Pelvic PET/MR attenuation correction in the image space using deep learning.

Introduction: The five-class Dixon-based PET/MR attenuation correction (AC) model, which adds bone information to the four-class model by registering major bones from a bone atlas, has been shown to be error-prone. In this study, we introduce a novel method of accounting for bone in pelvic PET/MR AC by directly predicting the errors in the PET image space caused by the lack of bone in four-class Dixon-based attenuation correction. Methods: A convolutional neural network was trained to predict the four-class AC error map relative to CT-based attenuation correction. Dixon MR images and the four-class attenuation correction µ-map were used as input to the models. CT and PET/MR examinations for 22 patients ([18F]FDG) were used for training and validation, and 17 patients were used for testing (6 [18F]PSMA-1007 and 11 [68Ga]Ga-PSMA-11). A quantitative analysis of PSMA uptake using voxel- and lesion-based error metrics was used to assess performance. Results: In the voxel-based analysis, the proposed model reduced the median root mean squared percentage error from 12.1% and 8.6% for the four- and five-class Dixon-based AC methods, respectively, to 6.2%. The median absolute percentage error in the maximum standardized uptake value (SUVmax) in bone lesions improved from 20.0% and 7.0% for four- and five-class Dixon-based AC methods to 3.8%. Conclusion: The proposed method reduces the voxel-based error and SUVmax errors in bone lesions when compared to the four- and five-class Dixon-based AC models.

Head and neck cancer treatment outcome prediction: a comparison between machine learning with conventional radiomics features and deep learning radiomics.

Background: Radiomics can provide in-depth characterization of cancers for treatment outcome prediction. Conventional radiomics rely on extraction of image features within a pre-defined image region of interest (ROI) which are typically fed to a classification algorithm for prediction of a clinical endpoint. Deep learning radiomics allows for a simpler workflow where images can be used directly as input to a convolutional neural network (CNN) with or without a pre-defined ROI. Purpose: The purpose of this study was to evaluate (i) conventional radiomics and (ii) deep learning radiomics for predicting overall survival (OS) and disease-free survival (DFS) for patients with head and neck squamous cell carcinoma (HNSCC) using pre-treatment 18F-fluorodeoxuglucose positron emission tomography (FDG PET) and computed tomography (CT) images. Materials and methods: FDG PET/CT images and clinical data of patients with HNSCC treated with radio(chemo)therapy at Oslo University Hospital (OUS; n = 139) and Maastricht University Medical Center (MAASTRO; n = 99) were collected retrospectively. OUS data was used for model training and initial evaluation. MAASTRO data was used for external testing to assess cross-institutional generalizability. Models trained on clinical and/or conventional radiomics features, with or without feature selection, were compared to CNNs trained on PET/CT images without or with the gross tumor volume (GTV) included. Model performance was measured using accuracy, area under the receiver operating characteristic curve (AUC), Matthew's correlation coefficient (MCC), and the F1 score calculated for both classes separately. Results: CNNs trained directly on images achieved the highest performance on external data for both endpoints. Adding both clinical and radiomics features to these image-based models increased performance further. Conventional radiomics including clinical data could achieve competitive performance. However, feature selection on clinical and radiomics data lead to overfitting and poor cross-institutional generalizability. CNNs without tumor and node contours achieved close to on-par performance with CNNs including contours. Conclusion: High performance and cross-institutional generalizability can be achieved by combining clinical data, radiomics features and medical images together with deep learning models. However, deep learning models trained on images without contours can achieve competitive performance and could see potential use as an initial screening tool for high-risk patients.

Serum cytokine profiles and metabolic tumor burden in patients with non-small cell lung cancer undergoing palliative thoracic radiation therapy.

PURPOSE: Radiation therapy effectively kills cancer cells and elicits local effects in the irradiated tissue. The aim of this study was to investigate the kinetics of cytokines in the serum of patients with lung cancer undergoing radiation therapy and to identify associations with metabolic tumor burden as determined by 2-deoxy-2-fluoro-D-glucose (18F-FDG) positron emission tomography (PET). METHODS AND MATERIALS: Forty-five patients with advanced non-small cell lung cancer were included in a phase 2 clinical trial and randomized between fractionated thoracic radiation therapy alone or concurrent with an epidermal growth factor receptor inhibitor. Blood was sampled at 4 different time points: prior to treatment, midtherapy, at the end of therapy, and 6 to 8 weeks after the start of treatment. The serum concentrations of 48 cytokines and 9 matrix metalloproteinases were measured with multiplex immunoassays. A subset of patients was examined by 18F-FDG PET/computed tomography before, during, and after radiation therapy. The maximum standardized uptake values (SUVmax) of the primary lung tumor, whole-body metabolic tumor volume, and total lesion glycolysis were calculated, and correlations between the PET parameters and cytokines were investigated. RESULTS: The SUVmax decreased from baseline through midtherapy to posttherapy 18F-FDG PET/computed tomography (P = .018). The serum levels of C-C motif chemokine ligand (CCL) 23, CCL24, C-X3-C motif chemokine ligand 1, and interleukin-8 (C-X-C motif ligand [CXCL]8) were significantly correlated to SUVmax, metabolic tumor volume, and total lesion glycolysis before, during, and after radiation therapy. CXCL2 (P = .030) and CXCL6 (P = .010) decreased after the start of therapy and changed significantly across the sample time points. Serum concentrations of CCL15 (P = .031), CXCL2 (P = .028), and interleukin-6 (P = .007) were positively correlated to the irradiated volume during the second week of treatment. CONCLUSIONS: Cytokine serum levels vary and correlate with metabolic tumor burden in patients with advanced non-small cell lung cancer undergoing palliative thoracic radiation therapy.

A new method to assess pulmonary changes using 18F-fluoro-2-deoxyglucose positron emission tomography for lung cancer patients following radiotherapy.

BACKGROUND: 18F-fluoro-2-deoxyglucose positron emission tomography (18F-FDG-PET) may be used for assessing radiation induced alterations in the lung. However, there is a need to further develop methodologies to improve quantification. Using computed tomography (CT), a local structure method has been shown to be superior to conventional CT-based analysis. Here, we investigate whether the local structure method based on 18F-FDG-PET improves radiotherapy (RT) dose-response quantification for lung cancer patients. MATERIAL AND METHODS: Sixteen patients with lung cancer undergoing fractionated RT were examined by 18F-FDG-PET/CT at three sessions (pre, mid, post) and the lung was delineated in the planning CT images. The RT dose matrix was co-registered with the PET images. For each PET image series, mean (µ) and standard deviation (σ) maps were calculated based on cubes in the lung (3 × 3 × 3 voxels), where the spread in pre-therapy µ and σ was characterized by a covariance ellipse in a sub-volume of 3 × 3 × 3 cubes. Mahalanobis distance was used to measure the distance of individual cube values to the origin of the ellipse and to further form local structure 'S' maps. The structural difference maps (ΔS) and mean difference maps (Δµ) were calculated by subtracting pre-therapy maps from maps at mid- and post-therapy. Corresponding maps based on CT images were also generated. RESULTS: ΔS identified new areas of interest in the lung compared to conventional Δµ maps. ΔS for PET and CT gave a significantly elevated lung signal compared to a control group during and post-RT (p < .05). Dose-response analyses by linear regression showed that ΔS between pre- and post-therapy for 18F-FDG-PET was the only parameter significantly associated with local lung dose (p = .04). CONCLUSIONS: The new method using local structures on 18F-FDG-PET provides a clearer uptake dose-response compared to conventional analysis and CT-based approaches and may be valuable in future studies addressing lung toxicity.

Assessment of pulmonary 18F-FDG-PET uptake and cytokine profiles in non-small cell lung cancer patients treated with radiotherapy and erlotinib.

PURPOSE: To investigate effects of radiotherapy (RT) and erlotinib on pulmonary glucose uptake using 2-deoxy-2-(18F)fluoro-D-glucose (18F-FDG) positron emission tomography (PET) during and after treatment of non-small cell lung cancer (NSCLC) and to identify associations between serum cytokine levels and lung glucose uptake. MATERIAL AND METHODS: Twenty-seven patients with advanced NSCLC, receiving RT alone or concomitant RT and erlotinib therapy, were examined by 18F-FDG PET before, during, and after treatment. A total of 57 18F-FDG PET scans were analyzed. Pulmonary 18F-FDG uptake and radiotherapy dose mapping were used to acquire dose-response curves for each patient, where subsequent linear regression gave a glucose uptake level in the un-irradiated parts of the lungs (SUV0) and a response slope (ΔSUV). Serum cytokine levels at corresponding time points were assessed using a multiplex bioassay. Correlations between the most robust cytokines and lung 18F-FDG dose response parameters were further investigated. RESULTS: From the dose response analysis, SUV0 at post-therapy was significantly higher (P < 0.001) than at mid- and pre-therapy (45% and 58%, respectively) for the group receiving RT + erlotinib. Also, SUV0 at post-therapy was higher for patients receiving RT + erlotinib compared to RT alone (42%; P < 0.001). No differences in ΔSUV were seen with treatments or time. SUV0 was positively associated (r = 0.47, P = 0.01) with serum levels of the chemokine C-C motif ligand 21 (CCL21) for patients receiving RT + erlotinib. CONCLUSIONS: Concomitant RT and erlotinib causes an elevation in pulmonary 18F-FDG uptake post treatment compared to RT alone. Pulmonary glucose uptake is associated with an upregulation of a chemokine (CCL21) involved in inflammatory reactions.

Validation of dose painting of lung tumours using alanine/EPR dosimetry.

Biologic image guided radiotherapy (RT) with escalated doses to tumour sub volumes challenges today's RT dose planning and delivery systems. In this phantom study, we verify the capability of a clinical dose planning and delivery system to deliver an 18F-FDG-PET based dose painted treatment plan to a lung tumour. Furthermore, we estimate the uncertainties of the dose painted treatment compared to conventional RT plans. An anthropomorphic thorax phantom of polystyrene and polyurethane was constructed based on CT images of a lung cancer patient. 101 EPR/alanine dosimeters were placed in separate cavities within the phantom. IMRT and VMAT plans were generated in Eclipse (version 10.0, Analytical Anisotropic Algorithm version 10.2.28, Varian Medical Systems, Inc.) for 6 and 15 MV photons, based on 18F-FDG-PET/CT images of the patient. A boost dose of 3.8 Gy/fraction was given to the 18F-FDG-avid region (biological planning volume; BTV), whereas 3.1 Gy/fraction was planned to the planning target volume (PTV, excluding the BTV). For the homogenous plans, 3.2 Gy/fraction was given to the PTV. Irradiation of the phantom was carried out at a Varian Trilogy linear accelerator (Varian Medical Systems, Inc.). Uncertainties involved in treatment planning and delivery were estimated from portal dosimetry gamma evaluation. Measured and calculated doses were compared by Bland-Altmann analysis. For all treatment plans, all dose-volume objectives could be achieved in the treatment planning system. The mean absolute differences between calculated and measured doses were small (<0.1 Gy) for BTV, PTV-BTV, lung and soft tissue. The estimated uncertainty of the planned doses was less than 3% for all plans, whereas the estimated uncertainty in the measured doses was less 2.3%. Our results show that planning and delivery of dose escalated lung cancer treatment on a clinical dose planning and delivery system has high dosimetric accuracy. The uncertainties associated with the dose escalated treatment plans are comparable to the conventional plans.

Dose painting by numbers in a standard treatment planning system using inverted dose prescription maps.

BACKGROUND: Dose painting by numbers (DPBN) is a method to deliver an inhomogeneous tumor dose voxel-by-voxel with a prescription based on biological medical images. However, planning of DPBN is not supported by commercial treatment planning systems (TPS) today. Here, a straightforward method for DPBN with a standard TPS is presented. MATERIAL AND METHODS: DPBN tumor dose prescription maps were generated from (18)F-FDG-PET images applying a linear relationship between image voxel value and dose. An inverted DPBN prescription map was created and imported into a standard TPS where it was defined as a mock pre-treated dose. Using inverse optimization for the summed dose, a planned DPBN dose distribution was created. The procedure was tested in standard TPS for three different tumor cases; cervix, lung and head and neck. The treatment plans were compared to the prescribed DPBN dose distribution by three-dimensional (3D) gamma analysis and quality factors (QFs). Delivery of the DPBN plans was assessed with portal dosimetry (PD). RESULTS: Maximum tumor doses of 149%, 140% and 151% relative to the minimum tumor dose were prescribed for the cervix, lung and head and neck case, respectively. DPBN distributions were well achieved within the tumor whilst normal tissue doses were within constraints. Generally, high gamma pass rates (> 89% at 2%/2 mm) and low QFs (< 2.6%) were found. PD showed that all DPBN plans could be successfully delivered. CONCLUSIONS: The presented methodology enables the use of currently available TPSs for DPBN planning and delivery and may therefore pave the way for clinical implementation.

Impact of PET reconstruction algorithm and threshold on dose painting of non-small cell lung cancer.

PURPOSE: In the current work, we investigate the impact of PET reconstruction methods (RMs) and threshold on two types of dose painting (DP) prescription strategies for non-small cell lung cancer (NSCLC). MATERIALS AND METHODS: Sixteen patients with NSCLC underwent an 18F-FDG-PET/CT examination prior to radiotherapy. Six different RMs were used. For both a dose painting by contours (DPBC) and a dose painting by numbers (DPBN) strategy, the prescribed radiation dose within the gross tumor volume (GTV) was mapped according to the spatial distribution of standardized uptake values (SUVs). SUVmax and SUVpeak were used for volume thresholding in DPBC and a linear SUV-dose scaling approach was used for DPBN. Deviations from the dose prescription as determined by the standard RM was scored by a quality factor (QF). RESULTS: For DPBC, the mean difference in thresholded boost volume between RMs was typically within 10%. The difference in dose prescription was systematically lower for thresholding based on SUVpeak (largest mean QF 2.8±2.0%) compared to SUVmax (largest mean QF 3.6±3.0%). For DPBN, the resulting dose prescriptions were less dependent on RM and threshold; the largest mean QFs were 1.3±0.3% both for SUVmax and SUVpeak. CONCLUSIONS: PET reconstruction algorithms will both influence DPBC and DPBN, although the impact is smaller for DPBN. For some patients, the resulting variations in dose prescriptions may result in clinically different dose distributions. SUVpeak is a more robust thresholding parameter than SUVmax.

An implementation of time-efficient respiratory-gated PET acquisition by repeated breath-holds.

BACKGROUND: Respiratory gating in positron emission tomography (PET) is used to improve detection of small tumors in the lower lung regions and in the liver, and to obtain a better estimate of the standardized uptake value (SUV). PURPOSE: To develop a time-efficient method for acquisition of respiratory-gated PET/CT that would produce one single high quality image volume corresponding to a breath-hold state. MATERIAL AND METHODS: An instrument was developed that displayed to the patient either red or green numbers, counting down from a chosen maximum to one at a rate of one dial per second. The patient was instructed to repeatedly hold the breath in moderate inspiration when red numbers were displayed and to breathe freely during display of green numbers. PET data were acquired in list mode and trigger signals were sent to the scanner and inserted into the list file each time the color of the countdown numbers switched from green to red. Data acquired during breath-holds were used to create one single image volume. RESULTS: High quality breath-hold images were obtained from 10 min data acquisition at one bed position. Improved image quality compared to standard whole-body PET was demonstrated by a significant reduction of noise (standard deviation) in regions of normal liver tissues. CONCLUSION: The instruction to perform repeated breath-holds was well understood by patients and they cooperated satisfactorily. When the new procedure is used the duration of the data acquisition may typically be reduced by a factor of 4 compared to conventional respiratory-gated protocols where the patient breathes freely.

Spatiotemporal analysis of tumor uptake patterns in dynamic (18)FDG-PET and dynamic contrast enhanced CT.

BACKGROUND: Molecular and functional imaging techniques such as dynamic positron emission tomography (DPET) and dynamic contrast enhanced computed tomography (DCECT) may provide improved characterization of tumors compared to conventional anatomic imaging. The purpose of the current work was to compare spatiotemporal uptake patterns in DPET and DCECT images. MATERIALS AND METHODS: A PET/CT protocol comprising DCECT with an iodine based contrast agent and DPET with (18)F-fluorodeoxyglucose was set up. The imaging protocol was used for examination of three dogs with spontaneous tumors of the head and neck at sessions prior to and after fractionated radiotherapy. Software tools were developed for downsampling the DCECT image series to the PET image dimensions, for segmentation of tracer uptake pattern in the tumors and for spatiotemporal correlation analysis of DCECT and DPET images. RESULTS: DCECT images evaluated one minute post injection qualitatively resembled the DPET images at most imaging sessions. Segmentation by region growing gave similar tumor extensions in DCECT and DPET images, with a median Dice similarity coefficient of 0.81. A relatively high correlation (median 0.85) was found between temporal tumor uptake patterns from DPET and DCECT. The heterogeneity in tumor uptake was not significantly different in the DPET and DCECT images. The median of the spatial correlation was 0.72. CONCLUSIONS: DCECT and DPET gave similar temporal wash-in characteristics, and the images also showed a relatively high spatial correlation. Hence, if the limited spatial resolution of DPET is considered adequate, a single DPET scan only for assessing both tumor perfusion and metabolic activity may be considered. However, further work on a larger number of cases is needed to verify the correlations observed in the present study.

Dynamic respiratory gated (18)FDG-PET of lung tumors - a feasibility study.

BACKGROUND: (18)FDG-PET/CT imaging is well established for diagnosis and staging of lung tumors. However, more detailed information regarding the distribution of FDG within the tumor, also as a function of time after injection may be relevant. In this study we explore the feasibility of a combined dynamic and respiratory gated (DR) PET protocol. MATERIAL AND METHODS: A DR FDG-PET protocol for a Siemens Biograph 16 PET/CT scanner was set up, allowing data acquisition from the time of FDG injection. Breath-hold (BH) respiratory gating was performed at four intervals over a total acquisition time of 50 minutes. Thus, the PET protocol provides both motion-free images and a spatiotemporal characterization of the glucose distribution in lung tumors. Software tools were developed in-house for tentative tumor segmentation and for extracting standard uptake values (SUVs) voxel by voxel, tumor volumes and SUV gradients in all directions. RESULTS: Four pilot patients have been investigated with the DR PET protocol. The procedure was well tolerated by the patients. The BH images appeared sharper, and SUV(max)/SUV(mean) was higher, compared to free breathing (FB) images. Also, SUV gradients in the periphery of the tumor in the BH images were in general greater than or equal to the gradients in the FB PET images. CONCLUSION: The DR FDG-PET protocol is feasible and the BH images have a superior quality compared to the FB images. The protocol may also provide information of relevance for radiotherapy planning and follow-up. A patient trial is needed for assessing the clinical value of the imaging protocol.

Feasibility of contrast-enhanced cone-beam CT for target localization and treatment monitoring.

A dog with a spontaneous maxillary tumour was given 40 Gy of fractionated radiotherapy. At five out of 10 fractions cone-beam CT (CBCT) imaging before and after administration of an iodinated contrast agent were performed. Contrast enhancement maps were overlaid on the pre-contrast CBCT images. The tumour was clearly visualized in the images thus produced.