Scalable radiotherapy data curation infrastructure for deep-learning based autosegmentation of organs-at-risk: A case study in head and neck cancer

In this era of patient-centered, outcomes-driven and adaptive radiotherapy, deep learning is now being successfully applied to tackle imaging-related workflow bottlenecks such as autosegmentation and dose planning. These applications typically require supervised learning approaches enabled by relatively large, curated radiotherapy datasets which are highly reflective of the contemporary standard of care. However, little has been previously published describing technical infrastructure, recommendations, methods or standards for radiotherapy dataset curation in a holistic fashion. Our radiation oncology department has recently embarked on a large-scale project in partnership with an external partner to develop deep-learning-based tools to assist with our radiotherapy workflow, beginning with autosegmentation of organs-at-risk. This project will require thousands of carefully curated radiotherapy datasets comprising all body sites we routinely treat with radiotherapy. Given such a large project scope, we have approached the need for dataset curation rigorously, with an aim towards building infrastructure that is compatible with efficiency, automation and scalability. Focusing on our first use-case pertaining to head and neck cancer, we describe our developed infrastructure and novel methods applied to radiotherapy dataset curation, inclusive of personnel and workflow organization, dataset selection, expert organ-at-risk segmentation, quality assurance, patient de-identification, data archival and transfer. Over the course of approximately 13 months, our expert multidisciplinary team generated 490 curated head and neck radiotherapy datasets. This task required approximately 6000 human-expert hours in total (not including planning and infrastructure development time). This infrastructure continues to evolve and will support ongoing and future project efforts.

The dose magnifying glass quality assurance system for daily proton therapy range verification

Proton therapy has a distinct dosimetric advantage over conventional photon therapy due to its Bragg peak profile. This allows greater accuracy in dose delivery and dose conformation to the target, however it requires greater precision in setup, delivery and for quality assurance (QA) procedures. The AAPM TG 224 report recommends daily range and spot position checks with tolerance on the order of a millimetre. Daily QA systems must therefore be efficient for daily use and be capable of sub-millimetre precision however few suitable commercial systems are available. In this work, a compact, real-time daily QA system is optimised and characterised for proton range verification using an ad-hoc Geant4 simulation. The system is comprised of a monolithic silicon diode array detector embedded in a perspex phantom. The detector is orientated at an angular offset to the incident proton beam to allow range in perspex to be determined for flat proton fields. The accuracy of the system for proton range in perspex measurements was experimentally evaluated over the full range of clinical proton energies. The meanR₁₀₀,R₉₀andR₈₀ranges measured with the system were accurate within ±0.6 mm of simulated ranges in a perspex phantom for all energies assessed. This system allows real-time read-out of individual detector channels also making it appropriate for temporal beam delivery diagnostics and for spot position monitoring along one axis. The system presented provides a suitable, economical and efficient alternative for daily QA in proton therapy.

Do we need cardiopulmonary exercise to determine the optimal time for intervention in valvular heart disease?

Introduction

Cardiopulmonary exercise testing (CPET) is underused in many clinical conditions other than HF. In valvular heart disease (VHD), CEPT can aid in choosing the right timing for surgery. The goals of this study were to compare the assessment of functional capacity (FC) by CEPT and NYHA scale, and to analyze the relationship between ventilatory efficiency (VE) parameters and time to surgery.

Methods

197 CPET were performed in 163 patients with moderate or severe VHD (51% female). Real METS (RM) were calculated as indexed peak VO2/3.5 (1 MET = 3.5 ml O2/kg/min) and compared to estimated METS (EM) derived by the exercise duration. An agreement analysis between RM, EM and NYHA was performed. The association among VE/VCO2 slope, pet CO2 at anaerobic threshold (AT), OUES and time to surgical indication was also studied using Cox logistic regression analysis.

Results

See Table. The RM and EM were 4,7±1,7and 6,2±2,9, respectively (p<0.01), and the correlation was low (ICC=0,7, p<0.001). The agreement between NYHA class and % of peak predicted V02 was very low (kappa index = 0.1, p<0.001). VE parameters were predictive of an earlier surgical indication: petCO2 AT (p=0.02), VE/VCO2 slope (p=0.069), OUES (p=0.014).

Conclusions

In asymptomatic VHD patients, the surgical indication should not rely solely in their FC assessed by either NYHA scale or EM derived by duration of exercise. In our series, these parameters clearly overestimated the FC of the population. Also, the ventilatory inefficiency may be a surrogate marker of advanced disease and lead to a closer surveillance for an earlier intervention.

Real and estimated METS by type of VHD

Funding Acknowledgement

Type of funding source: None

P4344An example that big data analysis is ready for the prime time

Introduction

Data about the epidemiology of valvular heart disease (VHD) is scarce. The increasing aging of the population may cause an augmented prevalence of VHD, with a great number of comorbidities that conveys a higher surgical risk. The aim of this study was to describe the prevalence of VHD in the patients attended at our institution from 2007 until 2017 and to describe the main characteristics of this population.

Methods

We used a new tool based on EHRead Technology to extract clinical relevant information from Electronic Health Records, designed for descriptive and predictive big data analysis. All medical reports generated at the outpatient clinic, ER or hospitalization ward were examined. Patients with a diagnosis of moderate or severe VHD were selected. The prevalence of VHD was also estimated in 2 quintiles, from 2008 until Feb 2013 and from March 2013 until Dec 2017.

Results

The total prevalence of VHD in our population was 1.04% (n=3431). Mitral regurgitation was the most frequent valvular lesion (0.4%, n=1318), followed by aortic stenosis (0.3%, n=967) and aortic regurgitation (0.28%, n=938). There was a clear female predominance (63%), and the median age was 76.4. In the 1st quintile the prevalence of VHD was 0.25%, and increased to 0.79% in the 2nd. This trend was consistent in all type of valvular lesions. The prevalence of comorbidities was higher than in other epidemiological studies (Table).

Prevalence of comorbidities Severe MR Severe AS Severe AR Euro Heart Valve Survey Hypertension 54,5% 69,1% 47,9% 49% Dyslipidemia 32,2% 40,6% 27,4% 35% Diabetes Mellitus 28,0% 31,5% 16,4% 15% Smoking (current) 5,6% 5,4% 13,7% 39% Coronary heart disease 12,0% 17,0% 12,3% 13% Stroke 7,0% 8,9% 5,5% 7% Chronic kidney disease 18,9% 16,9% 20,5% 15% Chronic obstructive pulmonary disease 11,2% 9,9% 11,0% 15% MR: Mitral regurgitation, AS: aortic stenosis, AR: aortic regurgitation, MS: mitral stenosis.

Sex Distribution

Conclusions

The older age and greater number of comorbidities seen in our series over the past ten years, compared to the Euroheart Valve Survey reinforce the idea that the percutaneous valvular therapies should play a major role in the treatment of patients with VHD. Although, the prevalence of VHD may be underestimated in our population, due to the methodology, it reflects an ever-growing pathology in an older and sicker population.

P2750Why should cardiopulmonary exercise testing be routinely used for assessing patients with valvular heart disease?

Introduction

The role of cardiopulmonary exercise testing (CPET) is unquestionable to assess prognosis in heart failure. In patients with valvular heart disease (VHD), the functional capacity (FC) is crucial to aid in the right timing of surgery. The aim of this study was to compare the assessment of the FC by CPET and NYHA and the correlation between ventilatory efficiency parameters and resting systolic pulmonary artery pressure (SPAP).

Methods

We studied 100 VHD patients (57% female) who underwent a CPET. We calculated the real METS (RM) as indexed peak VO2/3.5 (1 MET=3.5 ml O2/kg/min) and compared to estimated METS (EM) derived by the time of exercise. An agreement analysis between RM, EM and NYHA was calculated. The correlation among VE/Vslope CO2, EqCO2at anaerobic threshold (AT), PETCO2, partial pressure end-tidal CO2 at AT and SPAP was analyzed.

Results

The results are shown in Table and Figure. The RM and the EM were 4.7±1.7 and 5.5±3, respectively (p<0.01) with a low agreement (ICC=0.6, p<0.01). The agreement between NYHA and the classification obtained from peak % of predicted peak VO2 was very low (weighted kappa =0.06, p=0.28). In patients with severe mitral VHD, the ventilatory efficiency parameters were correlated with SPAP (PETCO2 (AT), r=−0.7, p=0.002; EqCO2 (AT), r=0.5, p=0.04:VE/Vslope CO2, r=0.3, p 0.2), whereas in those with severe aortic VHD, these correlations were much lower (PETCO2 (AT), r=−0.3, p=0.13; EqCO2 (AT), r=0.2, p=0.15; VE/Vslope CO2, r=0.18, p 0.31).

Total (n=100) Mitral regurgitation (n=35) Aortic regurgitation (n=23) Age 65 (29–86) 66 (30–84) 65 (11–87) LVEF (%) 62±6 63±6 61±7 SPAP (mmHg) 40±11 39±11 36±8 NYHA I (60%), II (33%), III (7%) I (63%), II (29%), III (9%) I (63%), II (33%), III (4%) Indexed peak VO2 (ml/min/kg) 16±6 17±6 19±8 Peak % predicted VO2 73±18 74±17 79±18 Predicted VO2 AT (%) 58±19 54±19 61±22 Eq CO2 AT 33±6 32±7 32±5 VE/VSlope CO2 33±6 32±7 33±8 PetCO2 AT 34±4 36±4 36±5

Type and degree of VHD

Conclusions

NYHA scale and estimation of METS derived from the time of exercise clearly overestimated the FC of our population. In our series, the ventilatory inefficiency in patients with mitral VHD could be a surrogate marker of advanced disease and could lead to an earlier intervention.

Highly efficient and sensitive patient-specific quality assurance for spot-scanned proton therapy

The purpose of this work was to develop an end-to-end patient-specific quality assurance (QA) technique for spot-scanned proton therapy that is more sensitive and efficient than traditional approaches. The patient-specific methodology relies on independently verifying the accuracy of the delivered proton fluence and the dose calculation in the heterogeneous patient volume. A Monte Carlo dose calculation engine, which was developed in-house, recalculates a planned dose distribution on the patient CT data set to verify the dose distribution represented by the treatment planning system. The plan is then delivered in a pre-treatment setting and logs of spot position and dose monitors, which are integrated into the treatment nozzle, are recorded. A computational routine compares the delivery log to the DICOM spot map used by the Monte Carlo calculation to ensure that the delivered parameters at the machine match the calculated plan. Measurements of dose planes using independent detector arrays, which historically are the standard approach to patient-specific QA, are not performed for every patient. The nozzle-integrated detectors are rigorously validated using independent detectors in regular QA intervals. The measured data are compared to the expected delivery patterns. The dose monitor reading deviations are reported in a histogram, while the spot position discrepancies are plotted vs. spot number to facilitate independent analysis of both random and systematic deviations. Action thresholds are linked to accuracy of the commissioned delivery system. Even when plan delivery is acceptable, the Monte Carlo second check system has identified dose calculation issues which would not have been illuminated using traditional, phantom-based measurement techniques. The efficiency and sensitivity of our patient-specific QA program has been improved by implementing a procedure which independently verifies patient dose calculation accuracy and plan delivery fidelity. Such an approach to QA requires holistic integration and maintenance of patient-specific and patient-independent QA.

MICRODOSIMETRIC APPLICATIONS IN PROTON AND HEAVY ION THERAPY USING SILICON MICRODOSIMETERS

Using the CMRP 'bridge' μ+ probe, microdosimetric measurements were undertaken out-of-field using a therapeutic scanning proton pencil beam and in-field using a 12C ion therapy field. These measurements were undertaken at Mayo Clinic, Rochester, USA and at HIMAC, Chiba, Japan, respectively. For a typical proton field used in the treatment of deep-seated tumors, we observed dose-equivalent values ranging from 0.62 to 0.99 mSv/Gy at locations downstream of the distal edge. Lateral measurements at depths close to the entrance and along the SOBP plateau were found to reach maximum values of 3.1 mSv/Gy and 5.3 mSv/Gy at 10 mm from the field edge, respectively, and decreased to ~0.04 mSv/Gy 120 mm from the field edge. The ability to measure the dose-equivalent with high spatial resolution is particularly relevant to healthy tissue dose calculations in hadron therapy treatments. We have also shown qualitatively and quantitively the effects critical organ motion would have in treatment using microdosimetric spectra. Large differences in spectra and RBE10 were observed for treatments where miscalculations of 12C ion range would result in critical structures being irradiated, showing the importance of motion management.

Stereotactic ablative radiotherapy delivered by image-guided helical tomotherapy for extracranial oligometastases

PURPOSE

To investigate the outcomes and risk factors of patients treated with stereotactic ablative radiotherapy (SABR) delivered by image-guided helical tomotherapy (HT) for extracranial oligometastases.

METHODS

From August 2006 through July 2011, 42 consecutive patients (median age 69 years [range 16-87]) with oligometastases (≤3) received HT to all known cancer sites (lung, n = 28; liver, n = 12; adrenal, n = 2). Prognostic factors were assessed by Cox's proportional hazards regression analysis.

RESULTS

A total of 60 lesions were treated with hypofractionated HT (median dose 39 Gy [range 36-72.5]; median dose per fraction 12 Gy [range 5-20]). Complete or partial response was observed in 40 (54 %) patients. With a median follow-up period of 15 months, 1- and 2-year overall survival (OS) was 84 and 63 %, respectively; and 1- and 2-year local control (LC) was 92 and 86 %, respectively. Four patients had pneumonitis Grade ≥2 and two patients had lower gastrointestinal toxicity Grade ≥2. Only the lack of complete/partial response was associated with higher risk of mortality on univariate (HR = 3.8, P = 0.04) and multivariate (HR = 6.6, P = 0.01) analyses.

CONCLUSIONS

SABR delivered by image-guided HT is well tolerated and offers adequate LC with low acute morbidity in patients with extracranial oligometastatic disease. We found that the response to HT was the only predictor for OS.