Espen Rusten | Theragnostic Imaging

Deep learning-based automatic delineation of anal cancer gross tumour volume: a multimodality comparison of CT, PET and MRI

Sat, 01 Jan 2022 00:00:00 +0000

Accurate target volume delineation is a prerequisite for high-precision radiotherapy. However, manual delineation is resource-demanding and prone to interobserver variation. An automatic delineation approach could potentially save time and increase delineation consistency. In this study, the applicability of deep learning for fully automatic delineation of the gross tumour volume (GTV) in patients with anal squamous cell carcinoma (ASCC) was evaluated for the first time. An extensive comparison of the effects single modality and multimodality combinations of computed tomography (CT), positron emission tomography (PET), and magnetic resonance imaging (MRI) have on automatic delineation quality was conducted. 18F-fluorodeoxyglucose PET/CT and contrast-enhanced CT (ceCT) images were collected for 86 patients with ASCC. A subset of 36 patients also underwent a study-specific 3T MRI examination including T2- and diffusion-weighted imaging. The resulting two datasets were analysed separately. A two-dimensional U-Net convolutional neural network (CNN) was trained to delineate the GTV in axial image slices based on single or multimodality image input. Manual GTV delineations constituted the ground truth for CNN model training and evaluation. Models were evaluated using the Dice similarity coefficient (Dice) and surface distance metrics computed from five-fold cross-validation. CNN-generated automatic delineations demonstrated good agreement with the ground truth, resulting in mean Dice scores of 0.65-0.76 and 0.74-0.83 for the 86 and 36-patient datasets, respectively. For both datasets, the highest mean Dice scores were obtained using a multimodal combination of PET and ceCT (0.76-0.83). However, models based on single modality ceCT performed comparably well (0.74-0.81). T2W-only models performed acceptably but were somewhat inferior to the PET/ceCT and ceCT-based models. CNNs provided high-quality automatic GTV delineations for both single and multimodality image input, indicating that deep learning may prove a versatile tool for target volume delineation in future patients with ASCC.

Anal cancer chemoradiotherapy outcome prediction using 18F-fluorodeoxyglucose positron emission tomography and clinicopathological factors

Wed, 01 May 2019 00:00:00 +0000

To assess the role of [¹⁸F]fluorodeoxyglucose (FDG) positron emission tomography (PET), obtained before and during chemoradiotherapy, in predicting locoregional failure relative to clinicopathological factors for patients with anal cancer. 93 patients with anal squamous cell carcinoma treated with chemoradiotherapy were included in a prospective observational study (NCT01937780). FDG-PET/CT was performed for all patients before treatment, and for a subgroup (n = 39) also 2 weeks into treatment. FDG-PET was evaluated with standardized uptake values (SUV_{max/peak/mean}), metabolic tumor volume (MTV), total lesion glycolysis (TLG), and a proposed Z-normalized combination of MTV and SUV_peak (ZMP). The objective was to predict locoregional failure using FDG-PET, tumor and lymph node stage, gross tumor volume (GTV) and human papilloma virus (HPV) status in univariate and bivariate Cox regression analysis. N3 lymph node stage, HPV negative tumor, GTV, MTV, TLG and ZMP were in univariate analysis significant predictors of locoregional failure (p < 0.01), while SUV_{max/peak/mean} were not (p > 0.2). In bivariate analysis HPV status was the most independent predictor in combinations with N3 stage, ZMP, TLG, and MTV (p < 0.02). The FDG-PET parameters at 2 weeks into radiotherapy decreased by 30-40 % of the initial values, but neither absolute nor relative decrease improved the prediction models. Pre-treatment PET parameters are predictive of chemoradiotherapy outcome in anal cancer, although HPV negativity and N3 stage are the strongest single predictors. Predictions can be improved by combining HPV with PET parameters such as MTV, TLG or ZMP. PET 2 weeks into treatment does not provide added predictive value. Pre-treatment PET parameters of anal cancer showed a predictive role independent of clinicopathological factors. Although the PET parameters show substantial reduction from pre- to mid-treatment, the changes were not predictive of chemoradiotherapy outcome.

Target volume delineation of anal cancer based on magnetic resonance imaging or positron emission tomography

Wed, 06 Sep 2017 00:00:00 +0000

To compare target volume delineation of anal cancer using positron emission tomography (PET) and magnetic resonance imaging (MRI) with respect to inter-observer and inter-modality variability. Nineteen patients with anal cancer undergoing chemoradiotherapy were prospectively included. Planning computed tomography (CT) images were co-registered with 18F-fluorodexocyglucose (FDG) PET/CT images and T2 and diffusion weighted (DW) MR images. Three oncologists delineated the Gross Tumor Volume (GTV) according to national guidelines and the visible tumor tissue (GTV_T). MRI and PET based delineations were evaluated by absolute volumes and Dice similarity coefficients. The median volume of the GTVs was 27 and 31 cm³ for PET and MRI, respectively, while it was 6 and 11 cm³ for GTV_T. Both GTV and GTV_T volumes were highly correlated between delineators (r = 0.90 and r = 0.96, respectively). The median Dice similarity coefficient was 0.75 when comparing the GTVs based on PET/CT (GTV_PET) with the GTVs based on MRI and CT (GTV_MRI). The median Dice coefficient was 0.56 when comparing the visible tumor volume evaluated by PET (GTV_{T_PET}) with the same volume evaluated by MRI (GTV_{T_MRI}). Margins of 1-2 mm in the axial plane and 7-8 mm in superoinferior direction were required for coverage of the individual observer’s GTVs. The rather good agreement between PET- and MRI-based GTVs indicates that either modality may be used for standard target delineation of anal cancer. However, larger deviations were found for GTV_T, which may impact future tumor boost strategies.