Prognostic value of tumor asphericity and 18F-FDG PET/CT texture features in colorectal cancer

Authors

Abstract

AimThe aim of this study was to investigate the prognostic value of tumor asphericity and radiomic features derived from 18F-FDG PET/CT in patients with colorectal cancer, and to evaluate their role in risk stratification for progression-free survival (PFS).
MethodsA retrospective analysis of 102 patients with diagnosed colorectal cancer who underwent pre-treatment 18F-FDG PET/CT imaging. Clinical variables, conventional PET parameters (SUVmax, MTV, TLG), Radiomic features, including tumor asphericity (ASP), GLCM (gray-level co-occurrence matrix)-entropy, GLSZM (gray-level size zone matrix) and GLRLM (gray-level run length matrix)-based parameters were extracted. PFS was the primary endpoint, and overall survival (OS) was the secondary endpoint. Univariate and multivariate Cox regression analyses, ROC curve analysis, and Kaplan–Meier curves were performed.
ResultsOver a median 26.5 months follow-up, 59 patients (57.8%) had disease progression and 50 (49%) died. Progression correlated with higher ASP (p=0.024), GLCM entropy (p<0.001), GLSZM-SZE (p=0.001), and lower GLSZM-LZE (p=0.030). Univariate Cox analysis showed that ASP, GLCM entropy, GLSZM-SZE, GLSZM-LZE, and TNM stage were associated with PFS and OS. Multivariate analysis found ASP (HR=5.6, p=0.021), GLCM entropy (HR=1.55, p=0.006) and TNM stage (all p<0.001) as independent factors. Higher than ROC cutoffs: 0.36 for ASP and 9.65 for GLCM entropy linked to shorter PFS (9.2 vs. 37.4 months, p<0.001) and (21.6 vs. 52.3 months, p=0.001), respectively.
ConclusionsTumor asphericity and GLCM-derived entropy obtained from pretreatment 18F-FDG PET/CT provide independent and complementary prognostic information beyond conventional metabolic parameters in colorectal cancer. These biomarkers may improve risk stratification and facilitate identification of patients at increased risk of disease progression and mortality.

Keywords

Introduction

Colorectal cancer (CRC) remains one of the leading causes of cancer-related morbidity and mortality worldwide, with survival largely determined by stage at diagnosis and the presence of metastatic disease.¹ However, patients within the same TNM stage frequently demonstrate markedly different clinical outcomes, indicating that anatomical staging alone is insufficient to fully characterize tumor behavior.² The biological heterogeneity of CRC plays a major role in treatment response, recurrence risk, and survival, underscoring the need for additional non-invasive biomarkers that reflect tumor aggressiveness.³
18F-FDG PET/CT provides functional insights beyond traditional morphological imaging and is commonly used for staging, restaging, and evaluating therapy in CRC.⁴ Common PET metrics, such as SUVmax, metabolic tumor volume (MTV), and total lesion glycolysis (TLG) have been studied as prognostic tools. However, their effectiveness varies across different studies, partly because of technical differences and the limited capacity of single-voxel metrics to reflect the complex biological structure of tumors.⁵
Radiomic features derived from GLCM and GLSZM reflect intratumoral heterogeneity, whereas asphericity characterizes global tumor shape irregularity. Together, these parameters provide complementary information regarding tumor biology.^5,6,7
Multiple studies across various types of cancer have shown that higher tumor asphericity is linked to more aggressive tumor behavior and worse clinical outcomes.^5,6,7,8,9 In non-small cell lung cancer, ASP provided independent prognostic information for both progression-free and overall survival, sometimes outperforming traditional PET parameters.⁸ Likewise, large-cohort research in head-and-neck and cervical cancers found that ASP enhances risk prediction when used alongside volumetric measurements, such as MTV.^7,9
Previous studies in colorectal cancer have associated radiomic and shape-based PET features, including asphericity, with survival outcomes and disease progression.¹⁰
Despite growing evidence in multiple tumor entities, the prognostic value of tumor asphericity in colorectal cancer remains insufficiently explored and has not yet been integrated into clinical risk stratification models. Given the clinical heterogeneity of CRC and the limitations of existing PET-derived parameters, evaluation of ASP may provide incremental prognostic information and improve the identification of high-risk patients.
Therefore, the aim of this study was to investigate the prognostic significance of tumor asphericity derived from pre-treatment 18F-FDG PET/CT in patients with colorectal cancer and to explore its potential to enhance prognostic stratification beyond conventional metabolic PET metrics.

Materials and Methods

Study Design and Patient PopulationPatients with histopathologically confirmed colorectal cancer who underwent pre-treatment 18F-FDG PET/CT at our institution between January 2020 and March 2025, within 6 weeks before therapy and who had a measurable primary tumor with available follow-up data were included. Patients were excluded if they had received prior oncologic treatment, had uncontrolled hyperglycemia (>200 mg/dL), had non-FDG-avid lesions, or had image artifacts that prevented reliable segmentation. Clinical information, including age, sex, tumor location, histopathology, grade, and TNM stage was retrieved from medical records. All patients received the best treatment based on their TNM stage.
PET/CT Acquisition ProtocolAll patients fasted for at least 6 hours before imaging. Blood glucose levels were confirmed to be <200 mg/dL prior to tracer administration. An intravenous injection of 18F-FDG (3.5–5.5 MBq/kg) was administered, followed by a resting uptake period of approximately 60 ± 10 minutes. Images were acquired using an integrated PET/CT scanner (Biograph mCT, Siemens Healthineers, Knoxville, USA). CT was performed for attenuation correction and anatomical localization (120 kV, automated mA). PET emission data were obtained from skull base to mid-thigh with 1.5 minutes per bed position and reconstructed using TrueX + Time-of-Flight (UltraHD-PET) with ordered-subset expectation maximization (OSEM) (2 iterations, 21 subsets), applying a 5-mm Gaussian post-reconstruction filter, into a 256 × 256 matrix.
Tumor Segmentation and Image AnalysisThe images on the PET/CT workstation were in DICOM format and transferred to the LIFEx software (Local Image Features Extraction, www.lifexsoft.org) on another computer for lesion segmentation and texture analysis. All VOI definitions and feature extractions were performed using this dedicated software to ensure consistency.¹¹
Clinical, metabolic, and radiomic features were extracted from pre-treatment 18F-FDG PET/CT images. Clinical parameters included age, sex, tumor location, and tumor size. Conventional PET metrics, including SUVmax, metabolic tumor volume (MTV), and total lesion glycolysis (TLG), were obtained from LIFEx software.
Shape-based heterogeneity was quantified using tumor asphericity (ASP). In addition, radiomic features reflecting intratumoral heterogeneity were calculated, including histogram-based entropy, GLCM-derived contrast and entropy, GLRLM-based short-run emphasis (SRE) and long-run emphasis (LRE), as well as GLSZM-based small zone emphasis (SZE) and large zone emphasis (LZE). All features were extracted using 3D analysis with 13 directions and a voxel distance of 1. Feature extraction followed IBSI definitions. For tumor segmentation, a threshold set at 40% of the maximum lesional metabolic activity was used. PET images were resampled to an isotropic voxel size of 4×4×4 mm. SUV intensities were rescaled using absolute bounds of 0–20 and discretized into 64 gray levels. Texture features were extracted in three dimensions (3D) from the segmented tumor volume, and only the largest connected tumor cluster was retained. Lesions containing fewer than 64 voxels were excluded to ensure the stability of radiomic features. When the primary tumor was near physiological high uptake areas, such as the bladder, or close to metastatic lymph nodes, VOI boundaries were carefully manually adjusted.
Follow-up and Outcome MeasuresPatients were followed clinically and radiologically at regular intervals according to institutional oncology protocols. The primary endpoint was Progression-free survival (PFS). The secondary endpoint was Overall survival (OS). PFS was defined as the time from pretreatment 18F-FDG PET/CT to disease progression, recurrence, or death from any cause, whichever occurred first. Patients without progression at last follow-up were censored on their last assessment. OS was the time from PET/CT to death from any cause; patients alive at last follow-up were censored on last contact. Patients’ data were obtained from the hospital information system.
Ethical ApprovalThe study was approved by the Sakarya University Faculty of Medicine Ethics Committee (Date: 16.03.2026, Decision No: E.572581).
Statistical AnalysisStatistical analyses were performed using SPSS software (version 24.0; IBM Corp., Armonk, NY, USA). The normality of continuous variables was assessed using the Shapiro–Wilk test. Continuous variables were expressed as mean ± standard deviation or median (min-max), as appropriate. Categorical variables were presented as frequencies and percentages. Comparisons between groups were performed using the independent samples t-test or Mann–Whitney U test for continuous variables, and the chi-square test or Fisher’s exact test for categorical variables. Continuous variables were primarily analyzed as continuous parameters in Cox proportional hazards regression models. Variables with p < 0.10 in univariate analysis were included in multivariate Cox regression models. Binary logistic regression analysis was performed to identify independent predictors of disease progression. Multicollinearity was assessed using the variance inflation factor (VIF) values, and highly correlated variables were not included in the same multivariable model. For Kaplan–Meier survival analysis, patients were dichotomized using two different approaches. First, a median-based dichotomization was applied to avoid overfitting and ensure balanced group sizes. Second, receiver operating characteristic (ROC) curve analysis was performed to determine optimal cut-off values based on the Youden index. Both dichotomization methods were used to generate Kaplan–Meier survival curves, and the results were compared to assess the consistency and robustness of the findings. A two-tailed p-value < 0.05 was considered statistically significant.
Reporting GuidelinesThis study was reported in accordance with the STROBE guideline.

Results

Patient Characteristics102 patients (59 men / 43 women; mean ± SD age 62.3 ± 14 years) underwent FDG PET/CT for colorectal tumor evaluation. Among them, 16 (%16) were at the right colon, 6 (%6) at the left colon, 5 (%5) at the sigmoid colon, 18 (%18) at the rectosigmoid colon, and 57 (%55) at the rectum. Of the 102 patients, 51 (50%) were stage 3, 33 (32.4%) were stage 4, and 18 (17.6%) were stage 2A or 2B. It was observed that 50 patients (49%) died in an average follow-up of 26.5 months (min-max;0.59-89.2 months). And 59 patients (57.8%) were observed progression events, such as relapse, progression, or death, with an average follow-up of 19.2 months (min-max; 0.59-89.2 months). Baseline clinical characteristics were summarized in Table 1.
Correlation Analysis Between Metabolic Tumor Parameters and Radiomic FeaturesSpearman correlation analysis was performed to evaluate the relationships between conventional PET parameters and radiomic features. A very strong positive correlation was observed between MTV and TLG (ρ = 0.816, p < 0.001), reflecting their shared volumetric-metabolic basis. SUVmax demonstrated a strong positive correlation with GLSZM-SZE (ρ = 0.765, p < 0.001), suggesting an association between higher metabolic activity and a greater prevalence of small, homogeneous-intensity zones within the tumor.
Among radiomic features, histogram-based entropy showed a strong positive correlation with GLRLM-SRE (ρ = 0.659, p < 0.001) and a moderate correlation with GLCM entropy (ρ = 0.442, p < 0.001), indicating that lesions with higher entropy exhibit greater textural heterogeneity.
Moderate positive correlations were found between MTV and Asphericity (ρ = 0.546, p < 0.001), as well as between TLG and Asphericity (ρ = 0.553, p < 0.001), suggesting that greater tumor burden is linked to increased morphological irregularity. Conversely, SUVmax was not significantly correlated with MTV (ρ = −0.073, p = 0.465), suggesting that peak uptake does not reflect tumor volume.
Overall, volumetric parameters showed strong interrelationships, while radiomic features demonstrated both intercorrelations and relative independence, suggesting that they capture complementary aspects of tumor heterogeneity.
Association of Tumor Heterogeneity Parameters with PFSThe distribution of clinical, metabolic, and radiomic parameters based on progression-free survival status is summarized in Table 2. Patients who experienced relapse, progression, or death showed significantly higher SUVmax values compared to those without progression (median: 25.6 vs. 21.3, p=0.018). Among shape-based parameters, asphericity was significantly elevated in the progression group (median: 0.29 vs. 0.25, p=0.024), indicating increased spatial heterogeneity in patients with poorer outcomes.
Regarding radiomic features, GLCM-derived entropy was significantly higher in patients with progression (median: 9.5 vs. 9.0, p<0.001), suggesting increased intratumoral heterogeneity. In contrast, GLSZM-SZE was significantly higher in the progression group (median: 0.73 vs. 0.69, p=0.001), indicating a predominance of small, fragmented intensity zones and thus greater intratumoral heterogeneity. Conversely, large zone emphasis (LZE) was significantly lower in patients with progression (median: 7.9 vs. 9.8, p=0.03), reflecting a relative loss of large, homogeneous regions. Together, these findings underscore the association between increased spatial heterogeneity and more aggressive tumor behavior (Table 2).
No significant differences were observed for MTV, TLG, GLCM contrast, or GLRLM-derived features (all p>0.05).
Univariate Analysis of Clinical and Radiomic Predictors of SurvivalUnivariate Cox regression analyses for progression-free survival (PFS) and overall survival (OS) are summarized in Supplement Table 1.
Among clinical variables, right-sided tumor location was linked to significantly worse outcomes for both PFS (HR: 2.2, p=0.017) and OS (HR: 2.3, p=0.014). Stage of cancer showed a strong relationship with survival (PFS: HR = 2.02, p < 0.001; OS: HR = 2.05, p < 0.001). Older age was significantly associated with worse OS (HR: 1.02, p = 0.026), but was not associated with PFS.
Among imaging-derived parameters, higher asphericity were significantly associated with poorer PFS (HR: 5.8, p=0.005) and poorer OS (HR: 7.9, p=0.003), suggesting that increased tumor shape irregularity is associated with poor survival outcomes. Furthermore, higher GLCM-based entropy was significantly associated with shorter survival (PFS: HR=1.74, p=0.006; OS: HR=1.9, p=0.002), reflecting increased intratumoral heterogeneity in aggressive tumors.
In addition, higher GLSZM-SZE was associated with increased risk, whereas higher GLSZM-LZE was associated with reduced risk (Supplement Table 1), indicating that a more fragmented and less homogeneous intratumoral texture pattern is linked to worse outcomes.
Histogram-based entropy, GLCM contrast, GLRLM-derived features, tumor diameter, and conventional metabolic parameters (SUVmax, MTV, TLG) were not significantly associated with survival outcomes (all p>0.05).
Multivariate Cox regression analyses for PFS & OSBefore conducting the Multivariate Cox regression analysis, the variance inflation factor (VIF) was used to assess multicollinearity among variables. All variables had low VIF values (around 1.0), indicating no significant multicollinearity in the model.
In a multivariate Cox regression analysis, Asphericity, GLCM entropy and TNM stage remained independent predictors of both PFS and OS (Omnibus Tests: p<0.001), highlighting the complementary prognostic value of clinical, shape-based, and texture-derived imaging features (Table 3).
ROC Curve analysis of Radiomic Predictors for PFSReceiver operating characteristic (ROC) curve analysis was applied to radiomic parameters that remained significant in multivariate Cox regression models (Supplement Table 2). The optimal cut-off for asphericity (0.36) achieved perfect specificity (100%) but limited sensitivity (29%), indicating strong rule-in capability for aggressive disease. In contrast, GLCM entropy (cut-off: 9.65) demonstrated a more balanced performance, with a sensitivity of 44% and specificity of 93%. GLCM entropy showed higher sensitivity than ASP while maintaining high specificity.
ROC curve analysis was also performed for additional radiomic parameters that were significant in univariate analysis, including GLSZM-SZE and GLSZM-LZE. However, only GLSZM-LZE had a meaningful Youden index with a cutoff of 10.44, whereas GLSZM-SZE did not , indicating limited discriminative ability for predicting PFS.
Kaplan–Meier Analysis of PFS According to Significant Radiomics Parameters
Kaplan–Meier analysis of asphericity using the ROC-derived cutoff value of 0.36 demonstrated a significant difference in PFS between groups (log-rank p < 0.001). Patients with high ASP had a markedly shorter median PFS than those with low ASP (9.2 vs. 37.4 months). When patients were dichotomized according to the median ASP value (0.27), the high ASP group also exhibited shorter PFS (14.0 vs. 31.5 months); however, this difference did not reach statistical significance (p = 0.066) (Figure 1).
Kaplan–Meier analysis of GLCM entropy using a ROC-derived cutoff value of 9.65 demonstrated significantly shorter PFS in patients with high GLCM entropy than those with low GLCM entropy (21.6 vs. 52.3 months) (log-rank p = 0.001). When patients were dichotomized according to the median GLCM entropy value (9.27), the median PFS was 14.0 months in the high-entropy group, whereas it was not reached in the low-entropy group (Figure 2). These findings suggest that increased textural heterogeneity, as reflected by higher GLCM entropy, is associated with an increased risk of disease progression and unfavorable clinical outcomes.

Discussion

Tumor heterogeneity influences aggressiveness and treatment resistance. Radiomics allows noninvasive measurement of this heterogeneity by analyzing tumor patterns.¹² Tumor asphericity (ASP) has proven prognostic in various cancers. In NSCLC and head and neck cancer, ASP independently predicts PFS and OS, often adding value beyond standard imaging.^5,6,7,8 Similar results in cervical cancer support ASP as a reliable biomarker.⁹
In contrast, evidence for colorectal cancer is limited and not yet conclusive. One study on colon cancer found that shape-related parameters such as asphericity, sphericity, and compactness were associated with disease-free survival. However, these factors lost significance in Cox regression analysis, emphasizing the need for further research.¹⁰ This gap in the literature provides the basis for the current study.
The present study demonstrates that both tumor asphericity (ASP), which represents macroscopic tumor heterogeneity, and GLCM-derived entropy, which reflects microscopic intratumoral heterogeneity, provide independent prognostic information in colorectal cancer. Importantly, these parameters remained significant after adjustment for TNM stage, indicating that imaging-derived heterogeneity captures biological characteristics that are not fully reflected by conventional anatomical staging.
In univariate analyses, ASP, GLCM-entropy, GLSZM-SZE, and GLSZM-LZE were significantly associated with both PFS and OS. However, after adjustment for potential confounders, only ASP and GLCM-entropy remained independent predictors in multivariate analyses. This finding suggests that shape-based and entropy-based heterogeneity metrics carry the most robust prognostic information, whereas GLSZM-derived features may provide complementary but not independent prognostic value. These results are consistent with previous studies reporting the superiority of heterogeneity-based features over single-voxel intensity metrics.
Our findings align with those of Tatlılıoğlu et al., who reported that asphericity and other shape-based parameters were significantly higher in colorectal cancer patients with disease progression than in those with disease-free status.¹⁰ In that study, ASP higher than 0.44 was associated with an HR of 7.6 for PFS, while Hofheinz et al. ¹³ and Apostolova et al. ⁶ identified lower ASP cut-offs (approximately 0.24–0.26) in head and neck cancer cohorts had a significantly higher PFS.
In the present study, the ROC-derived ASP cut-off of 0.36 identified patients with significantly shorter PFS. This threshold should be interpreted as a cohort-specific risk stratification value, while the main finding is the consistent association between higher ASP and poorer survival outcomes.
Previous NSCLC studies similarly identified ASP as an independent predictor of survival, supporting its biological relevance across tumor types.^5,14,15
The survival analyses further support the clinical importance of these findings. Patients with ASP values above the ROC-derived threshold of 0.36 experienced significantly shorter PFS compared to those with lower ASP values (9.2 vs. 37.4 months), with perfect specificity (100%) but limited sensitivity (29%). Similarly, higher GLCM entropy (cutoff of 9.65) identified a subgroup of patients with significantly poorer outcomes, with a sensitivity of 44% and a specificity of 93%. Notably, GLCM entropy showed a more balanced diagnostic profile than ASP, which achieved perfect specificity (100%), underscoring its potential as a highly specific marker for detecting patients at high risk of progression. Overall, these findings indicate that while ASP is a highly specific marker of aggressive disease, GLCM entropy offers a more balanced diagnostic capability and better overall discrimination by reflecting intratumoral heterogeneity at the voxel level. Similar findings have been reported by Burchardt et al., in which entropy-based parameters were associated with recurrence and treatment response, supporting their biological significance as indicators of tumor aggressiveness.¹⁶ Tatlıoğlu et al. also found that a GLCM-Entropy cutoff of 6.74 yielded an HR of 3.05 for disease-free survival in patients with newly diagnosed colon cancer.¹⁰
GLSZM-derived features, particularly LZE, demonstrated moderate predictive performance (PPV of 72.5% and an accuracy of 59.8%). Although their discriminative ability was lower compared to ASP and entropy, they may still contribute complementary structural information by reflecting the spatial distribution of homogeneous tumor regions. Together, these findings highlight that radiomic features derived from different domains provide complementary information rather than redundant signals.
Despite these findings, conventional metabolic parameters such as SUVmax did not show significant associations with survival outcomes. This suggests that single-voxel intensity measurements are insufficient to capture the complex biological behavior of tumors, further emphasizing the added value of radiomic heterogeneity features. This aligns with studies demonstrating ASP's prognostic value in head and neck, cervical and non-small cell lung cancer, whereas SUVmax has limited prognostic significance.^7,8,9
Tumor stage remained the strongest predictor of survival, as expected. However, the persistence of ASP and entropy as independent predictors in multivariate analysis indicates that radiomic features provide additional prognostic information beyond conventional staging. This supports the growing evidence that imaging-derived heterogeneity metrics can complement established clinical parameters. Şeker K. et al. showed that ASP values were significantly higher in late-stage small cell lung cancer (SCLC) than in early-stage SCLC, indicating that tumors in the extensive stage exhibit higher geometric irregularity ¹⁷ .
GLSZM-derived parameters quantify the distribution of homogeneous intensity zones within a tumor, with small zone emphasis (SZE) reflecting fine, fragmented heterogeneity and large zone emphasis (LZE) indicating the presence of larger, more homogeneous regions.¹² In our study, GLSZM-derived SZE and LZE, showed significant associations with survival outcomes, further supporting the role of intratumoral heterogeneity in colorectal cancer. Higher SZE reflects increased intratumoral heterogeneity (genomic instability); the more distinct "zones" a tumor has, the more likely it is to harbor subpopulations of cells resistant to treatment. and is generally associated with a poorer prognosis. Whereas higher LZE indicates more homogeneous tumor regions, though its prognostic implications may vary with metabolic activity and tumor burden.
Although GLSZM features have been less consistently reported compared to entropy-based metrics, they provide complementary information by characterizing the size distribution of homogeneous regions within the tumor. As in our study, Vural TÖ et al., investigating the value of radiomics features compared to tumor-infiltrating lymphocytes (TILs) among patients with breast cancer, found that GLSZM Small Zone Emphasis was independently associated with TIL grouping.¹⁸
In contrast to these studies, Prasanth et al.showed that all 11 Gray Level Zone Length Matrix (GLZLM) features were insignificant between responders and non-responders in patients with non-small cell lung cancer. Similarly, among the 11 GLRLM features, only GLNU (gray level non-uniformity) demonstrated predictive value. Among the shape-based features, only “surface” was found to be predictive. Other parameters such as volume, sphericity, and compacity did not demonstrate sufficient discriminative ability. Conventional parameters such as SUVmax and TLG did not show statistically significant differences, as in our study.¹⁹

Limitations

Despite promising findings, some limitations should be acknowledged. The retrospective, single-center design may limit generalizability. The relatively small sample size may have contributed to wide confidence intervals, particularly for ASP. Additionally, the inclusion of tumors from different colorectal subsites may have introduced biological heterogeneity, as tumor behavior varies across anatomical locations. In addition, radiomic analyses remain sensitive to image acquisition, reconstruction, segmentation, and discretization methods. Although standardized IBSI-compliant feature extraction procedures were applied in the present study, external validation in independent cohorts is necessary before clinical implementation. Future studies on more homogeneous tumor groups could offer clearer insights into radiomic features.

Conclusion

Asphericity and GLCM-derived entropy derived from pretreatment 18F-FDG PET/CT offer complementary prognostic information beyond conventional metabolic parameters such as SUVmax, MTV, and TLG in colorectal cancer. The independent prognostic value of ASP highlights the significance of macroscopic tumor structure, while entropy reflects intratumoral complexity at the voxel level. Together, these results support a multi-scale heterogeneity model for improved risk assessment. Incorporation of these radiomic biomarkers into clinical risk stratification models may improve the identification of high-risk patients who could benefit from intensified surveillance and personalized therapeutic strategies. Further prospective multicenter studies with external validation cohorts are warranted to confirm these findings and establish standardized radiomic biomarkers for routine clinical application in colorectal cancer.

Declarations

Abbreviations

ASP: Asphericity
CRC: Colorectal cancer
FDG: Fluorodeoxyglucose
GLCM: Gray-level co-occurrence matrix
GLRLM: Gray-level run length matrix
GLSZM: Gray-level size zone matrix
MTV: Metabolic tumor volume
OS: Overall survival
PET/CT: Positron emission tomography/computed tomography
PFS: Progression-free survival
ROC: Receiver operating characteristic
SUVmax: Maximum standardized uptake value
TLG: Total lesion glycolysis

References

1. Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209-249. doi:10.3322/caac.21660
   Article PubMed Google Scholar
2. Kasi PM, Hubbard JM, Grothey A. Selection of biologics for patients with metastatic colorectal cancer: the role of predictive markers. Expert Rev Gastroenterol Hepatol. 2015;9(2):273-276. doi:10.1586/17474124.2015.1001743
   Article PubMed Google Scholar
3. Chicklore S, Goh V, Siddique M, et al. Quantifying tumour heterogeneity in 18F-FDG PET/CT imaging by texture analysis. Eur J Nucl Med Mol Imaging. 2013;40(1):133-140. doi:10.1007/s00259-012-2247-0
   Article PubMed Google Scholar
4. Laurens ST, Oyen WJG. Impact of fluorodeoxyglucose PET/computed tomography on the management of patients with colorectal cancer. PET Clin. 2015;10(3):345-360. doi:10.1016/j.cpet.2015.03.007
   Article PubMed Google Scholar
5. Apostolova I, Rogasch J, Buchert R, et al. Quantitative assessment of the asphericity of pretherapeutic FDG uptake as an independent predictor of outcome in NSCLC. BMC Cancer. 2014;14:896. doi:10.1186/1471-2407-14-896
   Article PubMed Google Scholar
6. Apostolova I, Steffen IG, Wedel F, et al. Asphericity of pretherapeutic tumour FDG uptake provides independent prognostic value in head-and-neck cancer. Eur Radiol. 2014;24(9):2077-2087. doi:10.1007/s00330-014-3269-8
   Article PubMed Google Scholar
7. Hausmann P, Zschaeck S, Furth C, et al. Tumor asphericity in FDG PET is an independent prognostic parameter improving risk stratification in patients with head and neck squamous cell carcinoma. J Nucl Med. 2025;66(5):686-691. doi:10.2967/jnumed.124.268972
   Article PubMed Google Scholar
8. Rogasch JMM, Furth C, Chibolela C, et al. Validation of independent prognostic value of asphericity of 18F-fluorodeoxyglucose uptake in non-small-cell lung cancer patients undergoing treatment with curative intent. Clin Lung Cancer. 2020;21(3):264-272.e6. doi:10.1016/j.cllc.2019.10.001
   Article PubMed Google Scholar
9. Cegla P, Hofheinz F, Burchardt E, et al. Asphericity derived from [18F]FDG PET as a new prognostic parameter in cervical cancer patients. Sci Rep. 2023;13(1):8423. doi:10.1038/s41598-023-35191-8
   Article PubMed Google Scholar
10. Ural Tatlilioğlu B, Katipoğlu K, Okudan B. Contribution of texture features and LC3B immunohistochemical staining in colectomy materials to prognosis prediction in newly diagnosed colon cancer. Nucl Med Commun. 2025;46(10):980-988. doi:10.1097/mnm.0000000000002025
    Article PubMed Google Scholar
11. Nioche C, Orlhac F, Boughdad S, et al. LIFEx: a freeware for radiomic feature calculation in multimodality imaging to accelerate advances in the characterization of tumor heterogeneity. Cancer Res. 2018;78(16):4786-4789. doi:10.1158/0008-5472.can-18-0125
    Article PubMed Google Scholar
12. Ha S, Choi H, Paeng JC, et al. Radiomics in oncological PET/CT: a methodological overview. Nucl Med Mol Imaging. 2019;53(1):14-29. doi:10.1007/s13139-019-00571-4
    Article PubMed Google Scholar
13. Hofheinz F, Lougovski A, Zöphel K, et al. Increased evidence for the prognostic value of primary tumor asphericity in pretherapeutic FDG PET for risk stratification in patients with head and neck cancer. Eur J Nucl Med Mol Imaging. 2015;42(3):429-437. doi:10.1007/s00259-014-2953-x
    Article PubMed Google Scholar
14. Cook GJ, Yip C, Siddique M, et al. Are pretreatment 18F-FDG PET tumor textural features in non-small cell lung cancer associated with response and survival after chemoradiotherapy? J Nucl Med. 2013;54(1):19-26. doi:10.2967/jnumed.112.107375
    Article PubMed Google Scholar
15. Tixier F, Hatt M, Valla C, et al. Visual versus quantitative assessment of intratumor 18F-FDG PET uptake heterogeneity: prognostic value in non-small cell lung cancer. J Nucl Med. 2014;55(8):1235-1241. doi:10.2967/jnumed.113.133389
    Article PubMed Google Scholar
16. Burchardt E, Bos-Liedke A, Serkowska K, Cegla P, Piotrowski A, Malicki J. Value of [18F]FDG PET/CT radiomic parameters in the context of response to chemotherapy in advanced cervical cancer. Sci Rep. 2023;13(1):9092. doi:10.1038/s41598-023-35843-9
    Article PubMed Google Scholar
17. Şeker K, Ulaş Ö, Hasbek Z. Relationship between metabolic parameters and asphericity derived from 18F-FDG PET/CT and stage in small cell lung cancer. Rev Esp Med Nucl Imagen Mol (Engl Ed). 2026;45(3):500276. doi:10.1016/j.remnie.2025.500276
    Article PubMed Google Scholar
18. Vural Topuz Ö, Bağbudar S, Aksu A, Söylemez Akkurt T, Akkaş BE. Radiomic signatures derived from baseline 18F-FDG PET/CT imaging can predict tumor-infiltrating lymphocyte values in patients with primary breast cancer. Nuklearmedizin. 2025;64(3):194-204. doi:10.1055/a-2512-8212
    Article PubMed Google Scholar
19. Prasanth PN, Ravina M, Ganga RT, et al. Role of [18F]fluorodeoxyglucose positron emission tomography-computed tomography metabolic radiomics in predicting response to chemotherapy in patients with non-small cell lung cancer. Nucl Med Commun. Published online February 20, 2026. doi:10.1097/mnm.0000000000002127
    Article PubMed Google Scholar

Additional Information

Publisher’s Note
Bayrakol MP remains neutral with regard to jurisdictional and institutional claims.

Rights and Permissions

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). To view a copy of the license, visit https://creativecommons.org/licenses/by-nc/4.0/

View full license details →

About This Article

Received:

June 4, 2026

Published Online:

June 21, 2026