Automatic segmentation of head and neck primary tumors on MRI using a multi-view CNN

Background Accurate segmentation of head and neck squamous cell cancer (HNSCC) is important for radiotherapy treatment planning. Manual segmentation of these tumors is time-consuming and vulnerable to inconsistencies between experts, especially in the complex head and neck region. The aim of this study is to introduce and evaluate an automatic segmentation pipeline for HNSCC using a multi-view CNN (MV-CNN). Methods The dataset included 220 patients with primary HNSCC and availability of T1-weighted, STIR and optionally contrast-enhanced T1-weighted MR images together with a manual reference segmentation of the primary tumor by an expert. A T1-weighted standard space of the head and neck region was created to register all MRI sequences to. An MV-CNN was trained with these three MRI sequences and evaluated in terms of volumetric and spatial performance in a cross-validation by measuring intra-class correlation (ICC) and dice similarity score (DSC), respectively. Results The average manual segmented primary tumor volume was 11.8±6.70 cm3 with a median [IQR] of 13.9 [3.22-15.9] cm3. The tumor volume measured by MV-CNN was 22.8±21.1 cm3 with a median [IQR] of 16.0 [8.24-31.1] cm3. Compared to the manual segmentations, the MV-CNN scored an average ICC of 0.64±0.06 and a DSC of 0.49±0.19. Improved segmentation performance was observed with increasing primary tumor volume: the smallest tumor volume group (<3 cm3) scored a DSC of 0.26±0.16 and the largest group (>15 cm3) a DSC of 0.63±0.11 (p<0.001). The automated segmentation tended to overestimate compared to the manual reference, both around the actual primary tumor and in false positively classified healthy structures and pathologically enlarged lymph nodes. Conclusion An automatic segmentation pipeline was evaluated for primary HNSCC on MRI. The MV-CNN produced reasonable segmentation results, especially on large tumors, but overestimation decreased overall performance. In further research, the focus should be on decreasing false positives and make it valuable in treatment planning.

FLASH Radiotherapy Using Single-Energy Proton PBS Transmission Beams for Hypofractionation Liver Cancer: Dose and Dose Rate Quantification

PurposeThis work aims to study the dose and ultra-high-dose rate characteristics of transmission proton pencil beam scanning (PBS) FLASH radiotherapy (RT) for hypofractionation liver cancer based on the parameters of a commercially available proton system operating under FLASH mode.Methods and MaterialsAn in-house treatment planning software (TPS) was developed to perform intensity-modulated proton therapy (IMPT) FLASH-RT planning. Single-energy transmission proton PBS plans of 4.5 Gy × 15 fractions were optimized for seven consecutive hepatocellular carcinoma patients, using 2 and 5 fields combined with 1) the minimum MU/spot chosen between 100 and 400, and minimum spot time (MST) of 2 ms, and 2) the minimum MU/spot of 100, and MST of 0.5 ms, based upon considerations in target uniformities, OAR dose constraints, and OAR FLASH dose rate coverage. Then, the 3D average dose rate distribution was calculated. The dose metrics for the mean dose of Liver-GTV and other major OARs were characterized to evaluate the dose quality for the different combinations of field numbers and minimum spot times compared to that of conventional IMPT plans. Dose rate quality was evaluated using 40 Gy/s volume coverage (V40Gy/s).ResultsAll plans achieved favorable and comparable target uniformities, and target uniformity improved as the number of fields increased. For OARs, no significant dose differences were observed between plans of different field numbers and the same MST. For plans using shorter MST and the same field numbers, better sparing was generally observed in most OARs and was statistically significant for the chest wall. However, the FLASH dose rate coverage V40Gy/s was increased by 20% for 2-field plans compared to 5-field plans in most OARs with 2-ms MST, which was less evident in the 0.5-ms cases. For 2-field plans, dose metrics and V40Gy/s of select OARs have large variations due to the beam angle selection and variable distances to the targets. The transmission plans generally yielded inferior dosimetric quality to the conventional IMPT plans.ConclusionThis is the first attempt to assess liver FLASH treatment planning and demonstrates that it is challenging for hypofractionation with smaller fractional doses (4.5 Gy/fraction). Using fewer fields can allow higher minimum MU/spot, resulting in higher OAR FLASH dose rate coverages while achieving similar plan quality compared to plans with more fields. Shorter MST can result in better plan quality and comparable or even better FLASH dose rate coverage.

Deep Machine Learning for Oral Cancer: From Precise Diagnosis to Precision Medicine

Oral squamous cell carcinoma (OSCC) is one of the most prevalent cancers worldwide and its incidence is on the rise in many populations. The high incidence rate, late diagnosis, and improper treatment planning still form a significant concern. Diagnosis at an early-stage is important for better prognosis, treatment, and survival. Despite the recent improvement in the understanding of the molecular mechanisms, late diagnosis and approach toward precision medicine for OSCC patients remain a challenge. To enhance precision medicine, deep machine learning technique has been touted to enhance early detection, and consequently to reduce cancer-specific mortality and morbidity. This technique has been reported to have made a significant progress in data extraction and analysis of vital information in medical imaging in recent years. Therefore, it has the potential to assist in the early-stage detection of oral squamous cell carcinoma. Furthermore, automated image analysis can assist pathologists and clinicians to make an informed decision regarding cancer patients. This article discusses the technical knowledge and algorithms of deep learning for OSCC. It examines the application of deep learning technology in cancer detection, image classification, segmentation and synthesis, and treatment planning. Finally, we discuss how this technique can assist in precision medicine and the future perspective of deep learning technology in oral squamous cell carcinoma.

Building Sustainable Practice Transformation Through Payment Reform Initiatives

PURPOSE: Novel value-based payment approaches provide an opportunity to deploy and sustain health care delivery interventions, such as treatment planning documentation. However, limited data are available on implementation costs. METHODS: We described key factors affecting the cost of implementing care improvements under value-based payments, using treatment planning and Medicare's Oncology Care Model as examples. We estimated expected costs of implementing treatment plans for years 1 and 2-6 under (1) different staffing models, (2) use of technology, and (3) differences in the patients engaged. We compared costs to the payment amounts under the Oncology Care Model. RESULTS: Team-based models where staffing is aligned with skills needed for key tasks (eg, a combination of lay navigator, nurse, and physician) are more financially feasible when compared with using physicians or nurses alone. When existing staff are at or near capacity, hiring new staff focused on practice transformation activities allows adequate time for new initiatives without negative impacts on existing services. Investments in information technology can enhance staff productivity, but initial costs may be high. Interventions may not be financially feasible if implemented for a small patient volume or only for patients insured by a particular payer. Finally, costs may be higher for disadvantaged populations, and equity in care delivery may require higher payments from payers. CONCLUSION: Estimating the cost of implementing an intervention in different types of practice settings with various types of patients is essential to ensure that a value-based payment system will adequately support desired improvements in quality of care for all patients.

The Evolution of Lateral Dose Distributions of Helium Ion Beams in Air: From Measurement and Modeling to Their Impact on Treatment Planning

To start clinical trials with the first clinical treatment planning system supporting raster-scanned helium ion therapy, a comprehensive database of beam characteristics and parameters was required for treatment room-specific beam physics modeling at the Heidelberg Ion-Beam Therapy Center (HIT). At six different positions in the air gap along the beam axis, lateral beam profiles were systematically measured for 14 initial beam energies covering the full range of available energies at HIT. The 2D-array of liquid-filled ionization chambers OCTAVIUS from PTW was irradiated by a pencil beam focused at the central axis. With a full geometric representation of HIT’s monitoring chambers and beamline elements in FLUKA, our Monte Carlo beam model matches the measured lateral beam profiles. A second set of measurements with the detector placed in a water tank was used to validate the adjustments of the initial beam parameters assumed in the FLUKA simulation. With a deviation between simulated and measured profiles below ±0.8 mm for all investigated beam energies, the simulated profiles build part of the database for the first clinical treatment planning system for helium ions. The evolution of beamwidth was also compared to similar simulations of the clinically available proton and carbon beam. This allows a choice of treatment modality based on quantitative estimates of the physical beam properties. Finally, we investigated the influence of beamwidth variation on patient treatment plans in order to estimate the relevance and necessary precision limits for lateral beam width models.

Multi-parametric PET/MRI With FDG and RGD for Enhanced Tumor Characterization of Patients With Cervical Cancer

Aim: The concept of personalized medicine has brought increased awareness to the importance of inter- and intra-tumor heterogeneity for cancer treatment. The aim of this study was to explore simultaneous multi-parametric PET/MRI prior to chemoradiotherapy for cervical cancer for characterization of tumors and tumor heterogeneity. Methods: Ten patients with histologically proven primary cervical cancer were examined with multi-parametric 68Ga-NODAGA-E[c(RGDyK)]2-PET/MRI for radiation treatment planning after diagnostic 18F-FDG-PET/CT. Standardized uptake values (SUV) of RGD and FDG, diffusion weighted MRI and the derived apparent diffusion coefficient (ADC), and pharmacokinetic maps obtained from dynamic contrast-enhanced MRI with the Tofts model (iAUC60, Ktrans, ve, and kep) were included in the analysis. The spatial relation between functional imaging parameters in tumors was examined by a correlation analysis and joint histograms at the voxel level. The ability of multi-parametric imaging to identify tumor tissue classes was explored using an unsupervised 3D Gaussian mixture model-based cluster analysis.Results: Functional MRI and PET of cervical cancers appeared heterogeneous both between patients and spatially within the tumors, and the relations between parameters varied strongly within the patient cohort. The strongest spatial correlation was observed between FDG uptake and ADC (median r=-0.7). There was moderate voxel-wise correlation between RGD and FDG uptake, and weak correlations between all other modalities. Distinct relations between the ADC and RGD uptake as well as the ADC and FDG uptake were apparent in joint histograms. A cluster analysis using the combination of ADC, FDG and RGD uptake suggested tissue classes which could potentially relate to tumor sub-volumes. Conclusion: A multi-parametric PET/MRI examination of patients with cervical cancer integrated with treatment planning and including estimation of angiogenesis and glucose metabolism as well as MRI diffusion and perfusion parameters is feasible. A combined analysis of functional imaging parameters indicates a potential of multi-parametric PET/MRI to contribute to a better characterization of tumor heterogeneity than the modalities alone. However, the study is based on small patient numbers and further studies are needed prior to the future design of individually adapted treatment approaches based on multi-parametric functional imaging.