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

[This corrects the article DOI: 10.3389/fmed.2023.1217037.].

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.

Automatic gross tumor segmentation of canine head and neck cancer using deep learning and cross-species transfer learning.

Background: Radiotherapy (RT) is increasingly being used on dogs with spontaneous head and neck cancer (HNC), which account for a large percentage of veterinary patients treated with RT. Accurate definition of the gross tumor volume (GTV) is a vital part of RT planning, ensuring adequate dose coverage of the tumor while limiting the radiation dose to surrounding tissues. Currently the GTV is contoured manually in medical images, which is a time-consuming and challenging task. Purpose: The purpose of this study was to evaluate the applicability of deep learning-based automatic segmentation of the GTV in canine patients with HNC. Materials and methods: Contrast-enhanced computed tomography (CT) images and corresponding manual GTV contours of 36 canine HNC patients and 197 human HNC patients were included. A 3D U-Net convolutional neural network (CNN) was trained to automatically segment the GTV in canine patients using two main approaches: (i) training models from scratch based solely on canine CT images, and (ii) using cross-species transfer learning where models were pretrained on CT images of human patients and then fine-tuned on CT images of canine patients. For the canine patients, automatic segmentations were assessed using the Dice similarity coefficient (Dice), the positive predictive value, the true positive rate, and surface distance metrics, calculated from a four-fold cross-validation strategy where each fold was used as a validation set and test set once in independent model runs. Results: CNN models trained from scratch on canine data or by using transfer learning obtained mean test set Dice scores of 0.55 and 0.52, respectively, indicating acceptable auto-segmentations, similar to the mean Dice performances reported for CT-based automatic segmentation in human HNC studies. Automatic segmentation of nasal cavity tumors appeared particularly promising, resulting in mean test set Dice scores of 0.69 for both approaches. Conclusion: In conclusion, deep learning-based automatic segmentation of the GTV using CNN models based on canine data only or a cross-species transfer learning approach shows promise for future application in RT of canine HNC patients.

Semi-automatic tumor segmentation of rectal cancer based on functional magnetic resonance imaging.

Background and purpose: Tumor delineation is required both for radiotherapy planning and quantitative imaging biomarker purposes. It is a manual, time- and labor-intensive process prone to inter- and intraobserver variations. Semi or fully automatic segmentation could provide better efficiency and consistency. This study aimed to investigate the influence of including and combining functional with anatomical magnetic resonance imaging (MRI) sequences on the quality of automatic segmentations. Materials and methods: T2-weighted (T2w), diffusion weighted, multi-echo T2*-weighted, and contrast enhanced dynamic multi-echo (DME) MR images of eighty-one patients with rectal cancer were used in the analysis. Four classical machine learning algorithms; adaptive boosting (ADA), linear and quadratic discriminant analysis and support vector machines, were trained for automatic segmentation of tumor and normal tissue using different combinations of the MR images as input, followed by semi-automatic morphological post-processing. Manual delineations from two experts served as ground truth. The Sørensen-Dice similarity coefficient (DICE) and mean symmetric surface distance (MSD) were used as performance metric in leave-one-out cross validation. Results: Using T2w images alone, ADA outperformed the other algorithms, yielding a median per patient DICE of 0.67 and MSD of 3.6 mm. The performance improved when functional images were added and was highest for models based on either T2w and DME images (DICE: 0.72, MSD: 2.7 mm) or all four MRI sequences (DICE: 0.72, MSD: 2.5 mm). Conclusion: Machine learning models using functional MRI, in particular DME, have the potential to improve automatic segmentation of rectal cancer relative to models using T2w MRI alone.

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

BACKGROUND: 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. MATERIAL AND METHODS: 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. RESULTS: 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. CONCLUSION: 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.

MRI-based automatic segmentation of rectal cancer using 2D U-Net on two independent cohorts.

BACKGROUND: Tumor delineation is time- and labor-intensive and prone to inter- and intraobserver variations. Magnetic resonance imaging (MRI) provides good soft tissue contrast, and functional MRI captures tissue properties that may be valuable for tumor delineation. We explored MRI-based automatic segmentation of rectal cancer using a deep learning (DL) approach. We first investigated potential improvements when including both anatomical T2-weighted (T2w) MRI and diffusion-weighted MR images (DWI). Secondly, we investigated generalizability by including a second, independent cohort. MATERIAL AND METHODS: Two cohorts of rectal cancer patients (C1 and C2) from different hospitals with 109 and 83 patients, respectively, were subject to 1.5 T MRI at baseline. T2w images were acquired for both cohorts and DWI (b-value of 500 s/mm2) for patients in C1. Tumors were manually delineated by three radiologists (two in C1, one in C2). A 2D U-Net was trained on T2w and T2w + DWI. Optimal parameters for image pre-processing and training were identified on C1 using five-fold cross-validation and patient Dice similarity coefficient (DSCp) as performance measure. The optimized models were evaluated on a C1 hold-out test set and the generalizability was investigated using C2. RESULTS: For cohort C1, the T2w model resulted in a median DSCp of 0.77 on the test set. Inclusion of DWI did not further improve the performance (DSCp 0.76). The T2w-based model trained on C1 and applied to C2 achieved a DSCp of 0.59. CONCLUSION: T2w MR-based DL models demonstrated high performance for automatic tumor segmentation, at the same level as published data on interobserver variation. DWI did not improve results further. Using DL models on unseen cohorts requires caution, and one cannot expect the same performance.

A comparison of methods for fully automatic segmentation of tumors and involved nodes in PET/CT of head and neck cancers.

Target volume delineation is a vital but time-consuming and challenging part of radiotherapy, where the goal is to deliver sufficient dose to the target while reducing risks of side effects. For head and neck cancer (HNC) this is complicated by the complex anatomy of the head and neck region and the proximity of target volumes to organs at risk. The purpose of this study was to compare and evaluate conventional PET thresholding methods, six classical machine learning algorithms and a 2D U-Net convolutional neural network (CNN) for automatic gross tumor volume (GTV) segmentation of HNC in PET/CT images. For the latter two approaches the impact of single versus multimodality input on segmentation quality was also assessed. 197 patients were included in the study. The cohort was split into training and test sets (157 and 40 patients, respectively). Five-fold cross-validation was used on the training set for model comparison and selection. Manual GTV delineations represented the ground truth. Tresholding, classical machine learning and CNN segmentation models were ranked separately according to the cross-validation Sørensen-Dice similarity coefficient (Dice). PET thresholding gave a maximum mean Dice of 0.62, whereas classical machine learning resulted in maximum mean Dice scores of 0.24 (CT) and 0.66 (PET; PET/CT). CNN models obtained maximum mean Dice scores of 0.66 (CT), 0.68 (PET) and 0.74 (PET/CT). The difference in cross-validation Dice between multimodality PET/CT and single modality CNN models was significant (p ≤ 0.0001). The top-ranked PET/CT-based CNN model outperformed the best-performing thresholding and classical machine learning models, giving significantly better segmentations in terms of cross-validation and test set Dice, true positive rate, positive predictive value and surface distance-based metrics (p ≤ 0.0001). Thus, deep learning based on multimodality PET/CT input resulted in superior target coverage and less inclusion of surrounding normal tissue.

Deep learning-based auto-delineation of gross tumour volumes and involved nodes in PET/CT images of head and neck cancer patients.

PURPOSE: Identification and delineation of the gross tumour and malignant nodal volume (GTV) in medical images are vital in radiotherapy. We assessed the applicability of convolutional neural networks (CNNs) for fully automatic delineation of the GTV from FDG-PET/CT images of patients with head and neck cancer (HNC). CNN models were compared to manual GTV delineations made by experienced specialists. New structure-based performance metrics were introduced to enable in-depth assessment of auto-delineation of multiple malignant structures in individual patients. METHODS: U-Net CNN models were trained and evaluated on images and manual GTV delineations from 197 HNC patients. The dataset was split into training, validation and test cohorts (n= 142, n = 15 and n = 40, respectively). The Dice score, surface distance metrics and the new structure-based metrics were used for model evaluation. Additionally, auto-delineations were manually assessed by an oncologist for 15 randomly selected patients in the test cohort. RESULTS: The mean Dice scores of the auto-delineations were 55%, 69% and 71% for the CT-based, PET-based and PET/CT-based CNN models, respectively. The PET signal was essential for delineating all structures. Models based on PET/CT images identified 86% of the true GTV structures, whereas models built solely on CT images identified only 55% of the true structures. The oncologist reported very high-quality auto-delineations for 14 out of the 15 randomly selected patients. CONCLUSIONS: CNNs provided high-quality auto-delineations for HNC using multimodality PET/CT. The introduced structure-wise evaluation metrics provided valuable information on CNN model strengths and weaknesses for multi-structure auto-delineation.

In Quest of the Alanine R3 Radical: Multivariate EPR Spectral Analyses of X-Irradiated Alanine in the Solid State.

The amino acid l-α-alanine is the most commonly used material for solid-state electron paramagnetic resonance (EPR) dosimetry, due to the formation of highly stable radicals upon irradiation, with yields proportional to the radiation dose. Two major alanine radical components designated R1 and R2 have previously been uniquely characterized from EPR and electron-nuclear double resonance (ENDOR) studies as well as from quantum chemical calculations. There is also convincing experimental evidence of a third minor radical component R3, and a tentative radical structure has been suggested, even though no well-defined spectral signature has been observed experimentally. In the present study, temperature dependent EPR spectra of X-ray irradiated polycrystalline alanine were analyzed using five multivariate methods in further attempts to understand the composite nature of the alanine dosimeter EPR spectrum. Principal component analysis (PCA), maximum likelihood common factor analysis (MLCFA), independent component analysis (ICA), self-modeling mixture analysis (SMA), and multivariate curve resolution (MCR) were used to extract pure radical spectra and their fractional contributions from the experimental EPR spectra. All methods yielded spectral estimates resembling the established R1 spectrum. Furthermore, SMA and MCR consistently predicted both the established R2 spectrum and the shape of the R3 spectrum. The predicted shape of the R3 spectrum corresponded well with the proposed tentative spectrum derived from spectrum simulations. Thus, results from two independent multivariate data analysis techniques strongly support the previous evidence that three radicals are indeed present in irradiated alanine samples.

Autodelineation of cervical cancers using multiparametric magnetic resonance imaging and machine learning.

BACKGROUND: Tumour delineation is a challenging, time-consuming and complex part of radiotherapy planning. In this study, an automatic method for delineating locally advanced cervical cancers was developed using a machine learning approach. MATERIALS AND METHODS: A method for tumour segmentation based on image voxel classification using Fisher?s Linear Discriminant Analysis (LDA) was developed. This was applied to magnetic resonance (MR) images of 78 patients with locally advanced cervical cancer. The segmentation was based on multiparametric MRI consisting of T2- weighted (T2w), T1-weighted (T1w) and dynamic contrast-enhanced (DCE) sequences, and included intensity and spatial information from the images. The model was trained and assessed using delineations made by two radiologists. RESULTS: Segmentation based on T2w or T1w images resulted in mean sensitivity and specificity of 94% and 52%, respectively. Including DCE-MR images improved the segmentation model?s performance significantly, giving mean sensitivity and specificity of 85?93%. Comparisons with radiologists? tumour delineations gave Dice similarity coefficients of up to 0.44. CONCLUSION: Voxel classification using a machine learning approach is a flexible and fully automatic method for tumour delineation. Combining all relevant MR image series resulted in high sensitivity and specificity. Moreover, the presented method can be extended to include additional imaging modalities.

Daily changes in VPD during leaf development in high air humidity increase the stomatal responsiveness to darkness and dry air.

Previous studies have shown that plants developed under high relative air humidity (RH>85%) develop malfunctioning stomata and therefor have increased transpiration and reduced desiccation tolerance when transferred to lower RH conditions and darkness. In this study, plants developed at high RH were exposed to daily VPD fluctuations created by changes in temperature and/or RH to evaluate the potential improvements in stomatal functioning. Daily periods with an 11°C temperature increase and consequently a VPD increase (vpd: 0.36-2.37KPa) reduced the stomatal apertures and improved the stomatal functionality and desiccation tolerance of the rosette plant Arabidopsis thaliana. A similar experiment was performed with only a 4°C temperature increase and/or a RH decrease on tomato. The results showed that a daily change in VPD (vpd: 0.36-1.43KPa) also resulted in improved stomatal responsiveness and decreased water usage during growth. In tomato, the most effective treatment to increase the stomatal responsiveness to darkness as a signal for closure was daily changes in RH without a temperature increase.

Dynamic 2-Deoxy-2-[18F]Fluoro-D-Glucose Positron Emission Tomography for Chemotherapy Response Monitoring of Breast Cancer Xenografts.

PURPOSE: Non-invasive response monitoring can potentially be used to guide therapy selection for breast cancer patients. We employed dynamic 2-deoxy-2-[18F]fluoro-D-glucose positron emission tomography ([18F]FDG PET) to evaluate changes in three breast cancer xenograft lines in mice following three chemotherapy regimens. PROCEDURES: Sixty-six athymic nude mice bearing bilateral breast cancer xenografts (two basal-like and one luminal-like subtype) underwent three 60 min [18F]FDG PET scans. Scans were performed prior to and 3 and 10 days after treatment with doxorubicin, paclitaxel, or carboplatin. Tumor growth was monitored in parallel. A pharmacokinetic compartmental model was fitted to the tumor uptake curves, providing estimates of transfer rates between the vascular, non-metabolized, and metabolized compartments. Early and late standardized uptake values (SUVE and SUVL, respectively); the rate constants k 1, k 2, and k 3, and the intravascular fraction v B were estimated. Changes in tumor volume were used as a response measure. Multivariate partial least-squares regression (PLSR) was used to assess if PET parameters could model tumor response and to identify PET parameters with the largest impact on response. RESULTS: Treatment responders had significantly larger perfusion-related parameters (k 1 and k 2) and lower metabolism-related parameter (k 3) than non-responders 10 days after the start of treatment. These findings were further supported by the PLSR analysis, which showed that k 1 and k 2 at day 10 and changes in k 3 explained most of the variability in response to therapy, whereas SUVL and particularly SUVE were of lesser importance. CONCLUSIONS: Overall, rate parameters related to both tumor perfusion and metabolism were associated with tumor response. Conventional metrics of [18F]FDG uptake such as SUVE and SUVL apparently had little relation to tumor response, thus necessitating full dynamic scanning and pharmacokinetic analysis for optimal evaluation of chemotherapy-induced changes in breast cancers.

Cluster analysis of dynamic contrast enhanced MRI reveals tumor subregions related to locoregional relapse for cervical cancer patients.

BACKGROUND: Solid tumors are known to be spatially heterogeneous. Detection of treatment-resistant tumor regions can improve clinical outcome, by enabling implementation of strategies targeting such regions. In this study, K-means clustering was used to group voxels in dynamic contrast enhanced magnetic resonance images (DCE-MRI) of cervical cancers. The aim was to identify clusters reflecting treatment resistance that could be used for targeted radiotherapy with a dose-painting approach. MATERIAL AND METHODS: Eighty-one patients with locally advanced cervical cancer underwent DCE-MRI prior to chemoradiotherapy. The resulting image time series were fitted to two pharmacokinetic models, the Tofts model (yielding parameters Ktrans and νe) and the Brix model (ABrix, kep and kel). K-means clustering was used to group similar voxels based on either the pharmacokinetic parameter maps or the relative signal increase (RSI) time series. The associations between voxel clusters and treatment outcome (measured as locoregional control) were evaluated using the volume fraction or the spatial distribution of each cluster. RESULTS: One voxel cluster based on the RSI time series was significantly related to locoregional control (adjusted p-value 0.048). This cluster consisted of low-enhancing voxels. We found that tumors with poor prognosis had this RSI-based cluster gathered into few patches, making this cluster a potential candidate for targeted radiotherapy. None of the voxels clusters based on Tofts or Brix parameter maps were significantly related to treatment outcome. CONCLUSION: We identified one group of tumor voxels significantly associated with locoregional relapse that could potentially be used for dose painting. This tumor voxel cluster was identified using the raw MRI time series rather than the pharmacokinetic maps.

Classification of dynamic contrast enhanced MR images of cervical cancers using texture analysis and support vector machines.

Dynamic contrast enhanced MRI (DCE-MRI) provides insight into the vascular properties of tissue. Pharmacokinetic models may be fitted to DCE-MRI uptake patterns, enabling biologically relevant interpretations. The aim of our study was to determine whether treatment outcome for 81 patients with locally advanced cervical cancer could be predicted from parameters of the Brix pharmacokinetic model derived from pre-chemoradiotherapy DCE-MRI. First-order statistical features of the Brix parameters were used. In addition, texture analysis of Brix parameter maps was done by constructing gray level co-occurrence matrices (GLCM) from the maps. Clinical factors and first- and second-order features were used as explanatory variables for support vector machine (SVM) classification, with treatment outcome as response. Classification models were validated using leave-one-out cross-model validation. A random value permutation test was used to evaluate model significance. Features derived from first-order statistics could not discriminate between cured and relapsed patients (specificity 0%-20%, p-values close to unity). However, second-order GLCM features could significantly predict treatment outcome with accuracies (~70%) similar to the clinical factors tumor volume and stage (69%). The results indicate that the spatial relations within the tumor, quantified by texture features, were more suitable for outcome prediction than first-order features.

Visible foliar injury and infrared imaging show that daylength affects short-term recovery after ozone stress in Trifolium subterraneum.

Tropospheric ozone is a major air pollutant affecting plants worldwide. Plants in northern regions can display more ozone injury than plants at lower latitudes despite lower ozone levels. Larger ozone influx and shorter nights have been suggested as possible causes. However, the effects of the dim light present during northern summer nights have not been investigated. Young Trifolium subterraneum plants kept in environmentally controlled growth rooms under long day (10 h bright light, 14 h dim light) or short day (10 h bright light, 14 h darkness) conditions were exposed to 6 h of 70 ppb ozone during daytime for three consecutive days. Leaves were visually inspected and imaged in vivo using thermal imaging before and after the daily exposure. In long-day-treated plants, visible foliar injury within 1 week after exposure was more severe. Multivariate statistical analyses showed that the leaves of ozone-exposed long-day-treated plants were also warmer with more homogeneous temperature distributions than exposed short day and control plants, suggesting reduced transpiration. Temperature disruptions were not restricted to areas displaying visible damage and occurred even in leaves with only slight visible injury. Ozone did not affect the leaf temperature of short-day-treated plants. As all factors influencing ozone influx were the same for long- and short-day-treated plants, only the dim nocturnal light could account for the different ozone sensitivities. Thus, the twilight summer nights at high latitudes may have a negative effect on repair and defence processes activated after ozone exposure, thereby enhancing sensitivity.

Effects of the Nordic photoperiod on ozone sensitivity and repair in different clover species studied using infrared imaging.

Plants in Nordic regions can be more ozone sensitive at a given ozone concentration than plants at lower latitudes. A recent study shows that the Nordic summer photoperiod, particularly the dim nighttime light, can increase visible foliar injury and alter leaf transpiration in subterranean clover. Effects of photoperiod on the ozone sensitivity of white and red clover cultivars adapted to Nordic conditions were investigated. Although ozone induced visible foliar injury and leaf transpirational changes in white clover, the effects were independent of photoperiod. In red clover, ozone combined with a long photoperiod with dim nights (8 nights) induced more severe visible injuries than with a short photoperiod. Furthermore, transpirational changes in red clover depended on photoperiod. Thus, a long photoperiod can increase ozone sensitivity differently in clover cultivars with different degrees of adaptation to northern conditions, suggesting that ozone indices used in risk analysis should take this effect into account.