Improving the Classification of PCNSL and Brain Metastases by Developing a Machine Learning Model Based on 18F-FDG PET

Abstract

Background: The characteristic magnetic resonance imaging (MRI) and the positron emission tomography (PET) findings of PCNSL often overlap with other intracranial tumors, making definitive diagnosis challenging. PCNSL typically shows iso-hypointense to grey matter on T2-weighted imaging. However, a particular part of PCNSL can demonstrate T2-weighted hyperintensity as other intracranial tumors. Moreover, normal high uptake of FDG in the basal ganglia, thalamus, and grey matter can mask underlying PCNSL in 18F-FDG PET. In order to promote the efficiency of diagnosis, the MRI-based or PET/CT-based radiomics models combining histograms with texture features in diagnosing glioma and brain metastases have been widely established. However, the diagnosing model for PCNSL has not been widely reported. The study was designed to investigate a machine-learning (ML) model based on multiple parameters of 2-deoxy-2-[18F]-floor-D-glucose (18F-FDG) PET for differential diagnosis of PCNSL and metastases in the brain. Methods: Patients who underwent an 18F-FDG PET scan with untreated PCNSL or metastases in the brain were included between May 2016 and May 2022. A total of 126 lesions from 51 patients (43 patients with untreated brain metastases and eight patients with untreated PCNSL), including 14 lesions of PCNSL, and 112 metastatic lesions in the brain, met the inclusion criteria. PCNSL or brain metastasis was confirmed after pathology or clinical history. Principal component analysis (PCA) was used to decompose the datasets. Logistic regression (LR), support vector machine (SVM), and random forest classification (RFC) models were trained by two different groups of datasets, the group of multi-class features and the group of density features, respectively. The model with the highest mean precision score was selected. The testing sets and original data were used to examine the efficacy of models separately by using the weighted average F1 score and area under the curve (AUC) of the receiver operating characteristic curve (ROC). Results: The multi-class features-based RFC and SVM models reached identical weighted-average F1 scores in the testing set, and the score was 0.98. The AUCs of RFC and SVM models calculated from the testing set were 1.00 equally. Evaluated by the original dataset, the RFC model based on multi-class features performs better than the SVM model, whose weighted-average F1 scores of the RFC model calculated from the original data were 0.85 with an AUC of 0.93. Conclusions: The ML based on multi-class features of 18F-FDG PET exhibited the potential to distinguish PCNSL from brain metastases. The RFC models based on multi-class features provided comparatively high efficiency in our study.

The effect of harmonization on the variability of PET radiomic features extracted using various segmentation methods

“…For each radiomic feature, the intraclass correlation coefficient (ICC) was determined using R version 4.0.4 (The R Foundation, Vienna, Austria) both before and after harmonization. Considering Koo and Li’s guideline [ 48 ], two-way random effects with complete agreement and multi-raters were conducted to calculate the ICC. An ICC of less than 0.5, 0.5 ≤ ICC < 0.75, 0.75 ≤ ICC < 0.9, and ICC ≥ 0.9 reflect low, rational, promising, and outstanding reliability, respectively.…”

An analytic hierarchy process method weighted by an improved Random Forest model considering sample optimization selection for the evaluation of landslide susceptibility

“…In addition to these four common evaluation indicators, F1-Score is an essential evaluation indicator that cannot be ignored. It is the harmonic mean of the recall and accuracy rates (Cui et al 2023). The F1-Score in the established CF-combination model is 85.74%.…”

Clinical application of 18F-fluorodeoxyglucose positron emission tomography/computed tomography radiomics-based machine learning analyses in the field of oncology

“…Recently, 18 F-FDG PET/CT radiomics-based ML analysis has been applied to overcome these issues [ 75 ]. Previous studies have revealed that 18 F-FDG PET/CT radiomics-based ML analysis is useful in not only classifying tumors based on histological subtypes but also differentiating malignant lymphoma from other diseases [ 76 – 80 ].…”

“…AUC: 0.94 2. AUC: 0.95 Yang et al [ 79 ] 2023 Cervical lymph node Malignant lymphoma vs. metastasis n = 165 CNN model Combined PET radiomics-based + CNN-based model alone SVM Combined model Training and validation cohorts AUC: 0.948 Cui et al [ 80 ] 2023 Brain tumor Malignant lymphoma vs. metastasis n = 51 PET radiomics-based model alone RF – Training and validation cohorts AUC: 0.93 Predicting treatment response or survival Frood et al [ 84 ] 2022 DLBCL Recurrence after chemotherapy n = 229 Combined clinical + PET radiomics-based model alone Ridge regression – Training and validation cohorts AUC: 0.73 Cui et al [ 85 ] 2022 DLBCL PFS after chemotherapy n = 271 Clinical model PET radiomics-based model Combined clinical + PET radiomics-based model alone RF + cox proportional hazard Combined model Training and validation cohorts C-index: 0.853 Frood et al [ 86 ] 2022 HD Recurrence after chemotherapy or RT n = 289 Combined clinical + PET radiomics-based model alone …”