Single-Photon Emission Computed Tomography/Computed Tomography Image-Based Radiomics for Discriminating Vertebral Bone Metastases From Benign Bone Lesions in Patients With Tumors

Abstract

Purpose: The purpose of this study was to investigate the feasibility of Single-Photon Emission Computed Tomography/Computed Tomography (SPECT/CT) image-based radiomics in differentiating bone metastases from benign bone lesions in patients with tumors. Methods: A total of 192 lesions from 132 patients (134 in the training group, 58 in the validation group) diagnosed with vertebral bone metastases or benign bone lesions were enrolled. All images were evaluated and diagnosed independently by two physicians with more than 20 years of diagnostic experience for qualitative classification, the images were imported into MaZda software in Bitmap (BMP) format for feature extraction. All radiomics features were selected by least absolute shrinkage and selection operator (LASSO) regression and 10-fold cross-validation algorithms after the process of normalization and correlation analysis. Based on these selected features, two models were established: The CT model and SPECT model (radiomics features were derived from CT and SPECT images, respectively). In addition, a combination model (ComModel) combined CT and SPECT features was developed in order to better evaluate the predictive performance of radiomics models. Subsequently, the diagnostic performance between each model was separately evaluated by a confusion matrix. Results: There were 12, 13, and 18 features contained within the CT, SPECT, and ComModel, respectively. The constructed radiomics models based on SPECT/CT images to discriminate between bone metastases and benign bone lesions not only had high diagnostic efficacy in the training group (AUC of 0.894, 0.914, 0.951 for CT model, SPECT model, and ComModel, respectively), but also performed well in the validation group (AUC; 0.844, 0.871, 0.926). The AUC value of the human experts was 0.849 and 0.839 in the training and validation groups, respectively. Furthermore, both SPECT model and ComModel show higher classification performance than human experts in the training group (P = 0.021 and P = 0.001, respectively) and the validation group (P = 0.037 and P = 0.007, respectively). All models showed better diagnostic accuracy than human experts in the training group and the validation group. Conclusion: Radiomics derived from SPECT/CT images could effectively discriminate between bone metastases and benign bone lesions. This technique may be a new non-invasive way to help prevent unnecessary delays in diagnosis and a potential contribution in disease staging and treatment planning.

Keywords: SPECT/CT; benign bone lesions; bone metastases; diagnosis; radiomics.

Figure 1

The inclusion and exclusion criteria of our study.

Figure 2

The flowchart of our study.

Figure 3

(A–F) demonstrated the specific process of least absolute shrinkage and selection operator (LASSO) regression analysis screening features for CT model, SPECT model, and ComModel, respectively. (A,C,E) showed the process of features selection. The vertical line was plotted at the optimal λ of 0.064, 0.025, and 0.035 for CT, SPECT, and ComModel, respectively. Twelve, thirteen, and eighteen factors with non-zero coefficients were finally selected for CT, SPECT, and ComModel, respectively. (B,D,F) showed that features selection was performed by 10-fold cross-validation with the criterion of minimum deviance.

Figure 4

Comparison of the calibration curve and Brier score of different models. All three model's calibration curves were closed to ideal curves, indicating that the models had good fitness and predictive ability. The following figure shows the distribution of the probability of diagnosis for different models.

Figure 5

Comparison of decision curve analysis (DCA) of different models. When the threshold was 0–1, the ComModel always had a better overall net clinical gain than the other models, the SPECT model also had higher clinical gain than the human experts, and there was no significant difference between the CT model and the human experts.

Figure 6

Clinical cases SPECT/CT images of bone metastases (A) and benign bone lesions (B). The images shown are WBS image, axial CT, SPECT, fusion image, and sagittal CT (a–e, respectively). (A) bone metastases: a 53-year-old female with an adenocarcinoma of the left lung. Wedge-like changes of the T8 vertebral body with an abnormal concentration of radioactive tracer (arrows). (B) benign bone lesions: a 68-year-old female with breast cancer. Wedge-like changes of the L1 vertebral body with higher bone density and increased radioactive tracer distribution (arrows). It was difficult to determine whether lesions were metastasis with conventional images only. Lesion (A) was confirmed as pathological fracture due to bone metastases by pathological examination and showed systemic bone metastases at subsequent imaging follow-up. Lesion (B) was confirmed to be a benign compression fracture by imaging follow-up and clinical information.

Figure 7

(A) bone metastases: a 76-year-old male with prostate cancer. Nodular high-density shadow on the right lower edge of the L2 vertebral body with a concentrated radioactive tracer (arrows). (B) benign bone lesions: a 60-year-old male with prostate cancer. Nodular high-density shadow on the right upper edge of the T12 vertebral body with increased radioactive tracer distribution (arrows). The cluster of radioactive tracer concentrations in the left rib(a), with fusion images suggesting bone metastasis. Lesion (A) showed increased concentration of tracer and increased extent of concentration with systemic bone metastases at subsequent imaging follow-up. Lesion (B) was confirmed not a metastasis from prostate cancer at several subsequent imaging follow-ups.

Figure 8

Comparison of the diagnostic performance of different models. (A) Receiver operating characteristic for training cohort. (B) Receiver operating characteristic for validation cohort.