1
|
Dutta A, Chan J, Haworth A, Dubowitz DJ, Kneebone A, Reynolds HM. Robustness of magnetic resonance imaging and positron emission tomography radiomic features in prostate cancer: Impact on recurrence prediction after radiation therapy. Phys Imaging Radiat Oncol 2024; 29:100530. [PMID: 38275002 PMCID: PMC10809082 DOI: 10.1016/j.phro.2023.100530] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/28/2023] [Revised: 12/21/2023] [Accepted: 12/29/2023] [Indexed: 01/27/2024] Open
Abstract
Background and purpose Radiomic features from MRI and PET are an emerging tool with potential to improve prostate cancer outcomes. However, feature robustness due to image segmentation variations is currently unknown. Therefore, this study aimed to evaluate the robustness of radiomic features with segmentation variations and their impact on predicting biochemical recurrence (BCR). Materials and methods Multi-scanner, pre-radiation therapy imaging from 142 patients with localised prostate cancer was used. Imaging included T2-weighted (T2), apparent diffusion coefficient (ADC) MRI, and prostate-specific membrane antigen (PSMA)-PET. The prostate gland and intraprostatic tumours were manually and automatically segmented, and differences were quantified using Dice Coefficient (DC). Radiomic features including shape, first-order, and texture features were extracted for each segmentation from original and filtered images. Intraclass Correlation Coefficient (ICC) and Mean Absolute Percentage Difference (MAPD) were used to assess feature robustness. Random forest (RF) models were developed for each segmentation using robust features to predict BCR. Results Prostate gland segmentations were more consistent (mean DC = 0.78) than tumour segmentations (mean DC = 0.46). 112 (3.6 %) radiomic features demonstrated 'excellent' robustness (ICC > 0.9 and MAPD < 1 %), and 480 features (15.4 %) demonstrated 'good' robustness (ICC > 0.75 and MAPD < 5 %). PET imaging provided more features with excellent robustness than T2 and ADC. RF models showed strong predictive power for BCR with a mean area under the receiver-operator-characteristics curve (AUC) of 0.89 (range 0.85-0.93). Conclusion When using radiomic features for predictive modelling, segmentation variability should be considered. To develop BCR predictive models, radiomic features from the entire prostate gland are preferable over tumour segmentation-based features.
Collapse
Affiliation(s)
- Arpita Dutta
- Auckland Bioengineering Institute, The University of Auckland, Auckland, New Zealand
| | - Joseph Chan
- Department of Radiation Oncology, Royal North Shore Hospital, Sydney, New South Wales, Australia
| | - Annette Haworth
- Institute of Medical Physics, School of Physics, University of Sydney, Sydney, New South Wales, Australia
| | - David J. Dubowitz
- Department of Anatomy and Medical Imaging, Faculty of Medical and Health Sciences, The University of Auckland, Auckland, New Zealand
- Centre for Advanced MRI, The University of Auckland, Auckland, New Zealand
| | - Andrew Kneebone
- Department of Radiation Oncology, Royal North Shore Hospital, Sydney, New South Wales, Australia
| | - Hayley M. Reynolds
- Auckland Bioengineering Institute, The University of Auckland, Auckland, New Zealand
| |
Collapse
|
2
|
Yazdanpanah O, Mahadevan A, Sharma A, Benjamin DJ, Kalebasty AR. A comparative study of spinal cord compression management in metastatic prostate cancer: Teaching versus non-teaching hospitals in the United States. Cancer Med 2024; 13:e6845. [PMID: 38146897 PMCID: PMC10807570 DOI: 10.1002/cam4.6845] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/11/2023] [Revised: 12/04/2023] [Accepted: 12/10/2023] [Indexed: 12/27/2023] Open
Abstract
BACKGROUND Spinal cord compression (SCC) in metastatic prostate cancer (MPC) is a critical complication and multiple factors influence the optimal therapeutic strategy. We investigated the differences in practice patterns between teaching hospitals (TH) and non-teaching hospitals (NTH) across the United States. METHOD Using the National Inpatient Sample Database (NIS), we performed a retrospective study on hospitalizations with MPC and SCC between 2016 and 2020 in US. We compared demographic factors, comorbidities, treatment modalities, duration of hospitalization, financial expenditures, and mortality between TH and NTH. We also examined the patients' characteristics and outcomes in TH and NTH based on their chosen therapeutic strategy. RESULTS We identified 11,380 admissions with metastatic prostate cancer and SCC; 9610 in TH and 1770 in NTH. The median cost of hospitalization was $21,922 in TH and $15,141 in NTH. Although the median age and Charlson comorbidity score did not differ between two groups, patients in TH were more likely to receive intervention (radiation or surgery) compared to NTH (Surgery: 28.2% in TH vs. 23.0% in NTH & Radiation: 12.1% in TH vs. 8.2% in NTH). Mortality was lower in TH than NTH (4.5% vs. 7.9%). In both TH and NTH, a higher proportion of patients with private insurance underwent surgery (TH: Surgery 25.1% vs. Radiation 18.8% & NTH: Surgery 27.0% vs. 6.9%). Black patients were more likely to receive radiation than surgery in TH (34.2% vs. 26.8%). CONCLUSION This study showed a greater percentage of patients underwent surgical intervention at TH compared to NTH. Additionally, the type of insurance and racial background were associated with distinctive treatment approaches.
Collapse
Affiliation(s)
- Omid Yazdanpanah
- Division of Hematology and Oncology, Internal Medicine DepartmentUniversity of California IrvineOrangeCaliforniaUSA
| | - Aditya Mahadevan
- Division of Hematology and Oncology, Internal Medicine DepartmentUniversity of California IrvineOrangeCaliforniaUSA
| | - Aditi Sharma
- Division of Hematology/OncologyWayne State University School of Medicine/Karmanos Cancer InstituteDetroitMichiganUSA
| | | | - Arash Rezazadeh Kalebasty
- Division of Hematology and Oncology, Internal Medicine DepartmentUniversity of California IrvineOrangeCaliforniaUSA
| |
Collapse
|
3
|
Basso Dias A, Mirshahvalad SA, Ortega C, Perlis N, Berlin A, van der Kwast T, Ghai S, Jhaveri K, Metser U, Haider M, Avery L, Veit-Haibach P. The role of [ 18F]-DCFPyL PET/MRI radiomics for pathological grade group prediction in prostate cancer. Eur J Nucl Med Mol Imaging 2023; 50:2167-2176. [PMID: 36809425 DOI: 10.1007/s00259-023-06136-0] [Citation(s) in RCA: 6] [Impact Index Per Article: 6.0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/23/2022] [Accepted: 02/07/2023] [Indexed: 02/23/2023]
Abstract
PURPOSE To evaluate the diagnostic accuracy of [18F]-DCFPyL PET/MRI radiomics for the prediction of pathological grade group in prostate cancer (PCa) in therapy-naïve patients. METHODS Patients with confirmed or suspected PCa, who underwent [18F]-DCFPyL PET/MRI (n = 105), were included in this retrospective analysis of two prospective clinical trials. Radiomic features were extracted from the segmented volumes following the image biomarker standardization initiative (IBSI) guidelines. Histopathology obtained from systematic and targeted biopsies of the PET/MRI-detected lesions was the reference standard. Histopathology patterns were dichotomized as ISUP GG 1-2 vs. ISUP GG ≥ 3 categories. Different single-modality models were defined for feature extraction, including PET- and MRI-derived radiomic features. The clinical model included age, PSA, and lesions' PROMISE classification. Single models, as well as different combinations of them, were generated to calculate their performances. A cross-validation approach was used to evaluate the internal validity of the models. RESULTS All radiomic models outperformed the clinical models. The best model for grade group prediction was the combination of PET + ADC + T2w radiomic features, showing sensitivity, specificity, accuracy, and AUC of 0.85, 0.83, 0.84, and 0.85, respectively. The MRI-derived (ADC + T2w) features showed sensitivity, specificity, accuracy, and AUC of 0.88, 0.78, 0.83, and 0.84, respectively. PET-derived features showed 0.83, 0.68, 0.76, and 0.79, respectively. The baseline clinical model showed 0.73, 0.44, 0.60, and 0.58, respectively. The addition of the clinical model to the best radiomic model did not improve the diagnostic performance. The performances of MRI and PET/MRI radiomic models as per the cross-validation scheme yielded an accuracy of 0.80 (AUC = 0.79), whereas clinical models presented an accuracy of 0.60 (AUC = 0.60). CONCLUSION The combined [18F]-DCFPyL PET/MRI radiomic model was the best-performing model and outperformed the clinical model for pathological grade group prediction, indicating a complementary value of the hybrid PET/MRI model for non-invasive risk stratification of PCa. Further prospective studies are required to confirm the reproducibility and clinical utility of this approach.
Collapse
Affiliation(s)
- Adriano Basso Dias
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada.
| | - Seyed Ali Mirshahvalad
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Claudia Ortega
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Nathan Perlis
- Division of Urology, Department of Surgery, Princess Margaret Cancer Centre, University Health Network, Toronto, ON, Canada
| | - Alejandro Berlin
- Department of Radiation Oncology, Princess Margaret Cancer Center, University Health Network & University of Toronto, Toronto, ON, Canada
| | | | - Sangeet Ghai
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Kartik Jhaveri
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Ur Metser
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Masoom Haider
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| | - Lisa Avery
- Department of Biostatistics, Princess Margaret Cancer Centre, Toronto, ON, Canada
| | - Patrick Veit-Haibach
- Joint Department of Medical Imaging, University Medical Imaging Toronto (UMIT), University Health Network, Mount Sinai Hospital & Women's College Hospital; University of Toronto, Toronto, ON, Canada
| |
Collapse
|
4
|
Robustness of Radiomics in Pre-Surgical Computer Tomography of Non-Small-Cell Lung Cancer. J Pers Med 2022; 13:jpm13010083. [PMID: 36675744 PMCID: PMC9864775 DOI: 10.3390/jpm13010083] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/22/2022] [Accepted: 12/20/2022] [Indexed: 12/31/2022] Open
Abstract
Background: Radiomic features are increasingly used in CT of NSCLC. However, their robustness with respect to segmentation variability has not yet been demonstrated. The aim of this study was to assess radiomic features agreement across three kinds of segmentation. Methods: We retrospectively included 48 patients suffering from NSCLC who underwent pre-surgery CT. Two expert radiologists in consensus manually delineated three 3D-ROIs on each patient. To assess robustness for each feature, the intra-class correlation coefficient (ICC) across segmentations was evaluated. The ‘sensitivity’ of ICC upon some parameters affecting features computation (such as bin-width for first-order features and pixel-distances for second-order features) was also evaluated. Moreover, an assessment with respect to interpolator and isotropic resolution was also performed. Results: Our results indicate that ‘shape’ features tend to have excellent agreement (ICC > 0.9) across segmentations; moreover, they have approximately zero sensitivity to other parameters. ‘First-order’ features are in general sensitive to parameters variation; however, a few of them showed excellent agreement and low sensitivity (below 0.1) with respect to bin-width and pixel-distance. Similarly, a few second-order features showed excellent agreement and low sensitivity. Conclusions: Our results suggest that a limited number of radiomic features can achieve a high level of reproducibility in CT of NSCLC.
Collapse
|
5
|
Li L, Gu L, Kang B, Yang J, Wu Y, Liu H, Lai S, Wu X, Jiang J. Evaluation of the Efficiency of MRI-Based Radiomics Classifiers in the Diagnosis of Prostate Lesions. Front Oncol 2022; 12:934108. [PMID: 35865467 PMCID: PMC9295912 DOI: 10.3389/fonc.2022.934108] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/02/2022] [Accepted: 06/07/2022] [Indexed: 12/24/2022] Open
Abstract
ObjectiveTo compare the performance of different imaging classifiers in the prospective diagnosis of prostate diseases based on multiparameter MRI.MethodsA total of 238 patients with pathological outcomes were enrolled from September 2019 to July 2021, including 142 in the training set and 96 in the test set. After the regions of interest were manually segmented, decision tree (DT), Gaussian naive Bayes (GNB), XGBoost, logistic regression, random forest (RF) and support vector machine classifier (SVC) models were established on the training set and tested on the independent test set. The prospective diagnostic performance of each classifier was compared by using the AUC, F1-score and Brier score.ResultsIn the patient-based data set, the top three classifiers of combined sequences in terms of the AUC were logistic regression (0.865), RF (0.862), and DT (0.852); RF “was significantly different from the other two classifiers (P =0.022, P =0.005), while logistic regression and DT had no statistical significance (P =0.802). In the lesions-based data set, the top three classifiers of combined sequences in terms of the AUC were RF (0.931), logistic regression (0.922) and GNB (0.922). These three classifiers were significantly different from.ConclusionThe results of this experiment show that radiomics has a high diagnostic efficiency for prostate lesions. The RF classifier generally performed better overall than the other classifiers in the experiment. The XGBoost and logistic regression models also had high classification value in the lesions-based data set.
Collapse
Affiliation(s)
- Linghao Li
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Lili Gu
- Department of Pain, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Bin Kang
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Jiaojiao Yang
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Ying Wu
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Hao Liu
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Shasha Lai
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Xueting Wu
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
| | - Jian Jiang
- Department of Radiology, the First Affiliated Hospital, Nanchang University, Nanchang, China
- *Correspondence: Jian Jiang,
| |
Collapse
|