Prediction of Anal Cancer Recurrence After Chemoradiotherapy Using Quantitative Image Features Extracted From Serial 18F-FDG PET/CT

Jiahui Wang, Hao Zhang, Michael D. Chuong, Kujtim Latifi, Shan Tan, Wookjin Choi, Sarah E. Hoffe, Ravi Shridhar, Wei LüFrontiers in Oncology 2019 · DOI

Abstract

We extracted image features from serial <sup>18</sup>F-labeled fluorodeoxyglucose (FDG) positron emission tomography (PET) / computed tomography (CT) scans of anal cancer patients for the prediction of tumor recurrence after chemoradiation therapy (CRT). Seventeen patients (4 recurrent and 13 non-recurrent) underwent three PET/CT scans at baseline (Pre-CRT), in the middle of the treatment (Mid-CRT) and post-treatment (Post-CRT). Comprehensive image features were extracted from CT and PET (SUV) images within manually delineated gross tumor volume, including geometry features, intensity features and texture features. The difference of feature values between two time points were also computed and analyzed. In univariate analysis, six geometry, three intensity, and six texture features were identified as significant predictors of tumor recurrence. A geometry feature of Roundness between Post-CRT and Pre-CRT CTs was identified as the most important predictor with an AUC value of 1.00 by multivariate logistic regression model. To investigate the early prediction ability, we used features only from Pre-CRT and Mid-CRT scans. Orientation (geometry feature) of Pre-CRT SUV, Mean (intensity feature) of Pre-CRT CT, and Mean of Long Run High Gray Level Emphasis (LRHGLE) (texture feature) of Pre-CRT CT were identified as the most important feature set (AUC = 1.00) by multivariate logistic regression model. The experimental results demonstrated the potential of serial PET/CT scans in early prediction of anal tumor recurrence.

Introduction

Anal cancer treatment primarily relies on chemoradiotherapy (CRT). While effective for many, identifying patients at risk of recurrence remains a clinical challenge. Standard post-treatment assessments, such as digital rectal exams and biopsies performed 8–12 weeks after therapy, are often confounded by CRT-induced inflammation and tissue changes (mucositis/dermatitis). This study investigates the use of radiomics—the quantitative extraction of high-dimensional data from medical images—to predict tumor recurrence by analyzing serial <sup>18</sup>F-FDG PET/CT scans, potentially identifying high-risk patients much earlier than current clinical standards allow.

Methods

The study analyzed a retrospective cohort of 17 patients (4 recurrent, 13 non-recurrent) who underwent PET/CT imaging at three intervals: Pre-CRT (baseline), Mid-CRT (during treatment), and Post-CRT (after treatment).

Serial CT imaging showing tumor delineation in axial and coronal views at Pre-CRT and Post-CRT time points.

To ensure consistency, Mid-CRT and Post-CRT scans were aligned to the Pre-CRT baseline using B-spline deformable registration. Researchers extracted 46 quantitative features from both CT and PET (SUV) images within the manually delineated gross tumor volume (GTV). These features included geometry (shape/size), intensity (SUV/CT values), and texture (spatial voxel patterns). "Delta radiomics"—the change in feature values between time points—was also analyzed using univariate and multivariate logistic regression, as well as a Naïve Bayesian classifier.

Key Results and Insights

The analysis identified several significant predictors of recurrence. The most robust predictor was the change in "Roundness" (a geometry feature) on CT scans between the Pre-CRT and Post-CRT phases. A multivariate logistic regression model using this feature achieved an Area Under the Curve (AUC) of 1.00.

Post-CRT PET/CT fusion image used for extracting metabolic intensity and texture features.

A key insight from the geometry analysis is the "asymmetric regression" of non-recurrent tumors. While successful treatment often results in irregular tumor shrinkage (decreasing roundness), recurrent tumors tended to maintain a more spherical shape or even increase in roundness as they resisted therapy.

Furthermore, the study demonstrated the feasibility of early prediction. Using only Pre-CRT and Mid-CRT data, a multivariate model incorporating SUV orientation, CT intensity, and texture features (specifically Long Run High Gray Level Emphasis, or LRHGLE) also reached an AUC of 1.00. The study found that complex texture features like LRHGLE provided more predictive value than traditional metrics like SUVmax, likely because they capture subtle spatial variations in tumor metabolism and density.

Conclusion

Quantitative radiomic features extracted from serial PET/CT scans can accurately predict anal cancer recurrence. The ability to identify high-risk patients as early as mid-treatment could allow for personalized adjustments to therapy, potentially improving outcomes for those resistant to standard CRT. While limited by a small sample size, these results warrant further validation in larger prospective cohorts to confirm the clinical utility of features like "Roundness" and texture-based markers.

Summary updated

Abstract

We extracted image features from serial <sup>18</sup>F-labeled fluorodeoxyglucose (FDG) positron emission tomography (PET) / computed tomography (CT) scans of anal cancer patients for the prediction of tumor recurrence after chemoradiation therapy (CRT). Seventeen patients (4 recurrent and 13 non-recurrent) underwent three PET/CT scans at baseline (Pre-CRT), in the middle of the treatment (Mid-CRT) and post-treatment (Post-CRT) were included. For each patient, Mid-CRT and Post-CRT scans were aligned to Pre-CRT scan. Comprehensive image features were extracted from CT and PET (SUV) images within manually delineated gross tumor volume, including geometry features, intensity features and texture features. The difference of feature values between two time points were also computed and analyzed. We employed univariate logistic regression model, multivariate model, and naïve Bayesian classifier to analyze the image features and identify useful tumor recurrent predictors. The area under the receiver operating characteristic (ROC) curve (AUC) was used to evaluate the accuracy of the prediction. In univariate analysis, six geometry, three intensity, and six texture features were identified as significant predictors of tumor recurrence. A geometry feature of Roundness between Post-CRT and Pre-CRT CTs was identified as the most important predictor with an AUC value of 1.00 by multivariate logistic regression model. The difference of Number of Pixels on Border (geometry feature) between Post-CRT and Pre-CRT SUVs and Elongation (geometry feature) of Post-CRT CT were identified as the most useful feature set (AUC = 1.00) by naïve Bayesian classifier. To investigate the early prediction ability, we used features only from Pre-CRT and Mid-CRT scans. Orientation (geometry feature) of Pre-CRT SUV, Mean (intensity feature) of Pre-CRT CT, and Mean of Long Run High Gray Level Emphasis (LRHGLE) (texture feature) of Pre-CRT CT were identified as the most important feature set (AUC = 1.00) by multivariate logistic regression model. Standard deviation (intensity feature) of Mid-CRT SUV and difference of Mean of LRHGLE (texture feature) between Mid-CRT and Pre-CRT SUVs were identified as the most important feature set (AUC = 0.86) by naïve Bayesian classifier. The experimental results demonstrated the potential of serial PET/CT scans in early prediction of anal tumor recurrence.
BibTeX
@inproceedings{wang2019prediction,
  author = {Jiahui Wang and Hao Zhang and Michael D. Chuong and Kujtim Latifi and Shan Tan and Wookjin Choi and Sarah E. Hoffe and Ravi Shridhar and Wei Lü},
  title = {{Prediction of Anal Cancer Recurrence After Chemoradiotherapy Using Quantitative Image Features Extracted From Serial 18F-FDG PET/CT}},
  booktitle = {Frontiers in Oncology},
  year = {2019},
  doi = {10.3389/fonc.2019.00934},
}
Figure