How Will Big Data Impact Clinical Decision Making and Precision Medicine in Radiation Therapy?

Finite Element-Based Personalized Simulation of Duodenal Hydrogel Spacer: Spacer Location Dependent Duodenal Sparing and a Decision Support System for Spacer-Enabled Pancreatic Cancer Radiation Therapy

Purpose

Pancreatic cancer is the fourth leading cause of cancer-related death, with a very low 5-year overall survival rate (OS). Radiation therapy (RT) together with dose escalation significantly increases the OS at 2 and 3 years. However, dose escalation is very limited due to the proximity of the duodenum. Hydrogel spacers are an effective way to reduce duodenal toxicity, but the complexity of the anatomy and the procedure makes the success and effectiveness of the spacer procedure highly uncertain. To provide a preoperative simulation of hydrogel spacers, we presented a patient-specific spacer simulator algorithm and used it to create a decision support system (DSS) to provide a preoperative optimal spacer location to maximize the spacer benefits.

Materials and Methods

Our study was divided into three phases. In the validation phase, we evaluated the patient-specific spacer simulator algorithm (FEMOSSA) for the duodenal spacer using the dice similarity coefficient (DSC), overlap volume histogram (OVH), and radial nearest neighbor distance (RNND). For the simulation phase, we simulated four virtual spacer scenarios based on the location of the spacer in para-duodenal space. Next, stereotactic body radiation therapy (SBRT) plans were designed and dosimetrically analyzed. Finally, in the prediction phase, using the result of the simulation phase, we created a Bayesian DSS to predict the optimal spacer location and biological effective dose (BED).

Results

A realistic simulation of the spacer was achieved, reflected in a statistically significant increase in average target and duodenal DSC for the simulated spacer. Moreover, the small difference in average mean and 5th-percentile RNNDs (0.5 and 2.1 mm) and OVH thresholds (average of less than 0.75 mm) showed that the simulation attained similar separation as the real spacer. We found a spacer-location-independent decrease in duodenal V20Gy, a highly spacer-location-dependent change in V33Gy, and a strong correlation between L1cc and V33Gy. Finally, the Bayesian DSS predicted the change in BED with a root mean squared error of 3.6 Gys.

Conclusions

A duodenal spacer simulator platform was developed and used to systematically study the dosimetric effect of spacer location. Further, L1cc is an informative anatomical feedback to guide the DSS to indicate the spacer efficacy, optimum location, and expected improvement.

Rapid Progress in Intelligent Radiotherapy and Future Implementation

Radiotherapy is one of the major approaches to cancer treatment. Artificial intelligence in radiotherapy (shortly, Intelligent radiotherapy) mainly involves big data, deep learning, extended reality, digital twin, radiomics, Internet plus and Internet of Things (IoT), which establish an automatic and intelligent network platform consisting of radiotherapy preparation, target volume delineation, treatment planning, radiation delivery, quality assurance (QA) and quality control (QC), prognosis judgment and post-treatment follow-up. Intelligent radiotherapy is an interdisciplinary frontier discipline in infancy. The review aims to summary the important implements of intelligent radiotherapy in various areas and put forward the future of unmanned radiotherapy center.

Medical physics external beam plan review: What contributes to the variability?

PURPOSE

In this article we report on the results of a survey of physics plan review practices conducted by the Cancer Care Ontario Communities of Practice and the variations in practice between and within centers.

METHODS

The medical physicists at each center worked together to complete the survey and submit a single response for that center. A 4-point Likert scale, used to report the variation in practice at each center, was quantified into two parameters: "Intra-center variation", the distribution of responses within the center, and "Variation between centers", the difference between the center's response and the provincial mean. These metrics were correlated with center characteristics to identify factors that impacted on variations in practice.

RESULTS

Bolus and heterogeneity correction were the only two items checked by all physicists in all centers. In more than half of the centers, image registration and DVH binning are not likely checked by physics. A significant difference in the variation between centers is observed for centers that used a single vendor's products. Centers that used an official checklist indicated higher levels and a wider range of Intra-center variation. Higher workload did not affect the variation in checking patterns between physicists in the same center.

CONCLUSIONS

The effect of a center's resources on their checking practice suggest that local environment and workflow be accounted for when implementing TG275 guidelines. The observation that standardized checklists did not reduce checking variability point to the importance of following the checklist development guidelines in MPPG4 to avoid ineffective checklists.

Clinical Natural Language Processing for Radiation Oncology: A Review and Practical Primer

Natural language processing (NLP), which aims to convert human language into expressions that can be analyzed by computers, is one of the most rapidly developing and widely used technologies in the field of artificial intelligence. Natural language processing algorithms convert unstructured free text data into structured data that can be extracted and analyzed at scale. In medicine, this unlocking of the rich, expressive data within clinical free text in electronic medical records will help untap the full potential of big data for research and clinical purposes. Recent major NLP algorithmic advances have significantly improved the performance of these algorithms, leading to a surge in academic and industry interest in developing tools to automate information extraction and phenotyping from clinical texts. Thus, these technologies are poised to transform medical research and alter clinical practices in the future. Radiation oncology stands to benefit from NLP algorithms if they are appropriately developed and deployed, as they may enable advances such as automated inclusion of radiation therapy details into cancer registries, discovery of novel insights about cancer care, and improved patient data curation and presentation at the point of care. However, challenges remain before the full value of NLP is realized, such as the plethora of jargon specific to radiation oncology, nonstandard nomenclature, a lack of publicly available labeled data for model development, and interoperability limitations between radiation oncology data silos. Successful development and implementation of high quality and high value NLP models for radiation oncology will require close collaboration between computer scientists and the radiation oncology community. Here, we present a primer on artificial intelligence algorithms in general and NLP algorithms in particular; provide guidance on how to assess the performance of such algorithms; review prior research on NLP algorithms for oncology; and describe future avenues for NLP in radiation oncology research and clinics.

FEMOSSA: Patient-specific finite element simulation of the prostate-rectum spacer placement, a predictive model for prostate cancer radiotherapy

PURPOSE

Major advances in delivery systems in recent years have turned radiotherapy (RT) into a more effective way to manage prostate cancer. Still, adjacency of organs at risk (OARs) can severely limit RT benefits. Rectal spacer implant in recto-prostatic space provides sufficient separation between prostate and rectum, and therefore, the opportunity for potential dose escalation to the target and reduction of OAR dose. Pretreatment simulation of spacer placement can potentially provide decision support to reduce the risks and increase the efficacy of the spacer placement procedure.

METHODS

A novel finite element method-oriented spacer simulation algorithm, FEMOSSA, was developed in this study. We used the finite element (FE) method to model and predict the deformation of rectum and prostate wall, stemming from hydrogel injection. Ten cases of prostate cancer, which undergone hydrogel placement before the RT treatment, were included in this study. We used the pre-injection organ contours to create the FE model and post-injection spacer location to estimate the distribution of the virtual spacer. Material properties and boundary conditions specific to each patient's anatomy were assigned. The FE analysis was then performed to determine the displacement vectors of regions of interest (ROIs), and the results were validated by comparing the virtually simulated contours with the real post-injection contours. To evaluate the different aspects of our method's performance, we used three different figures of merit: dice similarity coefficient (DSC), nearest neighbor distance (NND), and overlapped volume histogram (OVH). Finally, to demonstrate a potential dosimetric application of FEMOSSA, the predicted rectal dose after virtual spacer placement was compared against the predicted post-injection rectal dose.

RESULTS

Our simulation showed a realistic deformation of ROIs. The post-simulation (virtual spacer) created the same separation between prostate and rectal wall, as post-injection spacer. The average DSCs for prostate and rectum were 0.87 and 0.74, respectively. Moreover, there was a statistically significant increase in rectal contour similarity coefficient (P < 0.01). Histogram of NNDs showed the same overall shape and a noticeable shift from lower to higher values for both post-simulation and post-injection, indicative of the increase in distance between prostate and rectum. There was less than 2.2- and 2.1-mm averaged difference between the mean and fifth percentile NNDs. The difference between the OVH distances and the corresponding predicted rectal dose was, on average, less than 1 mm and 1.5 Gy, respectively.

CONCLUSIONS

FEMOSSA provides a realistic simulation of the hydrogel injection process that can facilitate spacer placement planning and reduce the associated uncertainties. Consequently, it increases the robustness and success rate of spacer placement procedure that in turn improves prostate cancer RT quality.

Use of metrics to quantify IMRT and VMAT treatment plan complexity: A systematic review and perspectives

PURPOSE

Fixed-field intensity modulated radiation therapy (FF-IMRT) or volumetric modulated arc therapy (VMAT) beams complexity is due to fluence fluctuation. Pre-treatment Quality Assurance (PTQA) failure could be linked to it. Several plan complexity metrics (PCM) have been published to quantify this complexity but in a heterogeneous formalism. This review proposes to gather different PCM and to discuss their eventual PTQA failure identifier abilities.

METHODS AND MATERIALS

A systematic literature search and outcome extraction from MEDLINE/PubMed (National Center for Biotechnology Information, NCBI) was performed. First, a list and a synthesis of available PCM is made in a homogeneous formalism. Second, main results relying on the link between PCM and PTQA results but also on other uses are listed.

RESULTS

A total of 163 studies were identified and n = 19 were selected after inclusion and exclusion criteria application. Difference is made between fluence and degree of freedom (DOF)-based PCM. Results about the PCM potential as PTQA failure identifier are described and synthesized. Others uses are also found in quality, big data, machine learning and audit procedure.

CONCLUSIONS

A state of the art is made thanks to this homogeneous PCM classification. For now, PCM should be seen as a planning procedure quality indicator although PTQA failure identifier results are mitigated. However limited clinical use seems possible for some cases. Yet, addressing the general PTQA failure prediction case could be possible with the big data or machine learning help.

Retrospective analysis of breast radiotherapy treatment plans: Curating the 'non-curated'

INTRODUCTION

This paper provides a demonstration of how non-curated data can be retrospectively cleaned, so that existing repositories of radiotherapy treatment planning data can be used to complete bulk retrospective analyses of dosimetric trends and other plan characteristics.

METHODS

A non curated archive of 1137 radiotherapy treatment plans accumulated over a 12-month period, from five radiotherapy centres operated by one institution, was used to investigate and demonstrate a process of clinical data cleansing, by: identifying and translating inconsistent structure names; correcting inconsistent lung contouring; excluding plans for treatments other than breast tangents and plans without identifiable PTV, lung and heart structures; and identifying but not excluding plans that deviated from the local planning protocol. PTV, heart and lung dose-volume metrics were evaluated, in addition to a sample of personnel and linac load indicators.

RESULTS

Data cleansing reduced the number of treatment plans in the sample by 35.7%. Inconsistent structure names were successfully identified and translated (e.g. 35 different names for lung). Automatically separating whole lung structures into left and right lung structures allowed the effect of contralateral and ipsilateral lung dose to be evaluated, while introducing some small uncertainties, compared to manual contouring. PTV doses were indicative of prescription doses. Breast treatment work was unevenly distributed between oncologists and between metropolitan and regional centres.

CONCLUSION

This paper exemplifies the data cleansing and data analysis steps that may be completed using existing treatment planning data, to provide individual radiation oncology departments with access to information on their own patient populations. Clearly, the well-planned and systematic recording of new, high quality data is the preferred solution, but the retrospective curation of non-curated data may be a useful interim solution, for radiation oncology departments where the systems for recording of new data have yet to be designed and agreed.

National Cancer Institute Workshop on Proton Therapy for Children: Considerations Regarding Brainstem Injury

PURPOSE

Proton therapy can allow for superior avoidance of normal tissues. A widespread consensus has been reached that proton therapy should be used for patients with curable pediatric brain tumor to avoid critical central nervous system structures. Brainstem necrosis is a potentially devastating, but rare, complication of radiation. Recent reports of brainstem necrosis after proton therapy have raised concerns over the potential biological differences among radiation modalities. We have summarized findings from the National Cancer Institute Workshop on Proton Therapy for Children convened in May 2016 to examine brainstem injury.

METHODS AND MATERIALS

Twenty-seven physicians, physicists, and researchers from 17 institutions with expertise met to discuss this issue. The definition of brainstem injury, imaging of this entity, clinical experience with photons and photons, and potential biological differences among these radiation modalities were thoroughly discussed and reviewed. The 3 largest US pediatric proton therapy centers collectively summarized the incidence of symptomatic brainstem injury and physics details (planning, dosimetry, delivery) for 671 children with focal posterior fossa tumors treated with protons from 2006 to 2016.

RESULTS

The average rate of symptomatic brainstem toxicity from the 3 largest US pediatric proton centers was 2.38%. The actuarial rate of grade ≥2 brainstem toxicity was successfully reduced from 12.7% to 0% at 1 center after adopting modified radiation guidelines. Guidelines for treatment planning and current consensus brainstem constraints for proton therapy are presented. The current knowledge regarding linear energy transfer (LET) and its relationship to relative biological effectiveness (RBE) are defined. We review the current state of LET-based planning.

CONCLUSIONS

Brainstem injury is a rare complication of radiation therapy for both photons and protons. Substantial dosimetric data have been collected for brainstem injury after proton therapy, and established guidelines to allow for safe delivery of proton radiation have been defined. Increased capability exists to incorporate LET optimization; however, further research is needed to fully explore the capabilities of LET- and RBE-based planning.

Clinical Outcomes of Several IMRT Techniques for Patients With Head and Neck Cancer: A Propensity Score-Weighted Analysis

PURPOSE

The Advanced Radiotherapy Oto-Rhino-Laryngologie (ART-ORL) study (NCT02024035) was performed to prospectively evaluate the clinical and economic aspects of helical TomoTherapy and volumetric modulated arc therapy (RapidArc, Varian Medical Systems, Palo Alto, CA) for patients with head and neck cancer.

METHODS AND MATERIALS

Fourteen centers participated in this prospective comparative study. Randomization was not possible based on the availability of equipment. Patients with epidermoid or undifferentiated nasopharyngeal carcinoma or epidermoid carcinoma of the oropharynx and oral cavity (T1-T4, M0, N0-N3) were included between February 2010 and February 2012. Only the results of the clinical study are presented in this report, as the results of the economic assessment have been published previously. Inverse probability of treatment weighting using the propensity score analysis was undertaken in an effort to adjust for potential bias due to nonrandomization. Locoregional control, cancer-specific survival, and overall survival assessed 18 months after treatment, as well as long-term toxicity and salivary function, were evaluated.

RESULTS

The analysis included 166 patients. The following results are given after inverse probability of treatment weighting adjustment. The locoregional control rate at 18 months was significantly better in the TomoTherapy group: 83.3% (95% confidence interval [CI], 72.5%-90.2%) versus 72.7% (95% CI, 62.1%-80.8%) in the RapidArc group (P=.025). The cancer-specific survival rate was better in the TomoTherapy group: 97.2% (95% CI, 89.3%-99.3%) versus 85.5% (95% CI, 75.8%-91.5%) in the RapidArc group (P=.014). No significant difference was shown in progression-free or overall survival. TomoTherapy induced fewer acute salivary disorders (P=.012). Posttreatment salivary function degradation was worse in the RapidArc group (P=.012).

CONCLUSIONS

TomoTherapy provided better locoregional control and cancer-specific survival than RapidArc treatment, with fewer salivary disorders. No significant difference was shown in progression-free and overall survival. These results should be explored in a randomized trial.

[Big data and their perspectives in radiation therapy]

The concept of big data indicates a change of scale in the use of data and data aggregation into large databases through improved computer technology. One of the current challenges in the creation of big data in the context of radiation therapy is the transformation of routine care items into dark data, i.e. data not yet collected, and the fusion of databases collecting different types of information (dose-volume histograms and toxicity data for example). Processes and infrastructures devoted to big data collection should not impact negatively on the doctor-patient relationship, the general process of care or the quality of the data collected. The use of big data requires a collective effort of physicians, physicists, software manufacturers and health authorities to create, organize and exploit big data in radiotherapy and, beyond, oncology. Big data involve a new culture to build an appropriate infrastructure legally and ethically. Processes and issues are discussed in this article.

The big data effort in radiation oncology: Data mining or data farming?

Although large volumes of information are entered into our electronic health care records, radiation oncology information systems and treatment planning systems on a daily basis, the goal of extracting and using this big data has been slow to emerge. Development of strategies to meet this goal is aided by examining issues with a data farming instead of a data mining conceptualization. Using this model, a vision of key data elements, clinical process changes, technology issues and solutions, and role for professional societies is presented. With a better view of technology, process and standardization factors, definition and prioritization of efforts can be more effectively directed.