Role of Machine Learning (ML)-Based Classification Using Conventional 18F-FDG PET Parameters in Predicting Postsurgical Features of Endometrial Cancer Aggressiveness

Abstract

Purpose: to investigate the preoperative role of ML-based classification using conventional ¹⁸F-FDG PET parameters and clinical data in predicting features of EC aggressiveness.

Methods: retrospective study, including 123 EC patients who underwent ¹⁸F-FDG PET (2009-2021) for preoperative staging. Maximum standardized uptake value (SUVmax), SUVmean, metabolic tumour volume (MTV), and total lesion glycolysis (TLG) were computed on the primary tumour. Age and BMI were collected. Histotype, myometrial invasion (MI), risk group, lymph-nodal involvement (LN), and p53 expression were retrieved from histology. The population was split into a train and a validation set (80-20%). The train set was used to select relevant parameters (Mann-Whitney U test; ROC analysis) and implement ML models, while the validation set was used to test prediction abilities.

Results: on the validation set, the best accuracies obtained with individual parameters and ML were: 61% (TLG) and 87% (ML) for MI; 71% (SUVmax) and 79% (ML) for risk groups; 72% (TLG) and 83% (ML) for LN; 45% (SUVmax; SUVmean) and 73% (ML) for p53 expression.

Conclusions: ML-based classification using conventional ¹⁸F-FDG PET parameters and clinical data demonstrated ability to characterize the investigated features of EC aggressiveness, providing a non-invasive way to support preoperative stratification of EC patients.

Keywords: 18F-FDG PET; endometrial cancer; imaging parameters; machine learning; prognostic value.

Figure A1

Boxplots showing distributions of PET parameters according to the different tomographs before and after ComBat harmonization. Batches: 1 = fully hybrid 3 T PET/MRI system (SIGNA PET/MRI; GE Healthcare); 2 = Discovery STE (GE Healthcare); 3 = Discovery ST (GE Healthcare); 4 = Discovery 690 (GE Healthcare); 5 = Gemini–GXL (Philips Medical Systems); SUV: standardized uptake value; MTV: metabolic tumour volume; TLG: total lesion glycolysis.

Figure 1

Distributions of ¹⁸F-FDG PET and clinical parameters’ values with respect to the different features of EC aggressiveness. Boxplots represent parameter’s distribution according to: (a) presence of deep myometrial invasion (>50%); (b) risk group classification; (c) presence of lymph nodes (LN) involvement; (d) p53 expression. Differences that are statistically significant (Mann-Whitney U test’s adjusted p-value < 0.05) are marked with a red *. SUV = standardized uptake value; MTV = metabolic tumour volume; TLG = total lesion glycolysis; BMI = body mass index.

Figure 2

ROC curves of ¹⁸F–FDG PET and clinical parameters as predictors of the different features of EC aggressiveness. Solid lines represent the area under the curve (AUC) obtained for SUVmax (light blue). SUVmean (orange). MTV (green). TLG (red) and age (purple) in predicting: (a) presence of deep myometrial invasion (>50%); (b) risk group classification; (c) presence of lymph nodes involvement; (d) p53 expression. AUC values coupled with a statistically significant 95% confidence interval are marked with *. SUV = standardized uptake value; MTV = metabolic tumour volume; TLG = total lesion glycolysis.

Figure 3

¹⁸F–FDG PET performed for EC staging. A 71-year-old patient with endometrial cancer (Stage: III C1; Grade: 3; histotype: non-endometrioid EC; MI: 85%; risk group: high–intermediate/high; presence of LNs) who underwent ¹⁸F–FDG PET/CT for staging purpose. Red arrows indicate pathological uptake in correspondence of the primary tumour. green arrows indicate pathological uptake in correspondence of bilateral iliac–obturator lymph nodes and yellow arrows indicate pathological uptake in correspondence of lomboaortic and interaortocaval lymph nodes ((a): MIP; (b–d): transaxial PET/CT images). PET parameters of the primary tumour were as follow: SUVmax = 17.43; SUVmean = 11.82; MTV = 28.60; TLG = 338.04. Histological analysis: the tumor is constituted by large cells with a high grade of nuclear atypia and numerous mitotic figures (e). The tumoral growth is mainly in papillary projections. Myometrial infiltration (f) has a tubulo–glandular architecture with micro–papillary structure into the lumen; the way of myometrial invasion is infiltrative/destructive. Lymph nodal metastasis (g) is nodular and constituted by serous atypical cells arranged in cords and small nests.