Research articleImpact of experimental design on PET radiomics in predicting somatic mutation status
Introduction
Functional positron emission tomography imaging with the glucose analog 18F-fluorodoxyglucose (18F-FDG PET) is widely used for diagnosis and staging in oncology. 18F FDG PET plays an increasingly important role in the assessment of treatment response [1] and planning of radiotherapy [2]. Furthermore, 18F FDG PET imaging can capture the different metabolic phenotypes that exist between tumors [3], [4]. Quantification of these phenotypes may improve tumor characterization for individualized therapy [4], [5]. Radiomics extracts an atlas of features that describe the tumor phenotype from standard of care images through the utilization of advanced mathematical algorithms that quantify the relationship between image voxels [6], [7]. Recently, PET radiomic features have demonstrated tremendous promise in predicting clinical outcomes [8], [9], treatment response [10], [11], and genomic data [3] in various malignancies.
As part of the feature extraction pipeline, PET images undergo two image processing methods [5]. First, the voxel sizes are resampled. Thus far, radiomic studies have relied on retrospective PET imaging data, which are often acquired on different scanners and are reconstructed differently, leading to variability in the voxel size [12], [13]. Therefore, for radiomic analysis, many research groups resample the voxel size with an interpolation filter in order to obtain a uniform size across all patients [5], [14]. Second, the image intensity of PET images are often discretized (or binned) into a limited range of intensity values to reduce image noise [5] and increase feature computational efficiency [6].
The values and methods chosen for these image processing steps are often selected without clear justification or are simply not mentioned in many radiomic studies. Therefore, it is unclear whether these parameters impact the performance of PET radiomic features in predicting outcomes or status and what parameters will maximize the predictability of the features. Researchers have reported that the quantitative value of PET radiomic features is sensitive to the choice of discretization methods [15], [16]. However, the effects of the resampled voxel size and interpolation filter on PET feature quantification have not been investigated. Furthermore, the impact of these settings on the ability of radiomic features to predict clinically valuable information has not been reported. With the overall goal of the clinical implementation of radiomics, it is important to study the impact of these parameters on the predictive performance of radiomic features.
Due to the high incidence and recurrence rate of non-small cell lung cancer (NSCLC) [17], there has been considerable interest in applying radiomics to characterize the disease in order to improve prognosis and better tailor treatments for individual patients [4]. Initiation and progression of NSCLC tumors are driven by somatic mutations in key oncogenes (e.g. epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma viral oncogene homolog (KRAS)) [18]. Early prediction of the oncogene mutation status may provide information for individualized therapy for NSCLC patients [18]. PET-based radiomic features have shown to be significantly associated with NSCLC mutation status [3]. Furthermore, unlike prediction of treatment outcomes (e.g. overall survival), the prediction of somatic mutations is not confounded by treatment methods as they are an intrinsic characteristic of the tumor. In this study, we investigated how various experimental settings influence the feature quantification and performance of radiomic features in predicting somatic mutations.
Section snippets
Patient imaging
This retrospective study was conducted under a Dana-Farber Cancer Institute and Brigham and Women’s Hospital Institutional Review Board approved protocol. This study included 348 patients with NSCLC who received diagnostic 18F FDG PET/CT scans prior to treatment between September 2003 and December 2013.
Patients were injected with 348–921 MBq of 18F FDG and PET images were acquired approximately 65 min after injection on a GE Discovery (GE Healthcare, Waukesha, WI), Siemens Biograph (Siemens AG,
Results
This study investigated how various experimental settings influence the quantification of radiomic features as well as the performance of these features in predicting somatic mutations in a large cohort of 348 NSCLC patients. The cohort was predominantly female (61%) and Caucasian (91%). 42% (146/348) and 58% (202/348) of the patients had early and advanced stages of NSCLC, respectively. 13% (44/348) of the patients harbored EGFR mutations.
Discussion
PET radiomic features have great potential to predict mutation status, treatment response, and prognosis by quantifying the tumor metabolic phenotype. Before feature extraction, PET images need to be resampled and interpolated to uniform voxel grids. Furthermore, the intensity histogram of PET images need to be discretized. However, the impact of these image processing steps on the predictability of radiomic features has not been carefully investigated. Here, we report how different settings
Conclusions
Radiomic features with excellent predictive performance for EGFR mutation status were robust to the parameters chosen for image processing steps, and therefore, these parameters may potentially be arbitrarily chosen for mutation prediction in NSCLC patients. However, the parameters need to be carefully chosen for other radiomic features that may be less predictive. Regardless of the experimental settings, some radiomic features were still unable to significantly predict EGFR mutations, and
Ethical consideration
This retrospective study was approved by a Dana-Farber Cancer Institute and Brigham and Women’s Hospital Institutional Review Board.
Conflict of interest
The authors of this manuscript declared no conflict of interest.
Acknowledgements
This study was funded by the National Institute of Health Award NumberU01CA190234 and U24CA194354, and the Research Seed Funding Grant from the American Association of Physicists in Medicine. The authors would also like to thank the PROFILE team for their help with somatic mutation testing.
References (30)
- et al.
Modeling pathologic response of esophageal cancer to chemoradiation therapy using spatial-temporal 18F-FDG PET features, clinical parameters, and demographics
Int. J. Radiat. Oncol. Biol. Phys.
(2014) - et al.
Predicting malignant nodules from screening CTs
J. Thorac. Oncol.
(2016) - et al.
Radiomics: the process and the challenges
Magn. Reson. Imaging
(2012) - et al.
CT-based radiomic analysis of stereotactic body radiation therapy patients with lung cancer
Radiother. Oncol.
(2016) - et al.
Patterns of locoregional failure in stage III non-small cell lung cancer treated with definitive chemoradiation therapy
Pract. Radiat. Oncol.
(2014) - et al.
New driver mutations in non-small-cell lung cancer
Lancet Oncol.
(2011) Texture analysis using gray level run lengths
Comput. Graph. Image Processing
(1975)- et al.
Variability of image features computed from conventional and respiratory-gated PET/CT images of lung cancer
Transl. Oncol.
(2015) - et al.
From RECIST to PERCIST: evolving considerations for PET response criteria in solid tumors
J. Nucl. Med.
(2009) - et al.
18F-FLT PET/CT for early response monitoring and dose escalation in oropharyngeal tumors
J. Nucl. Med.
(2010)
Associations between somatic mutations and metabolic imaging phenotypes in non-small cell lung cancer
J. Nucl. Med.
Radiomics in PET: principles and applications
Clin. Transl. Imaging
Characterization of PET/CT images using texture analysis: the past, the present… any future?
Eur. J. Nucl. Med. Mol. Imaging
Applications and limitations of radiomics
Phys. Med. Biol.
The potential of radiomic-based phenotyping in precision medicine: a review
JAMA Oncol.
Cited by (38)
Artificial intelligence in cancer diagnosis and therapy: Current status and future perspective
2023, Computers in Biology and MedicineClinical application of AI-based PET images in oncological patients
2023, Seminars in Cancer BiologyPreoperative prediction of mediastinal lymph node metastasis in non-small cell lung cancer based on <sup>18</sup>F-FDG PET/CT radiomics
2023, Clinical RadiologyCitation Excerpt :Kang et al.20 reported that PET/CT radiomics signatures could effectively decrease the false-positive rate of manual diagnosis of lung cancer (manual diagnosis 30.6%; PET/CT radiomics 5.4%). Numerous past studies revealed that 18F-FDG PET/CT-based radiomic prediction models could identify molecular subtypes or disease-free survival of lung cancer.21–25 Additionally, PET/CT radiomics has the capacity to predict the prognosis of patients undergoing radiotherapy, with a sensitivity of 100% and a specificity of 96%.26
Artificial intelligence-based prediction of clinical outcome in immunotherapy and targeted therapy of lung cancer
2022, Seminars in Cancer Biology[<sup>18</sup>F]FDG-PET/CT Radiomics and Artificial Intelligence in Lung Cancer: Technical Aspects and Potential Clinical Applications
2022, Seminars in Nuclear MedicineCitation Excerpt :Tumors with EGFR mutation tend to have more heterogeneity,168, 169 which might be captured by radiomics. Employing [18F]FDG-PET/CT conventional radiomics and ML for the prediction of EGFR mutation, a wide range of predictive power has been reported (AUC = 0.5-0.87), mostly around 0.75-0.80.168, 170-174 [ 18F]FDG-PET/CT-based features seemed superior over CT and PET-only radiomics (AUC = 0.80-0.84, 0.67-0.72, and 0.74-0.79, respectively).168, 175
CT-based radiomics for predicting brain metastases as the first failure in patients with curatively resected locally advanced non-small cell lung cancer
2021, European Journal of RadiologyCitation Excerpt :As shown in the results, the radiomics signature is composed of the following features: (original.shape.Maximum2DDiameterColumn), (original.glcm.Correlation), (wavelet.LHL.glszm.LargeAreaLowGrayLevelEmphasis), (wavelet.LLH.ngtdm.Strength), (wavelet.HLH.glcm.MCC), (wavelet.HLH.glcm.Imc1), and (wavelet.HLH.firstorder.Median). Some radiomics features in this study have been reported to be related to EGFR mutations in previous studies, such as SHAPE, GLCM, and GLSZM [38]. EGFR mutations play an essential role in the BM of NSCLC and may support risk stratification, especially in East Asian patients.