1
|
Dudas D, Dilling TJ, El Naqa I. Improved outcome models with denoising diffusion. Phys Med 2024; 119:103307. [PMID: 38325221 PMCID: PMC10939775 DOI: 10.1016/j.ejmp.2024.103307] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 11/09/2023] [Revised: 01/12/2024] [Accepted: 01/31/2024] [Indexed: 02/09/2024] Open
Abstract
PURPOSE Radiotherapy outcome modelling often suffers from class imbalance in the modelled endpoints. One of the main options to address this issue is by introducing new synthetically generated datapoints, using generative models, such as Denoising Diffusion Probabilistic Models (DDPM). In this study, we implemented DDPM to improve performance of a tumor local control model, trained on imbalanced dataset, and compare this approach with other common techniques. METHODS A dataset of 535 NSCLC patients treated with SBRT (50 Gy/5 fractions) was used to train a deep learning outcome model for tumor local control prediction. The dataset included complete treatment planning data (planning CT images, 3D planning dose distribution and patient demographics) with sparsely distributed endpoints (6-7 % experiencing local failure). Consequently, we trained a novel conditional 3D DDPM model to generate synthetic treatment planning data. Synthetically generated treatment planning datapoints were used to supplement the real training dataset and the improvement in the model's performance was studied. Obtained results were also compared to other common techniques for class imbalanced training, such as Oversampling, Undersampling, Augmentation, Class Weights, SMOTE and ADASYN. RESULTS Synthetic DDPM-generated data were visually trustworthy, with Fréchet inception distance (FID) below 50. Extending the training dataset with the synthetic data improved the model's performance by more than 10%, while other techniques exhibited only about 4% improvement. CONCLUSIONS DDPM introduces a novel approach to class-imbalanced outcome modelling problems. The model generates realistic synthetic radiotherapy planning data, with a strong potential to increase performance and robustness of outcome models.
Collapse
Affiliation(s)
- D Dudas
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL, USA; Czech Technical University in Prague, Faculty of Nuclear Sciences and Physical Engineering, Prague, Czechia.
| | - T J Dilling
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL, USA
| | - I El Naqa
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL, USA
| |
Collapse
|
2
|
Dudas D, Ghasemi P, Dilling TJ, Perez BA, Rosenberg SA, El Naqa I. Novel Dose Criteria for Lung Cancer SBRT to Improve Local Control in Patients Treated to 50 Gy/5 Fractions Using Deep Learning Methods and Explainability Techniques. Int J Radiat Oncol Biol Phys 2023; 117:e662. [PMID: 37785961 DOI: 10.1016/j.ijrobp.2023.06.2099] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/04/2023]
Abstract
PURPOSE/OBJECTIVE(S) To some radiation oncologists, 50 Gy/5 fractions has been considered controversial, as they feel the nominal BED of 100 Gy might be too low for long-term local control of some lesions. We analyzed a large cohort of these patients using a deep learning model to predict local recurrence (LR) and used explainability techniques to extract new dose features important to the model's prediction. Subsequently, we determined optimal cut-points for the most significant metrics to provide actionable criteria for treatment planning in these patients. MATERIALS/METHODS A total of 535 SBRT lung cancer patients treated between 2009 and 2017 were retrospectively analyzed using a deep learning approach. All patients had NSCLC and all of them were treated with 50 Gy in 5 fractions (100 Gy BED, α/β = 10). Mean clinical maximum tumor diameter was 2.2 cm. There were 31 LR in the dataset with mean follow-up time of 28 months. Mean age was 75 years. CT images, 3D dose distribution and patient demographic details were used to train a deep learning survival model to predict time to failure and probability of local control. Validation, training, and testing were in accordance with TRIPOD criteria. 80 % of the data were used for 5-fold cross-validation (10 iterations) and 20 % was held for independent testing. The Grad-CAM method was applied to identify regions of the dose distribution that are the most significant to the model's decision-making. Based on the results, appropriate dose metrics were proposed, and optimal cut-points were determined to distinguish between lower and higher LR-risk patients. RESULTS The model has an acceptable performance (c-index: 0.72, 95% CI: 0.68-0.75); the testing c-index was 0.69. Grad-CAM showed that the model's spatial attention was mostly concentrated in the tumor's "PTV-GTV" region. Statistically significant criteria are in Table 1. CONCLUSION A novel deep learning model for prediction of LR, incorporating 3D dose data, CT images and patient demographics, was developed and tested. Grad-CAM demonstrated superior significance of peripheral (PTV-GTV) dose features. Subsequently determined optimal cut-points have significant prognostic power (log rank, p<0.001) and could be used as additional criteria in treatment planning. While these data have repercussions in treatment planning, they do not suggest that a significantly higher BED for the prescription dose is necessary for tumor control in NSCLC. Nevertheless, it might be effective to slightly elevate the prescribed dose, i.e., from 100 Gy BED to 104 Gy BED.
Collapse
Affiliation(s)
- D Dudas
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL
| | - P Ghasemi
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL
| | - T J Dilling
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - B A Perez
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - S A Rosenberg
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - I El Naqa
- H. Lee Moffitt Cancer Center and Research Institute, Department of Machine Learning, Tampa, FL
| |
Collapse
|
3
|
Sandoval ML, Rishi A, Liveringhouse C, Dohm AE, Palm RF, Perez BA, Frakes JM, Rosenberg SA, Hoffe S, Dilling TJ. Outcomes of Cytoreductive Stereotactic Body Radiotherapy (SBRT) in Patients with Oligometastatic or Oligoprogressive Dominant Lung Metastases from Colorectal Primary. Int J Radiat Oncol Biol Phys 2023; 117:e53. [PMID: 37785644 DOI: 10.1016/j.ijrobp.2023.06.764] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/04/2023]
Abstract
PURPOSE/OBJECTIVE(S) Oxaliplatin based systemic therapy regimens have improved the prognosis of patients with colorectal cancer (CRC) and with this, there has been increased interest in the integration of local therapies to oligometastatic and oligoprogressive sites. There is a vast body of literature exploring the benefits of cytoreduction with surgery and stereotactic body radiation therapy (SBRT) approaches. We report our rates of local control (LC) and overall survival (OS) for patients with oligometastatic/progressive CRC with lung metastases treated with SBRT. MATERIALS/METHODS Single institution retrospective review of patients diagnosed with oligometastatic or oligoprogressive CRC with dominant metastases to the lungs who were treated with SBRT between September 2009 and December 2022. Oligometastatic disease was defined as newly diagnosed, untreated CRC with up to 5 metastases, up to 3 in one organ. Oligoprogressive disease was defined as CRC with 1 - 2 distant sites that continued to progress on active treatment while the primary site was controlled. Survival was estimated using Kaplan-Meier. Association between local control and patient factors was analyzed using log-rank test. RESULTS A total of 84 patients with oligometastatic or oligoprogressive CRC were treated with SBRT to 124 lung lesions. Colon cancer was the primary site for 54 patients with a median age at time of SBRT of 66 years (IQR 57 - 73) and a median tumor diameter of 1.20 cm (IQR 0.93 - 1.90). Rectal cancer was the primary site for 30 patients, median age was 60 years (IQR 49 - 70) and median tumor diameter was 1.10 cm (IQR 0.80 - 1.48). Median dose for the entire cohort was 6000 cGy (range 5000 - 6000) with median number of fractions 5 (range 3 - 5). Median follow-up after SBRT was 24 months. Overall, there were 9 local failures at last follow-up. Almost half (n = 42) of the patients experienced distant recurrence. Median local control (LC) for the entire cohort was not reached, 2-yr LC and 5-yr LC were 94.6% and 85.7% respectively. There were no differences in LC between colon and rectal cancer (p = 0.29). Actuarial median overall survival was 71 months (95% CI 44.3 - 97.7) and 5-yr OS was 50.2%. Due to the small number of events, we were unable to identify patient factors associated with local failure on univariate or multivariate analysis. CONCLUSION Cytoreductive SBRT is an effective treatment option for patients with oligometastatic or oligoprogressive CRC with dominant lung metastases offering excellent rates of LC. Most patients failed distantly highlighting the importance of additional systemic therapies.
Collapse
Affiliation(s)
- M L Sandoval
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - A Rishi
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - C Liveringhouse
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - A E Dohm
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - R F Palm
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - B A Perez
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - J M Frakes
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - S A Rosenberg
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - S Hoffe
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| | - T J Dilling
- H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL
| |
Collapse
|
4
|
Saeed N, Latifi K, Hoffe SE, Cruz A, Opp DW, Moros EG, Zhang GG, Budzevich MM, Shridhar R, Dilling TJ. Optimizing Options for Re-irradiation With Deformable Image Registration of Prior Plans. Pract Radiat Oncol 2014; 3:S16-7. [PMID: 24674497 DOI: 10.1016/j.prro.2013.01.060] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.2] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/16/2022]
Affiliation(s)
- N Saeed
- Brown University, Providence, RI
| | - K Latifi
- Moffitt Cancer Center, Tampa, FL
| | | | - A Cruz
- University of South Florida, Tampa, FL
| | - D W Opp
- Moffitt Cancer Center, Tampa, FL
| | | | | | | | | | | |
Collapse
|
5
|
Figura N, Latifi K, Dilling TJ, Kuykendall CC, Eikman EA, Moros EG, Zhang GG, Leuthold S, Mehra C, Hoffe SE. Dosimetric Implications of Treating 4D PET/CT-Defined Maximum Inhale Versus Exhale Target Volumes in Esophageal Cancer. Pract Radiat Oncol 2013; 3:S34-5. [PMID: 24674556 DOI: 10.1016/j.prro.2013.01.116] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.1] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/27/2022]
Affiliation(s)
- N Figura
- University of South Florida, Tampa, FL
| | - K Latifi
- Moffitt Cancer Center, Tampa, FL
| | | | | | | | | | | | | | - C Mehra
- Moffitt Cancer Center, Tampa, FL
| | | |
Collapse
|
6
|
Lomas H, Hoffe SE, Weber J, Dilling TJ. Post Chemoradiation PET SUV is highly Predictive of Overall Survival in Esophageal Cancer. ACTA ACUST UNITED AC 2012. [DOI: 10.4172/2155-9619.1000125] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [What about the content of this article? (0)] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/09/2022]
|
7
|
Benjamin I, Dilling TJ, Goldwein JW. Administration of a World Wide Web site during a period of rapid growth: the OncoLink experience. MD Comput 1997; 14:365-70. [PMID: 9308345] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Subscribe] [Scholar Register] [Indexed: 02/05/2023]
Abstract
OncoLink is a cancer information resource on the World Wide Web that provides a wide variety of information for both patients and health care providers. Introduced in March 1994, OncoLink has enjoyed a 50-fold increase in use since then, with more than 1.8 million accesses per month as of February 1997. New items are added daily, and the OncoLink Web site currently contains more than 10,000 files. During this period of rapid growth, the complexity of managing and maintaining OncoLink has increased as well. Consequently, we developed administrative procedures to handle our workload, which involves content editing, technical (or production) editing, and Web site maintenance. The new strategies have greatly reduced the need for face-to-face meetings of our editorial and production staffs. The rapid growth of OncoLink would not have been possible without these efficient new strategies for managing its daily operation.
Collapse
Affiliation(s)
- I Benjamin
- Department of Obstetrics and Gynecology, School of Medicine, University of Pennsylvania Cancer Center, Philadelphia 19104, USA
| | | | | |
Collapse
|
8
|
Emrich JG, Hand CM, Dilling TJ, Class R, Bender H, Brady LW. Biodistribution of 125I-MAb 425 in a human glioma xenograft model: effect of chloroquine. Hybridoma (Larchmt) 1997; 16:93-100. [PMID: 9085135 DOI: 10.1089/hyb.1997.16.93] [Citation(s) in RCA: 7] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 02/04/2023]
Abstract
Chloroquine has been shown to increase the cellular retention and nuclear incorporation of 125I-labeled monoclonal antibody (MAb) 425, a murine anti-epidermal growth factor receptor monoclonal antibody, in human high-grade glioma cells in vitro. The objective of this study was to examine the effect of chloroquine on the biodistribution of 125I-MAb 425 in an intracerebral xenogeneic transplant of glioma cells. Nude rats were stereotaxically implanted in the right hemisphere with A1207 human high-grade glioma cells. After 14 days, animals were injected i.v. with chloroquine (40 mg/kg) followed 2 h later by an 125I-MAb 425 (9 MBq) infusion. Tissue distributions were performed up to 168 h post 125I-MAb 425 injection. From 24 to 168 h, tumor-to-contralateral left brain ratios increased from 9 to 15 for 125I-MAb 425 alone, and 7 to 13 for the 125I-MAb 425/chloroquine combination, respectively. A single administration of chloroquine did not result in any significant difference in radiolabeled MAb accumulation in either the tumor site or other tissues. We conclude that chloroquine did not increase the amount of 125I-MAb 425 into the tumor; however, it is safe to administer i.v. at the 40 mg/kg dose. Under these experimental conditions, the increased radioactive accumulation observed for in vitro data did not translate into similar in vivo results.
Collapse
Affiliation(s)
- J G Emrich
- Department of Radiation Oncology, Allegheny University of the Health Sciences, Philadelphia, Pennsylvania 19102, USA
| | | | | | | | | | | |
Collapse
|
9
|
Benjamin I, Dilling TJ, Campbell KC, Maraqa A, Liang B, Medbery R, Goldwein JW. Technical and editorial administration of a World-Wide-Web site during a period of rapid growth: the OncoLink experience. Proc AMIA Annu Fall Symp 1996:398-402. [PMID: 8947696 PMCID: PMC2233173] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Subscribe] [Scholar Register] [Indexed: 02/03/2023]
Abstract
OncoLink is a cancer information resource on the World-Wide-Web (WWW) that provides a wide variety of information for cancer patients and healthcare providers. Since its introduction in March, 1994 it has enjoyed success as demonstrated by an over 31-fold increase in usage as of February, 1996. Current utilization exceeds 1.1 million accesses per month. The content of OncoLink has also expanded greatly, with new items being added daily. In addition, OncoLink has been the recipient of numerous awards from a variety of agencies and organizations. During this period of rapid growth, the complexity of managing and maintaining OncoLink has likewise increased. This work may be divided into three categories: content editing, technical (or production) editing, and web site maintenance. Consequently, we have developed numerous administrative procedures to handle this workload. After implementing these new administrative strategies, we were able to greatly reduce the need for face-to-face meetings of our Editorial and Production Staffs. This paper describes our experience with developing efficient strategies for managing the daily operation of OncoLink during a period of rapid growth.
Collapse
Affiliation(s)
- I Benjamin
- Department of Obstetrics & Gynecology, School of Medicine University of Pennsylvania Cancer Center, Philadelphia, USA
| | | | | | | | | | | | | |
Collapse
|