of image features of computed tomography (CT) images extracted by radiomics, self-supervised learning and end-to-end deep learning for local control (LC), regional control (RC), locoregional control (LRC), distant metastasis-free survival (DMFS), tumor-specific survival (TSS), overall survival (OS) and disease-free survival (DFS) of oropharyngeal squamous cell carcinoma (OPSCC) patients after (chemo)radiotherapy. Methods and materials The OPC-Radiomics dataset was used for model development and independent internal testing and the UMCG-OPC set for external testing. Image features were extracted from the Gross Tumor Volume contours of the primary tumor (GTVt) regions in CT scans when using radiomics or a self-supervised learning-based method (autoencoder). Clinical and combined (radiomics, autoencoder or end-to-end) models were built using multivariable Cox proportional-hazard analysis with clinical features only and both clinical and image features for LC, RC, LRC, DMFS, TSS, OS and DFS prediction, respectively. Results In the internal test set, combined autoencoder models performed better than clinical models and combined radiomics models for LC, RC, LRC, DMFS, TSS and DFS prediction (largest improvements in C-index: 0.91 vs. 0.76 in RC and 0.74 vs. 0.60 in DMFS). In the external test set, combined radiomics models performed better than clinical and combined autoencoder models for all endpoints (largest improvements in LC, 0.82 vs. 0.71). Furthermore, combined models performed better in risk stratification than clinical models and showed good calibration for most endpoints. Conclusions Image features extracted using self-supervised learning showed best internal prediction performance while radiomics features have better external generalizability.
Collapse
Affiliation(s)
- Baoqiang Ma
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
| | - Jiapan Guo
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
- Machine Learning Lab, Data Science Center in Health (DASH), Groningen, Netherlands
- Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence , University of Groningen, Groningen, Netherlands
| | - Hung Chu
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
- Machine Learning Lab, Data Science Center in Health (DASH), Groningen, Netherlands
- Center for Information Technology, University of Groningen ,Groningen, Netherlands
| | - Lisanne V. van Dijk
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
- Department of Radiation Oncology, University of Texas MD Anderson Cancer Center, Houston, TX USA
| | - Peter M.A. van Ooijen
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
- Machine Learning Lab, Data Science Center in Health (DASH), Groningen, Netherlands
| | - Johannes A. Langendijk
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
| | - Stefan Both
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
| | - Nanna M. Sijtsema
- Department of Radiation Oncology, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
| |
Collapse
|
2
|
Connor S, Sit C, Anjari M, Lei M, Guerrero-Urbano T, Szyszko T, Cook G, Bassett P, Goh V. The ability of post-chemoradiotherapy DWI ADC mean and 18F-FDG SUV max to predict treatment outcomes in head and neck cancer: impact of human papilloma virus oropharyngeal cancer status. J Cancer Res Clin Oncol 2021; 147:2323-2336. [PMID: 34159420 PMCID: PMC8236463 DOI: 10.1007/s00432-021-03662-y] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/07/2021] [Accepted: 05/10/2021] [Indexed: 11/26/2022]
Abstract
OBJECTIVES To evaluate the ability of post-chemo-radiotherapy (CRT) diffusion-weighted-MRI apparent diffusion coefficient (ADCmean) and 18F-FDG PET maximum standardized uptake value (SUVmax) to predict disease-free survival (DFS) in head and neck squamous cell carcinoma (HNSCC), and to determine whether this ability is influenced by human papillomavirus oropharyngeal cancer (HPV-OPC) status. METHODS This prospective cohort observational study included 65 participants (53 male, mean ± SD age 59.9 ± 7.9 years, 46 HPV-OPC) with stage III or IV HNSCC. Primary tumour and nodal ADCmean (pre-treatment, 6- and 12-weeks post-CRT) and SUVmax (12-weeks post-CRT) were measured. Variables were compared with 2-year DFS (independent t-test/Mann-Whitney test) and overall DFS (Cox regression), before and after accounting for HPV-OPC status. Variables were also compared between HPV-OPC and other HNSCC subgroups after stratifying for DFS. RESULTS Absolute post-CRT ADCmean values predicted 2-year DFS and overall DFS for all participants (p = 0.03/0.03, 6-week node; p = 0.02/0.03 12-week primary tumour) but not in the HPV-OPC subgroup. In participants with DFS, percentage interval changes in primary tumour ADCmean at 6- and 12-weeks were higher in HPV-OPC than other HNSCC (p = 0.01, 6 weeks; p = 0.005, 12 weeks). The 12-week post-CRT SUVmax did not predict DFS. CONCLUSION Absolute post-CRT ADCmean values predicted DFS in HNSCC but not in the HPV-OPC subgroup. Amongst participants with DFS, post-CRT percentage interval changes in primary tumour ADCmean were significantly higher in HPV-OPC than in other HNSCC. Knowledge of HPV-OPC status is crucial to the clinical utilisation of post-CRT DWI-MRI for the prediction of outcomes.
Collapse
Affiliation(s)
- S Connor
- School of Biomedical Engineering and Imaging Sciences, St Thomas' Hospital, King's College, London, SE1 7EH, UK.
- Department of Neuroradiology, Ruskin Wing, Kings College Hospital, Denmark Hill, London, SE5 9RS, UK.
- Department of Radiology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK.
| | - C Sit
- Department of Radiology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| | - M Anjari
- Department of Radiology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| | - M Lei
- Department of Oncology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| | - T Guerrero-Urbano
- Department of Oncology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| | - T Szyszko
- King's College London & Guy's and St. Thomas' PET Centre, London, SE1 7EH, UK
| | - G Cook
- School of Biomedical Engineering and Imaging Sciences, St Thomas' Hospital, King's College, London, SE1 7EH, UK
- King's College London & Guy's and St. Thomas' PET Centre, London, SE1 7EH, UK
| | - P Bassett
- Department of Oncology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| | - V Goh
- School of Biomedical Engineering and Imaging Sciences, St Thomas' Hospital, King's College, London, SE1 7EH, UK
- Department of Radiology, Guy's Hospital, 2nd Floor, Tower Wing, Great Maze Pond, London, SE1 9RT, UK
| |
Collapse
|