Research articlePrognostic modeling for patients with colorectal liver metastases incorporating FDG PET radiomic features
Introduction
Colorectal cancer is a common cancer worldwide, often burdened by liver metastases [1]. About 15% of patients have liver metastases at the time of diagnosis and an additional 15% developed liver metastases over time [2]; 5-year survival in patients with liver metastases was reported as low as 5% in untreated patients [2]. However, recent studies report a 5-year survival rate of about 40% following surgical resection of colorectal liver metastases [3]. Treatment options for colorectal liver metastases have expanded with new therapeutic modalities such as radiofrequency ablation, which imply a clinical need for improved prognostication to assist choice of therapy.
The emerging area of precision (or personalized) cancer medicine involves efforts towards the discovery and validation of biomarkers that move beyond diagnosis, to domains such as prognostication, disease progression tracking, and therapy response prediction and assessment. To this end, PET imaging provides valuable capabilities for non-invasive assessment and quantification of disease burden, and towards the development of effective imaging biomarkers of disease [4]. Overall, PET images present a wide array of information related to disease. However, in common clinical practice, only intensity-based standard-uptake-value (SUV) metrics are utilized, particularly SUVmax or SUVpeak. This is due to the simplicity in the computation of these metrics, not requiring accurate segmentation of the tumors. Specifically, SUVmax is computed as the maximum uptake in an area of interest, and SUVpeak is obtained by moving a 1-cm3 spherical region of interest over the area with increased tracer uptake (not necessarily conforming to the precise tumor outline) to maximize the enclosed average uptake [5,6].
Quantitative volumetric tumor parameters, though less straightforward to compute, provide a notable frontier towards improved assessment of disease. In fact, there is increasing evidence that volumetric measures, particularly metabolic tumor volume (MTV) or total lesion glycolysis (TLG) can outperform their SUV counterparts, in a range of human solid tumors such as head & neck cancer, lung cancer, breast cancer, colorectal cancer and lymphoma [[7], [8], [9], [10], [11], [12], [13], [14], [15], [16]]. Tumor volumetric parameters facilitate estimation of total tumor burden in a patient at the time of diagnosis or recurrence. Furthermore, segmentation of PET images enables generation of SUVmean, which is also sometimes reported in the literature.
In the present work, we have performed extensive comparisons, including univariate and multivariate analyses involving a range of quantitative measures of tumor uptake, to assess optimal methods for prediction of clinical outcome in patients with liver metastases from colorectal cancer. Our analyses includes the use of volumetric parameters, as well as other advanced radiomic features which quantify heterogeneity [[17], [18], [19], [20], [21]] as increasingly studied in the emerging field of radiomics. The ultimate aim is that enhanced predictive models would result in significant improvements in management of patients, including non-invasive selection of patients with poor prognosis who could benefit from earlier and more intensive treatment strategies. These high-risk patients could also be identified for participation in clinical trials in order to better power discovery of effective therapies.
Section snippets
Subjects
We analyzed data from 52 patients with colorectal intrahepatic-only metastases (29 males and 23 females; mean age 62.9 years [SD 9.8; range 32–82]). The patients had FDG PET/CT scans obtained before treatment, in years 2005 to 2010 (with patient outcome follow-ups up to 2017). The scans were performed as part of the clinical workup prior to final decision on treatment, most often in patients considered for liver surgery, as PET/CT was not part of primary standard workup for all patients with
Results
An example of segmentation for a subject with liver metastasis is depicted in Fig. 1. When using SUV metrics, the four segmentation methods performed relatively similarly, but when performing volumetric analysis, 40% and 50% background-corrected SUVmax thresholding resulted in relatively improved performance especially in PFS (elaborated in the discussion section). Rest of the paper describes results for 40% background-corrected SUVmax thresholding.
Of the original 51 metrics, 26 were retained
Conventional measures vs. volumetric and heterogeneity parameters
In our univariate survival analyses of OS, PFS and EFS, SUV measures (max/mean/peak) did not perform as well as volumetric measures MTV or TLG (Fig. 2, Fig. 3, Fig. 4). Furthermore, in multivariate analyses, only in the case of PFS, SUV added value in combination with number of liver mets.
In a study by de Geus-Oei et al. [28] of 152 colorectal metastatic patients (majority with involvement of the liver), only SUVmean was evaluated for OS. The resulting HR, though statistically significant, was
Conclusion
The present work shows that conventional, commonly-employed SUV metrics (SUVmax, SUVpeak, SUVmean) perform relatively poorly in outcome prediction tasks (OS, PFS, EFS) when assessing colorectal liver metastases from FDG PET images. By contrast, use of the number of liver metastasis provided significant performance. This was also the case for volumetric MTV and TLG measures. Furthermore, use of multivariate prognostic modeling while including radiomic features further improved outcome
Author Contributions
AR: Led the overall project in terms of data analysis and writing the manuscript.
KPBF: PET image analyses, collecting non-PET data, interpretation of results, contributed to the final manuscript.
SA: Developed standardized metrics (radiomics) and pipeline used in data analysis.
LL: Contributed to statistical modeling and analysis of data.
CRS: Discussion of data and interpretation of results, and contributed to the final manuscript.
RMS: Clinical motivation for the work, discussion of data and
Financial disclosure
Dr. Keiding: Danish Council for Independent Research (Medical Sciences, 4004-00022).
Dr. Schmidtlein: MSK Cancer Center Support Grant/Core Grant (P30 CA008748).
Dr. Lu: National Natural Science Foundation of China. No disclosures exist for any other authors.
IRB Statement
Formal approval for access and analysis of patient data was obtained from the Danish Patient Safety Authority and the study was also approved by the Danish Data Protection Agency.
Acknowledgements
This work was supported by the Danish Council for Independent Research (Medical Sciences, 4004-00022), the MSK Cancer Center Support Grant/Core Grant (P30 CA008748), and the National Natural Science Foundation of China under grants 61628105. We also acknowledge very helpful discussions with Dr. Ciprian Crainiceanu.
References (45)
- et al.
Managing synchronous liver metastases from colorectal cancer: a multidisciplinary international consensus
Cancer Treat. Rev.
(2015) - et al.
Prognostic value of metabolic tumor volume measured by 18F-FDG PET/CT in advanced-stage squamous cell carcinoma of the larynx and hypopharynx
Ann. Oncol.
(2013) - et al.
Radiomics: the process and the challenges
Magn. Reson. Imaging
(2012) - et al.
Quantifying heterogeneity in human tumours using MRI and PET
Eur. J. Cancer
(2012) - et al.
Radiomics: Extracting more information from medical images using advanced feature analysis
Eur. J. Cancer
(2012) - et al.
Exploring feature-based approaches in PET images for predicting cancer treatment outcomes
Pattern Recognit.
(2009) - et al.
FDG-PET for prediction of survival of patients with metastatic colorectal carcinoma
Ann. Oncol.
(2006) - et al.
Metabolic tumor volume and total lesion glycolysis on FDG-PET/CT can predict overall survival after 90Y radioembolization of colorectal liver metastases: A comparison with SUVmax, SUVpeak, and RECIST 1.0
Eur. J. Radiol.
(2016) - et al.
Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012
Int. J. Cancer
(2015) - et al.
Epidemiology and management of liver metastases from colorectal cancer
Ann. Surg.
(2006)
Principles and Practice of PET and PET/CT
From RECIST to PERCIST: evolving considerations for PET response criteria in solid tumors
J. Nucl. Med.
Noise considerations for PET quantification using maximum and peak standardized uptake value
J. Nucl. Med.
18F-FDG metabolic tumor volume and total glycolytic activity of oral cavity and oropharyngeal squamous cell cancer: adding value to clinical staging
J. Nucl. Med.
PET-based primary tumor volumetric parameters and survival of patients with non—small cell lung carcinoma
Am J Roentgenol
Prognostic value of metabolic tumor volume and total lesion glycolysis in head and neck cancer: a systematic review and meta-analysis
J. Nucl. Med.
Prognostic significance of preoperative metabolic tumour volume and total lesion glycolysis measured by (18)F-FDG PET/CT in squamous cell carcinoma of the oral cavity
Eur. J. Nucl. Med. Mol. Imaging
Prognostic value of volume-based (18)F-fluorodeoxyglucose PET/CT parameters in patients with clinically node-negative oral tongue squamous cell carcinoma
Korean J. Radiol.
FDG avidity and tumor burden: survival outcomes for patients with recurrent breast cancer
Am. J. Roentgenol.
JOURNAL CLUB: Value of quantitative FDG PET/CT volumetric biomarkers in recurrent colorectal Cancer patient survival
Am. J. Roentgenol.
Pretherapeutic FDG-PET total metabolic tumor volume predicts response to induction therapy in pediatric Hodgkin’s lymphoma
BMC Cancer
Prognostic role of pretreatment 18F-FDG PET/CT in primary brain lymphoma
Ann. Nucl. Med.
Cited by (43)
Liver Transplantation for Colorectal Liver Metastases
2024, Surgical Clinics of North AmericaA different perspective on<sup>18</sup>F-FDG PET radiomics in colorectal cancer patients: The relationship between intra & peritumoral analysis and pathological findings
2023, Revista Espanola de Medicina Nuclear e Imagen MolecularThe management of colorectal liver metastases amenable of surgical resection: How to shape treatment strategies according to clinical, radiological, pathological and molecular features
2022, Cancer Treatment ReviewsCitation Excerpt :Moreover, in another study by Rahmim et al., the radiomic analysis of 52 pre-surgical FDG PET scans showed that the combination of metabolic and volumetric criteria can predict survival better than commonly-used SUV metrics [137]. However, both chemo-naïve and pre-treated patients were included in these analyses, thus hampering the interpretation of the identified radiomic features [136,137]. With regard to MRI, Shur et al. built a radiomic prognostic index that incorporated two significant MRI radiomic signatures [minimum pixel (MIN) and gray level size zone (glszm) small area emphasis (SAE)], to predict disease recurrence in 102 patients undergoing upfront CRLM resection [138].
Rectal MRI radiomics for predicting pathological complete response: Where we are
2022, Clinical ImagingCitation Excerpt :Radiomics approach also shows potential use in discriminating responding from non-responding liver metastases based on the pre-treatment CT scan, in patients with esophagogastric cancer.21 A PET-derived research demonstrated that SUV parameters hold significant potential in predicting clinical outcome in patients with colorectal liver metastases, especially when utilizing multivariate models.22 According to a recent published review,23 high discriminatory power and accuracy of individual radiomic features and radiomics models were reported for the prediction of response to treatment in patients with different gastrointestinal tumors, in particular for rectal and esophageal cancer.
Colorectal Cancer Liver Metastasis—State-of-the-Art and Future Perspectives
2023, Gastrointestinal DisordersPET/CT in assessment of colorectal liver metastases: a comprehensive review with emphasis on <sup>18</sup>F-FDG
2023, Clinical and Experimental Metastasis