A.S.R. Mohamed

TU‐F‐CAMPUS‐I‐01: Head and Neck Squamous Cell Carcinoma: Short‐Term Repeatability of Apparent Diffusion Coefficient and Intravoxel Incoherent Motion Parameters at 3.0T

Purpose: Many published studies have recently demonstrated the potential value of intravoxel incoherent motion (IVIM) analysis for disease evaluation. However, few have questioned its measurement repeatability/reproducibility when applied. The purpose of this study was to determine the short-term measurement repeatability of apparent diffusion coefficient ADC, true diffusion coefficient D, pseudodiffusion coefficient D* and perfusion fraction f, in head and neck squamous cell carcinoma (HNSCC) primary tumors and metastatic nodes. Methods: Ten patients with known HNSCC were examined twice using echo-planar DW-MRI with 12 b values (0 to 800 s/mm2) 1hour to 24 hours apart before radiation treatment. All patients were scanned with the customized radiation treatment immobilization devices to reduce motion artifacts and to improve image registration in repeat scans. Regions of interests were drawn in primary tumor and metastases node in each patient (Fig. 1). ADC and IVIM parameters D, D* and f were calculated by least squares data fitting. Short-term test–retest repeatability of ADC and IVIM parameters were assessed by measuring Bland–Altman limits of agreements (BA-LA). Results: Sixteen HNSCC lesions were assessed in 10 patients. Repeatability of perfusion-sensitive parameters, D* and f, in HNSCC lesions was poor (BA-LA: -144% to 88% and −57% to 96% for D* and f, respectively); a lesser extent was observed for the diffusion-sensitive parameters of ADC and D (BA-LA: −34% to 39% and −37% to 40%, for ADC and D, respectively) (Fig. 2). Conclusion: Poor repeatability of D*/f and good repeatability for ADC/D were observed in HNSCC primary tumors and metastatic nodes. Efforts should be made to improve the measurement repeatability of perfusion-sensitive IVIM parameters.

SU‐E‐J‐220: Assessment of MRI Geometric Distortion in Head and Neck Cancer Patients Scanned in Immobilized Radiation Treatment Position

Purpose: Uncertainties about geometric distortion have somewhat hindered MRI simulation in radiation therapy. Most of the geometric distortion studies were performed with phantom measurements but another major aspect of MR distortion is patient related. We studied the geometric distortion in patient images by comparing their MRI scans with the corresponding CT, using CT as the non-distorted gold standard. Methods: Ten H&N cancer patients were imaged with MRI as part of a prospective IRB approved study. All patients had their treatment planning CT done on the same day or within one week of the MRI. MR Images were acquired with a T2 SE sequence (1×1×2.5mm voxel size) in the same immobilization position as in the CT scans. MRI to CT rigid registration was then done and geometric distortion comparison was done by measuring the corresponding anatomical landmarks on both the MRI and the CT images by two observers. Several skin to skin (9 landmarks), bone to bone (8 landmarks), and soft tissue (3 landmarks) were measured at specific levels in horizontal and vertical planes of both scans. Results: The mean distortion for all landmark measurements in all scans was 1.8±1.9mm. For each patient 11 measurements were done in the horizontal planemorexa0» while 9 were done in the vertical plane. The measured geometric distortion were significantly lower in the horizontal axis compared to the vertical axis (1.3±0.16 mm vs 2.2±0.19 mm, respectively, P=0.003*). The magnitude of distortion was lower in the bone to bone landmarks compared to the combined soft tissue and skin to skin landmarks (1.2±0.19 mm vs 2.3±0.17 mm, P=0.0006*). The mean distortion measured by observer one was not significantly different compared toobserver 2 (2.3 vs 2.4 mm, P=0.4). Conclusion: MRI geometric distortions were quantified in H&N patients with mean error of less than 2 mm. JW received a corporate sponsored research grant from Elekta.«xa0less

SU‐E‐J‐225: CEST Imaging in Head and Neck Cancer Patients

Purpose: Chemical Exchange Saturation Transfer (CEST) imaging is an MRI technique enables the detection and imaging of metabolically active compounds in vivo. It has been used to differentiate tumor types and metabolic characteristics. Unlike PET/CT,CEST imaging does not use isotopes so it can be used on patient repeatedly. This study is to report the preliminary results of CEST imaging in Head and Neck cancer (HNC) patients. Methods: A CEST imaging sequence and the post-processing software was developed on a 3T clinical MRI scanner. Ten patients with Human papilloma virus positive oropharyngeal cancer were imaged in their immobilized treatment position. A 5 mm slice CEST image was acquired (128×128, FOV=20∼24cm) to encompass the maximum dimension of tumor. Twenty-nine off-set frequencies (from −7.8ppm to +7.8 ppm) were acquired to obtain the Z-spectrum. Asymmetry analysis was used to extract the CEST contrasts. ROI at the tumor, node and surrounding tissues were measured. Results: CEST images were successfully acquired and Zspectrum asymmetry analysis demonstrated clear CEST contrasts in tumor as well as the surrounding tissues. 3∼5% CEST contrast in the range of 1 to 4 ppm was noted in tumor as well as grossly involved nodes. Injection of glucose produced a marked increase of CEST contrast in tumor region (∼10%). Motion and pulsation artifacts tend to smear the CEST contrast, making the interpretation of the image contrast difficult. Field nonuniformity, pulsation in blood vesicle and susceptibility artifacts caused by air cavities were also problematic for CEST imaging. Conclusion: We have demonstrated successful CEST acquisition and Z-spectrum reconstruction on HNC patients with a clinical scanner. MRI acquisition in immobilized treatment position is critical for image quality as well as the success of CEST image acquisition. CEST images provide novel contrast of metabolites in HNC and present great potential in the pre- and post-treatment assessment of patients undergoing radiation therapy.

Precision Risk Analysis of Cancer Therapy with Interactive Nomograms and Survival Plots

We present the design and evaluation of an integrated problem solving environment for cancer therapy analysis. The environment intertwines a statistical martingale model and a K Nearest Neighbor approach with visual encodings, including novel interactive nomograms, in order to compute and explain a patients probability of survival as a function of similar patient results. A coordinated views paradigm enables exploration of the multivariate, heterogeneous and few-valued data from a large head and neck cancer repository. A visual scaffolding approach further enables users to build from familiar representations to unfamiliar ones. Evaluation with domain experts show how this visualization approach and set of streamlined workflows enable the systematic and precise analysis of a patient prognosis in the context of cohorts of similar patients. We describe the design lessons learned from this successful, multi-site remote collaboration.

Imaging-Genomics Study Of Head-Neck Squamous Cell Carcinoma: Associations Between Radiomic Phenotypes And Genomic Mechanisms Via Integration Of TCGA And TCIA

Purpose Recent data suggest that imaging radiomics features for a tumor could predict important genomic biomarkers. Understanding the relationship between radiomic and genomic features is important for basic cancer research and future patient care. For Head and Neck Squamous Cell Carcinoma (HNSCC), we perform a comprehensive study to discover the imaging-genomics associations and explore the potential of predicting tumor genomic alternations using radiomic features. Methods Our retrospective study integrates whole-genome multi-omics data from The Cancer Genome Atlas (TCGA) with matched computed tomography imaging data from The Cancer Imaging Archive (TCIA) for the same set of 126 HNSCC patients. Linear regression analysis and gene set enrichment analysis are used to identify statistically significant associations between radiomic imaging features and genomic features. Random forest classifier is used to predict two key HNSCC molecular biomarkers, the status of human papilloma virus (HPV) and disruptive TP53 mutation, based on radiomic features. Results Wide-spread and statistically significant associations are discovered between genomic features (including miRNA expressions, protein expressions, somatic mutations, and transcriptional activities, copy number variations, and promoter region DNA methylation changes of pathways) and radiomic features characterizing the size, shape, and texture of tumor. Prediction of HPV and TP53 mutation status using radiomic features achieves an area under the receiver operating characteristics curve (AUC) of 0.71 and 0.641, respectively. Conclusion Our analysis suggests that radiomic features are associated with genomic characteristics in HNSCC and provides justification for continued development of radiomics as biomarkers for relevant genomic alterations in HNSCC.

Radiation Therapy is Independently Associated With Worse Survival After R0 Resection for Stage I-II Non–Small Cell Lung cancer: An Analysis of the National Cancer Data Base

Background nThe 1998 post-operative radiotherapy meta-analysis for lung cancer showed a survival detriment associated with radiation for stage I–II resected non-small cell lung cancer (NSCLC), but has been criticized for including antiquated radiation techniques. We analyzed the National Cancer Database (NCDB) to determine the impact of radiation after margin-negative (R0) resection for stage I–II NSCLC on survival.

SU-C-BRA-05: Delineating High-Dose Clinical Target Volumes for Head and Neck Tumors Using Machine Learning Algorithms

PURPOSEnTo develop and test population-based machine learning algorithms for delineating high-dose clinical target volumes (CTVs) in H&N tumors. Automating and standardizing the contouring of CTVs can reduce both physician contouring time and inter-physician variability, which is one of the largest sources of uncertainty in H&N radiotherapy.nnnMETHODSnTwenty-five node-negative patients treated with definitive radiotherapy were selected (6 right base of tongue, 11 left and 9 right tonsil). All patients had GTV and CTVs manually contoured by an experienced radiation oncologist prior to treatment. This contouring process, which is driven by anatomical, pathological, and patient specific information, typically results in non-uniform margin expansions about the GTV. Therefore, we tested two methods to delineate high-dose CTV given a manually-contoured GTV: (1) regression-support vector machines(SVM) and (2) classification-SVM. These models were trained and tested on each patient group using leave-one-out cross-validation. The volume difference(VD) and Dice similarity coefficient(DSC) between the manual and auto-contoured CTV were calculated to evaluate the results. Distances from GTV-to-CTV were computed about each patients GTV and these distances, in addition to distances from GTV to surrounding anatomy in the expansion direction, were utilized in the regression-SVM method. The classification-SVM method used categorical voxel-information (GTV, selected anatomical structures, else) from a 3×3×3cm3 ROI centered about the voxel to classify voxels as CTV.nnnRESULTSnVolumes for the auto-contoured CTVs ranged from 17.1 to 149.1cc and 17.4 to 151.9cc; the average(range) VD between manual and auto-contoured CTV were 0.93 (0.48-1.59) and 1.16(0.48-1.97); while average(range) DSC values were 0.75(0.59-0.88) and 0.74(0.59-0.81) for the regression-SVM and classification-SVM methods, respectively.nnnCONCLUSIONnWe developed two novel machine learning methods to delineate high-dose CTV for H&N patients. Both methods showed promising results that hint to a solution to the standardization of the contouring process of clinical target volumes. Varian Medical Systems grant.

WE-FG-202-12: Investigation of Longitudinal Salivary Gland DCE-MRI Changes

PURPOSEnTo determine the correlation between dose and changes through treatment in dynamic contrast enhanced (DCE) MRI voxel parameters (Ktrans, kep, Ve, and Vp) within salivary glands of head and neck oropharyngeal squamous cell carcinoma (HNSCC) patients.nnnMETHODSn17 HNSCC patients treated with definitive radiation therapy completed DCE-MRI scans on a 3T scanner at pre-treatment, mid-treatment, and post-treatment time points. Mid-treatment and post-treatment DCE images were deformably registered to pre-treatment DCE images (Velocity software package). Pharmacokinetic analysis of the DCE images used a modified Tofts model to produce parameter maps with an arterial input function selected from each patients perivertebral space on the image (NordicICE software package). In-house software was developed for voxel-by-voxel longitudinal analysis of the salivary glands within the registered images. The planning CT was rigidly registered to the pre-treatment DCE image to obtain dose values in each voxel. Voxels within the lower and upper dose quartiles for each gland were averaged for each patient, then an average of the patients means for the two quartiles were compared. Dose-relationships were also assessed by Spearman correlations between dose and voxel parameter changes for each patients gland.nnnRESULTSnChanges in parameters means between time points were observed, but inter-patient variability was high. Ve of the parotid was the only parameter that had a consistently significant longitudinal difference between dose quartiles. The highest Spearman correlation was Vp of the sublingual gland for the change in the pre-treatment to mid-treatment values with only a ρ=0.29.nnnCONCLUSIONnIn this preliminary study, there was large inter-patient variability in the changes of DCE voxel parameters with no clear relationship with dose. Additional patients may reduce the uncertainties and allow for the determination of the existence of parameter and dose relationships.

SU-G-IeP1-05: Diffusion Kurtosis Imaging for Oropharyngeal Cancer Detection

PURPOSEnDiffusion kurtosis imaging (DKI) is an emerging diffusion MRI technique in cancer diagnosis applications. The objective of this study is to compare DKI with conventional diffusion weighted imaging (DWI) for diagnosis of head and neck cancer.nnnMETHODSnFive male patients with histologically documented phase II/III squamous cell carcinoma of the oropharynx were included in this study. DKI (with 6 b-values of 0-2000 s/mm2) and conventional DWI data were acquired at a 3.0 T GE MRI scanner. Monoexponential (calculating apparent diffusion coefficient (ADC) using DWI data) and non-Gaussian kurtosis (calculating mean diffusion coefficient (MD) and mean kurtosis coefficient (MK) using DKI data) fits were performed on a voxel-by-voxel basis in selected regions of interest (primary tumor, metastatic nodes, contralateral region of tumor (tongue muscle), and submandibular glands). The non-parametric Wilcoxon test of ADC and DKI parameters between primary tumor and tongue muscle in contralateral region of tumor were calculated for comparison.nnnRESULTSnExamples of T1 post-contrast images and diffusion parametric maps (ADC, MD and MK) were illustrated in Figure 1 for an oropharynx cancer patient. MK parameters were significantly higher in the primary tumor than in the contralateral tongue muscle (0.89 ± 0.17 vs 0.68 ± 0.13, respectively; P <.05). MD and ADC in the primary tumor were significantly lower than those in the contralateral tongue muscle (1.20 ± 0.23 vs 2.57 ± 1.01; P <0.02 and 0.88 ± 0.22 vs 1.41 ± 0.18; P <0.01, respectively) (Table 1). Figure 2 showed statistical distribution of DWI and DKI parameters in selected regions of interest (primary tumor, metastatic nodes, tongue muscle, and submandibular glands).nnnCONCLUSIONnThe preliminary results of this study showed DKI could be a new option for increasing diagnostic confidence of oropharyngeal lesions. We continue to accrue study patients to evaluate the potential correlation between DKI parameters and tumor pathologic factors.

SU-E-QI-05: Denoising Intravoxel Incoherent Motion Magnetic Resonance Images Using Non-Local Mean Technique for Oropharyngeal Cancer Study

PURPOSEnIntravoxel incoherent motion (IVIM) magnetic resonance imaging (MRI) normally shows a low signal to noise ratio (SNR) due to the presence of noise which complicates and biases the estimation of quantitative diffusion parameters. In this study, a Non-local Means (NLM) approach was applied to remove the noise in oropharyngeal cancer IVIMMRI images.nnnMETHODSnEight male patients with squamous cell carcinoma of the oropharynx were included in this study under an approved IRB protocol. IVIM-MRI was carried out on a 3.0-T GE MRI. NLM denoising technique was performed using the MIPAV (V7, NIH). IVIM parameters (D, pure diffusion coefficient; f, perfusion fraction; D*, pseudodiffusion coefficient) were calculated on a pixel-by-pixel basis using a bi-exponential model implemented within ImageJ (V1.47, NIH) with ten b-values ranging from 0 - 800 s/mm2 . SNR with and without denoising processing was calculated and compared in tumor regions. The agreements of IVIM-MRI parameters estimated between with and without NLM processing in tumor region were assessed by Bland-Altman plots.nnnRESULTSnExamples of IVIM images (b=800 s/mm2 ) and parametric maps (D, D* and f) with and without NLM applied are illustrated in Fig. 1 for an oropharynx cancer patient. Results of SNR and IVIM diffusion parameters between two different processing, as mean and standard deviation on 8 patients, are reported in Table 1. Bland-Altman plots (Fig. 2) between the two approaches show better concordance for D (0.22% ± 1.42%) and f (0.56% ± 3.49%), whereas larger discrepancies were found for D* (0.86% ± 6.25%).nnnCONCLUSIONnNLM approach gives better SNR and quantitative data quality, also shows a good agreement between the IVIM parameters processed with and without NLM approaches. Therefore, NLM denoising technique can be applied to improve performance in terms of denoising quality and estimation of IVIM parameter as a post-processing step without increasing the scanning time.