Semiautomated segmentation of head and neck cancers in 18F-FDG PET scans: A just-enough-interaction approach

Prediction of EGFR-TP53 genes co-mutations in patients with lung adenocarcinoma (LUAD) by 18F-FDG PET/CT radiomics

PURPOSE

This retrospective study was undertaken to assess the predictive efficacy of 18F-FDG PET/CT -derived radiomic features concerning the co-mutation status of epidermal growth factor receptor (EGFR) and TP53 in LUAD.

METHODS

A cohort of 150 LUAD patients underwent pretreatment 18F-FDG PET/CT scans with known mutation status of EGFR and TP53 were collected. The feature extraction based on their PET/CT images utilized the Pyradiomics package based on the 3D Slicer. The optimal radiomic features were selected through correlation analysis and the Gradient Boosting Decision Tree (GBDT) algorithm, followed by the construction of the radiomic model. The clinical model incorporated meaningful clinical variables, whereas the complex model integrated both the radiomic and clinical models. The area under the receiver operating characteristic curve (AUC) facilitated the comparison of prediction performance across the three models. The DCA gauged the clinical utility of these models.

RESULTS

The patient cohort was randomly allocated into a training set (n = 105) and a validation set (n = 45) in a 7:3 ratio. Eleven PET and eleven CT optimal radiomic features were selected to construct the radiomic model. The model showed a good ability to discriminate the co-occurrence of EGFR and TP53, with AUC equal to 0.850 in the training set, and 0.748 in the validation set, compared with 0.750 and 0.626 for the clinical model. The complex model exhibited the highest AUC values, with 0.880 and 0.794 in both sets, but there were no significant differences compared to the radiomic model. The DCA revealed favorable clinical value.

CONCLUSION

Deep learning-based image analysis predicts PD-L1 status from 18F-FDG PET/CT images in non-small-cell lung cancer

Background

There is still a lack of clinically validated biomarkers to screen lung cancer patients suitable for programmed dead cell-1 (PD-1)/programmed dead cell receptor-1 (PD-L1) immunotherapy. Detection of PD-L1 expression is invasively operated, and some PD-L1-negative patients can also benefit from immunotherapy; thus, the joint modeling of both deep learning images and clinical features was used to improve the prediction performance of PD-L1 expression in non-small cell lung cancer (NSCLC).

Methods

Retrospective collection of 101 patients diagnosed with pathology in our hospital who underwent 18F FDG PET/CT scans, with lung cancer tissue Tumor Propulsion Score (TPS) ≥1% as a positive expression. Lesions were extracted after preprocessing PET/CT images, and using deep learning 3D DenseNet121 to learn lesions in PET, CT, and PET/CT images, 1,024 fully connected features were extracted; clinical features (age, gender, smoking/no smoking history, lesion diameter, lesion volume, maximum standard uptake value of lesions [SUVmax], mean standard uptake value of lesions [SUVmean], total lesion glycolysis [TLG]) were combined for joint modeling based on the structured data Category Embedding Model.

Results

Area under a receiver operating characteristic (ROC) curve (AUC) and accuracy of predicting PD-L1 positive for PET, CT, and PET/CT test groups were 0.814 ± 0.0152, 0.7212 ± 0.0861, and 0.90 ± 0.0605, 0.806 ± 0.023, 0.70 ± 0.074, and 0.950 ± 0.0250, respectively. After joint clinical feature modeling, the AUC and accuracy of predicting PD-L1 positive for PET/CT were 0.96 ± 0.00905 and 0.950 ± 0.0250, respectively.

Conclusion

This study combines the features of 18F-FDG PET/CT images with clinical features using deep learning to predict the expression of PD-L1 in NSCLC, suggesting that 18F-FDG PET/CT images can be conducted as biomarkers for PD-L1 expression.

Diagnostic efficacy of [68Ga]Ga-DOTA-GPFAPI-04 in patients with solid tumors in a head-to-head comparison with [18F]F-FDG: results from a prospective clinical study

PURPOSE

To identify the biodistribution and diagnostic performance of a novel fibroblast activation protein (FAP) targeted positron emission tomography (PET) tracer, [68Ga]Ga-DOTA-GPFAPI-04, in patients with solid tumors in a head-to-head comparison with [18F]F-FDG.

METHODS

Twenty-six patients histologically proven with cancers of nasopharyngeal (n = 5), esophagus (n = 5), gastro-esophagus (n = 1), stomach (n = 7), liver (n = 3), and colorectum (n = 5) were recruited for [68Ga]Ga-DOTA-GPFAPI-04 and [18F]F-FDG PET/CT scans on consecutive days. The primary endpoint was the diagnostic efficacy, with the histological diagnosis and the follow-up results selected as the gold standard. The secondary endpoint was the background uptake pattern. Two experienced nuclear medicine physicians who were blinded to the gold standard results while having essential awareness of the clinical context reviewed the images and labeled lesions by consensus for subsequent software-assisted lesion segmentation. Additionally, background organs were automatically segmented, assisted by artificial intelligence. Volume, mean, and maximum standard uptake values (SUVmean and SUVmax) of all segmentations were recorded. P < 0.05 was deemed as statistically significant.

RESULTS

Significant glandular uptake of [68Ga]Ga-DOTA-GPFAPI-04 was detected in the thyroid, pancreas, and submandibular glands, while moderate uptake was observed in the parotid glands. The myocardium and myometrium exhibited 2-3 times higher uptake of the radiotracer than that of the background levels in blood and liver. A total of 349 targeted lesions, consisting of 324 malignancies and 25 benign lesions, were segmented. [68Ga]Ga-DOTA-GPFAPI-04 is more sensitive than [18F]F-FDG, especially for abdominopelvic dissemination (1.000 vs. 0.475, P < 0.001). Interestingly, [18F]F-FDG demonstrated higher sensitivity for lung metastasis compared to [68Ga]Ga-DOTA-GPFAPI-04 (0.845 vs. 0.682, P = 0.003). The high glandular uptake made it difficult to delineate lesions in close proximity and masked two metastatic lesions in these organs.

CONCLUSION

Despite prominent glandular uptake, [68Ga]Ga-DOTA-GPFAPI-04 demonstrates favorable diagnostic performance. It is a promising probe scaffold for further development of FAP-targeted tumor theranostic agents.

Peri- and intra-nodular radiomic features based on 18F-FDG PET/CT to distinguish lung adenocarcinomas from pulmonary granulomas

Objective

To compare the effectiveness of radiomic features based on 18F-FDG PET/CT images within (intranodular) and around (perinodular) lung nodules/masses in distinguishing between lung adenocarcinoma and pulmonary granulomas.

Methods

For this retrospective study, 18F-FDG PET/CT images were collected for 228 patients. Patients diagnosed with lung adenocarcinoma (n = 156) or granulomas (n = 72) were randomly assigned to a training (n = 159) and validation (n = 69) groups. The volume of interest (VOI) of intranodular, perinodular (1-5 voxels, termed Lesion_margin1 to Lesion_margin5) and total area (intra- plus perinodular region, termed Lesion_total1 to Lesion_total5) on PET/CT images were delineated using PETtumor and Marge tool of segmentation editor. A total of 1,037 radiomic features were extracted separately from PET and CT images, and the optimal features were selected to develop radiomic models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC).

Results

Good and acceptable performance was, respectively, observed in the training (AUC = 0.868, p < 0.001) and validation (AUC = 0.715, p = 0.004) sets for the intranodular radiomic model. Among the perinodular models, the Lesion_margin2 model demonstrated the highest AUC in both sets (0.883 and 0.616, p < 0.001 and p = 0.122). Similarly, in terms of total models, Lesion_total2 model was found to outperform others in the training (AUC = 0.879, p < 0.001) and validation (AUC = 0.742, p = 0.001) sets, slightly surpassing the intranodular model.

Conclusion

When intra- and perinodular radiomic features extracted from the immediate vicinity of the nodule/mass up to 2 voxels distance on 18F-FDG PET/CT imaging are combined, improved differential diagnostic performance in distinguishing between lung adenocarcinomas and granulomas is achieved compared to the intra- and perinodular radiomic features alone.

Interim 18F-FDG-PET based response-adaptive dose escalation of proton therapy for head and neck cancer: a treatment planning feasibility study

BACKGROUND

Image-driven dose escalation to tumor subvolumes has been proposed to improve treatment outcome in head and neck cancer (HNC). We used 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) acquired at baseline and into treatment (interim) to identify biologic target volumes (BTVs). We assessed the feasibility of interim dose escalation to the BTV with proton therapy by simulating the effects to organs at risk (OARs).

METHODS

We used the semiautomated just-enough-interaction (JEI) method to identify BTVs in 18F-FDG-PET images from nine HNC patients. Between baseline and interim FDG-PET, patients received photon radiotherapy. BTV was identified assuming that high standardized uptake value (SUV) at interim reflected tumor radioresistance. Using Eclipse (Varian Medical Systems), we simulated a 10% (6.8 Gy(RBE1.1)) and 20% (13.6 Gy(RBE1.1)) dose escalation to the BTV with protons and compared results with proton plans without dose escalation.

RESULTS

At interim 18F-FDG-PET, radiotherapy resulted in reduced SUV compared to baseline. However, spatial overlap between high-SUV regions at baseline and interim allowed for BTV identification. Proton therapy planning demonstrated that dose escalation to the BTV was feasible, and except for some 20% dose escalation plans, OAR doses did not significantly increase.

CONCLUSION

Our in silico analysis demonstrated the potential for interim 18F-FDG-PET response-adaptive dose escalation to the BTV with proton therapy. This approach may give more efficient treatment to HNC with radioresistant tumor subvolumes without increasing normal tissue toxicity. Studies in larger cohorts are required to determine the full potential for interim 18F-FDG-PET-guided dose escalation of proton therapy in HNC.

Evaluation of the performance of both machine learning models using PET and CT radiomics for predicting recurrence following lung stereotactic body radiation therapy: A single-institutional study

PURPOSE

Predicting recurrence following stereotactic body radiotherapy (SBRT) for non-small cell lung cancer provides important information for the feasibility of the individualized radiotherapy and allows to select the appropriate treatment strategy based on the risk of recurrence. In this study, we evaluated the performance of both machine learning models using positron emission tomography (PET) and computed tomography (CT) radiomic features for predicting recurrence after SBRT.

METHODS

Planning CT and PET images of 82 non-small cell lung cancer patients who performed SBRT at our hospital were used. First, tumors were delineated on each CT and PET of each patient, and 111 unique radiomic features were extracted, respectively. Next, the 10 features were selected using three different feature selection algorithms, respectively. Recurrence prediction models based on the selected features and four different machine learning algorithms were developed, respectively. Finally, we compared the predictive performance of each model for each recurrence pattern using the mean area under the curve (AUC) calculated following the 0.632+ bootstrap method.

RESULTS

The highest performance for local recurrence, regional lymph node metastasis, and distant metastasis were observed in models using Support vector machine with PET features (mean AUC = 0.646), Naive Bayes with PET features (mean AUC = 0.611), and Support vector machine with CT features (mean AUC = 0.645), respectively.

CONCLUSIONS

We comprehensively evaluated the performance of prediction model developed for recurrence following SBRT. The model in this study would provide information to predict the recurrence pattern and assist in making treatment strategies.

Comparison of semi-automatic and manual segmentation methods for tumor delineation on head and neck squamous cell carcinoma (HNSCC) positron emission tomography (PET) images

Objective. Accurate and reproducible tumor delineation on positron emission tomography (PET) images is required to validate predictive and prognostic models based on PET radiomic features. Manual segmentation of tumors is time-consuming whereas semi-automatic methods are easily implementable and inexpensive. This study assessed the reliability of semi-automatic segmentation methods over manual segmentation for tumor delineation in head and neck squamous cell carcinoma (HNSCC) PET images.Approach. We employed manual and six semi-automatic segmentation methods (just enough interaction (JEI), watershed, grow from seeds (GfS), flood filling (FF), 30% SUVmax and 40%SUVmax threshold) using 3D slicer software to extract 128 radiomic features from FDG-PET images of 100 HNSCC patients independently by three operators. We assessed the distributional properties of all features and considered 92 log-transformed features for subsequent analysis. For each paired comparison of a feature, we fitted a separate linear mixed effect model using the method (two levels; manual versus one semi-automatic method) as a fixed effect and the subject and the operator as the random effects. We estimated different statistics-the intraclass correlation coefficient agreement (aICC), limits of agreement (LoA), total deviation index (TDI), coverage probability (CP) and coefficient of individual agreement (CIA)-to evaluate the agreement between the manual and semi-automatic methods.Main results. Accounting for all statistics across 92 features, the JEI method consistently demonstrated acceptable agreement with the manual method, with median values of aICC = 0.86, TDI = 0.94, CP = 0.66, and CIA = 0.91.Significance. This study demonstrated that JEI method is a reliable semi-automatic method for tumor delineation on HNSCC PET images.

Transfer learning-based PET/CT three-dimensional convolutional neural network fusion of image and clinical information for prediction of EGFR mutation in lung adenocarcinoma

BACKGROUND

To introduce a three-dimensional convolutional neural network (3D CNN) leveraging transfer learning for fusing PET/CT images and clinical data to predict EGFR mutation status in lung adenocarcinoma (LADC).

METHODS

Retrospective data from 516 LADC patients, encompassing preoperative PET/CT images, clinical information, and EGFR mutation status, were divided into training (n = 404) and test sets (n = 112). Several deep learning models were developed utilizing transfer learning, involving CT-only and PET-only models. A dual-stream model fusing PET and CT and a three-stream transfer learning model (TS_TL) integrating clinical data were also developed. Image preprocessing includes semi-automatic segmentation, resampling, and image cropping. Considering the impact of class imbalance, the performance of the model was evaluated using ROC curves and AUC values.

RESULTS

TS_TL model demonstrated promising performance in predicting the EGFR mutation status, with an AUC of 0.883 (95%CI = 0.849-0.917) in the training set and 0.730 (95%CI = 0.629-0.830) in the independent test set. Particularly in advanced LADC, the model achieved an AUC of 0.871 (95%CI = 0.823-0.919) in the training set and 0.760 (95%CI = 0.638-0.881) in the test set. The model identified distinct activation areas in solid or subsolid lesions associated with wild and mutant types. Additionally, the patterns captured by the model were significantly altered by effective tyrosine kinase inhibitors treatment, leading to notable changes in predicted mutation probabilities.

CONCLUSION

PET/CT deep learning model can act as a tool for predicting EGFR mutation in LADC. Additionally, it offers clinicians insights for treatment decisions through evaluations both before and after treatment.

Antihormonal-Treatment Status Affects 68Ga-PSMA-HBED-CC PET Biodistribution in Patients with Prostate Cancer

Androgen deprivation therapy (ADT) is known to influence the prostate-specific membrane antigen (PSMA) expression of prostate cancer, potentially complicating the interpretation of PSMA ligand PET findings and affecting PSMA radioligand therapy. However, the impact of ADT on PSMA ligand biodistribution in nontumorous organs is not well understood. Methods: Men (n = 112) with histologically proven prostate cancer who underwent 68Ga-PSMA-HBED-CC (68Ga-PSMA-11) PET/CT between November 2015 and July 2021 at the Medical University Vienna with known ADT status were retrospectively recruited. Fifty-six patients were on gonadotropin-releasing hormone-interfering ADT at the time of imaging (ADT group), whereas 56 patients with no history of ADT served as a control group. Physiologically PSMA-expressing organs (salivary glands, kidneys, liver, and spleen) were delineated, and their uptake was compared according to their data distributions. Multivariate regression analysis assessed the relationship between renal, hepatic, splenic, and salivary gland uptake and the explanatory variables metabolic tumor volume, glomerular filtration rate, and ADT status. Results: ADT was associated with lower levels of PSMA uptake in the kidneys (SUVmean: Δ[ADT - control] = -7.89; 95% CI, -10.73 to -5.04; P < 0.001), liver (SUVpeak: Δ[ADT - control] = -2.3; 95% CI, -5.72 to -0.93; P = 0.003), spleen (SUVpeak: Δ[ADT - control] = -1.27; 95% CI, -3.61 to -0.16; P = 0.033), and salivary glands (SUVmean: Δ[ADT - control] = -1.04; 95% CI, -2.48 to -0.13; P = 0.027). In a multivariate analysis, ADT was found to be associated with lower renal (SUVmean: β = -7.95; 95% CI, -11.06 to -4.84; P < 0.0001), hepatic (SUVpeak: β = -7.85; 95% CI, -11.78 to -3.91; P < 0.0001), splenic (SUVpeak: β = -5.83; 95% CI, -9.95 to -1.7; P = 0.006), and salivary gland (SUVmean: β = -1.47; 95% CI, -2.76 to -0.17; P = 0.027) uptake. A higher glomerular filtration rate was associated with a higher renal SUVmean (β = 0.16; 95% CI, 0.05 to 0.26; P = 0.0034). Conclusion: These findings suggest that ADT systemically modulates PSMA expression, which may have implications for treatment-optimizing and side-effect-minimizing strategies for PSMA radioligand therapies, particularly those using more potent 225Ac-labeled PSMA conjugates.

Head and Neck Cancer Segmentation in FDG PET Images: Performance Comparison of Convolutional Neural Networks and Vision Transformers

Convolutional neural networks (CNNs) have a proven track record in medical image segmentation. Recently, Vision Transformers were introduced and are gaining popularity for many computer vision applications, including object detection, classification, and segmentation. Machine learning algorithms such as CNNs or Transformers are subject to an inductive bias, which can have a significant impact on the performance of machine learning models. This is especially relevant for medical image segmentation applications where limited training data are available, and a model's inductive bias should help it to generalize well. In this work, we quantitatively assess the performance of two CNN-based networks (U-Net and U-Net-CBAM) and three popular Transformer-based segmentation network architectures (UNETR, TransBTS, and VT-UNet) in the context of HNC lesion segmentation in volumetric [F-18] fluorodeoxyglucose (FDG) PET scans. For performance assessment, 272 FDG PET-CT scans of a clinical trial (ACRIN 6685) were utilized, which includes a total of 650 lesions (primary: 272 and secondary: 378). The image data used are highly diverse and representative for clinical use. For performance analysis, several error metrics were utilized. The achieved Dice coefficient ranged from 0.833 to 0.809 with the best performance being achieved by CNN-based approaches. U-Net-CBAM, which utilizes spatial and channel attention, showed several advantages for smaller lesions compared to the standard U-Net. Furthermore, our results provide some insight regarding the image features relevant for this specific segmentation application. In addition, results highlight the need to utilize primary as well as secondary lesions to derive clinically relevant segmentation performance estimates avoiding biases.

Fully Automated, Fast Motion Correction of Dynamic Whole-Body and Total-Body PET/CT Imaging Studies

We introduce the Fast Algorithm for Motion Correction (FALCON) software, which allows correction of both rigid and nonlinear motion artifacts in dynamic whole-body (WB) images, irrespective of the PET/CT system or the tracer. Methods: Motion was corrected using affine alignment followed by a diffeomorphic approach to account for nonrigid deformations. In both steps, images were registered using multiscale image alignment. Moreover, the frames suited to successful motion correction were automatically estimated by calculating the initial normalized cross-correlation metric between the reference frame and the other moving frames. To evaluate motion correction performance, WB dynamic image sequences from 3 different PET/CT systems (Biograph mCT, Biograph Vision 600, and uEXPLORER) using 6 different tracers (18F-FDG, 18F-fluciclovine, 68Ga-PSMA, 68Ga-DOTATATE, 11C-Pittsburgh compound B, and 82Rb) were considered. Motion correction accuracy was assessed using 4 different measures: change in volume mismatch between individual WB image volumes to assess gross body motion, change in displacement of a large organ (liver dome) within the torso due to respiration, change in intensity in small tumor nodules due to motion blur, and constancy of activity concentration levels. Results: Motion correction decreased gross body motion artifacts and reduced volume mismatch across dynamic frames by about 50%. Moreover, large-organ motion correction was assessed on the basis of correction of liver dome motion, which was removed entirely in about 70% of all cases. Motion correction also improved tumor intensity, resulting in an average increase in tumor SUVs by 15%. Large deformations seen in gated cardiac 82Rb images were managed without leading to anomalous distortions or substantial intensity changes in the resulting images. Finally, the constancy of activity concentration levels was reasonably preserved (<2% change) in large organs before and after motion correction. Conclusion: FALCON allows fast and accurate correction of rigid and nonrigid WB motion artifacts while being insensitive to scanner hardware or tracer distribution, making it applicable to a wide range of PET imaging scenarios.

Predictive value of PET metabolic parameters for occult lymph node metastases in PET/CT defined node-negative patients with advanced epithelial ovarian cancer

Accurate lymph node metastasis (LNM) prediction is crucial for patients with advanced epithelial ovarian cancer (AEOC) since it guides the decisions about lymphadenectomy. Previous studies have shown that occult lymph node metastasis (OLNM) is common in AEOC. The objective of our study is to quantitatively assess the probability of occult lymph node metastasis defined by 18F-Fluorodeoxyglucose PET/CT in AEOC and explore relationship between OLNM and PET metabolic parameters. The patients with pathologically confirmed AEOC who underwent PET/CT for preoperative staging at our institute were reviewed. Univariate and multivariate analysis were performed to evaluate the predictive value of PET/CT-related metabolic parameters for OLNM. The result of our study showed metastatic TLG index had a better diagnostic performance than other PET/CT-related metabolic parameters. Two variables were independently and significantly associated with OLNM in multivariate analysis: metastatic TLG index and primary tumor location. The logistic model combining metastatic TLG index, primary tumor location, and CA125 might be a promising tool to effectively predict the individualized possibility of OLNM for AEOC patients.

The predictive value of [18F]FDG PET/CT radiomics combined with clinical features for EGFR mutation status in different clinical staging of lung adenocarcinoma

BACKGROUND

This study aims to construct radiomics models based on [¹⁸F]FDG PET/CT using multiple machine learning methods to predict the EGFR mutation status of lung adenocarcinoma and evaluate whether incorporating clinical parameters can improve the performance of radiomics models.

METHODS

A total of 515 patients were retrospectively collected and divided into a training set (n = 404) and an independent testing set (n = 111) according to their examination time. After semi-automatic segmentation of PET/CT images, the radiomics features were extracted, and the best feature sets of CT, PET, and PET/CT modalities were screened out. Nine radiomics models were constructed using logistic regression (LR), random forest (RF), and support vector machine (SVM) methods. According to the performance in the testing set, the best model of the three modalities was kept, and its radiomics score (Rad-score) was calculated. Furthermore, combined with the valuable clinical parameters (gender, smoking history, nodule type, CEA, SCC-Ag), a joint radiomics model was built.

RESULTS

Compared with LR and SVM, the RF Rad-score showed the best performance among the three radiomics models of CT, PET, and PET/CT (training and testing sets AUC: 0.688, 0.666, and 0.698 vs. 0.726, 0.678, and 0.704). Among the three joint models, the PET/CT joint model performed the best (training and testing sets AUC: 0.760 vs. 0.730). The further stratified analysis found that CT_RF had the best prediction effect for stage I-II lesions (training set and testing set AUC: 0.791 vs. 0.797), while PET/CT joint model had the best prediction effect for stage III-IV lesions (training and testing sets AUC: 0.722 vs. 0.723).

CONCLUSIONS

Combining with clinical parameters can improve the predictive performance of PET/CT radiomics model, especially for patients with advanced lung adenocarcinoma.

Predicting pathological highly invasive lung cancer from preoperative [18F]FDG PET/CT with multiple machine learning models

PURPOSE

The efficacy of sublobar resection of primary lung cancer have been proven in recent years. However, sublobar resection for highly invasive lung cancer increases local recurrence. We developed and validated multiple machine learning models predicting pathological invasiveness of lung cancer based on preoperative [¹⁸F]fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) radiomic features.

METHODS

Overall, 873 patients who underwent lobectomy or segmentectomy for primary lung cancer were enrolled. Radiomics features were extracted from preoperative PET/CT images with the PyRadiomics package. Seven machine learning models and an ensemble of all models (ENS) were evaluated after 100 iterations. In addition, the probability of highly invasive lung cancer was calculated in a nested cross-validation to assess the calibration plot and clinical usefulness and to compare to consolidation tumour ratio (CTR) on CT images, one of the generally used diagnostic criteria.

RESULTS

In the training set, when PET and CT features were combined, all models achieved an area under the curve (AUC) of ≥ 0.880. In the test set, ENS showed the highest mean AUC of 0.880 and smallest standard deviation of 0.0165, and when the cutoff was 0.5, accuracy of 0.804, F1 of 0.851, precision of 0.821, and recall of 0.885. In the nested cross-validation, the AUC of 0.882 (95% CI: 0.860-0.905) showed a high discriminative ability, and the calibration plot indicated consistency with a Brier score of 0.131. A decision curve analysis showed that the ENS was valid with a threshold probability ranging from 3 to 98%. Accuracy showed an improvement of more than 8% over the CTR.

CONCLUSION

The machine learning model based on preoperative [¹⁸F]FDG PET/CT images was able to predict pathological highly invasive lung cancer with high discriminative ability and stability. The calibration plot showed good consistency, suggesting its usefulness in quantitative risk assessment.

Using stacked deep learning models based on PET/CT images and clinical data to predict EGFR mutations in lung cancer

Purpose

To determine whether stacked deep learning models based on PET/CT images and clinical data can help to predict epidermal growth factor receptor (EGFR) mutations in lung cancer.

Methods

We analyzed data from two public datasets of patients who underwent ¹⁸F-FDG PET/CT. Three PET deep learning ResNet models and one CT deep learning ResNet model were trained as low-level predictors based on PET and CT images, respectively. A high-level Support Vector Machine model (Stack PET/CT and Clinical model) was trained using the prediction results of the low-level predictors and clinical data. The clinical data included sex, age, smoking history, SUVmax and SUVmean of the lesion. Fivefold cross-validation was used in this study to validate the prediction performance of the models. The predictive performance of the models was evaluated by receiver operator characteristic (ROC) curves. The area under the curve (AUC) was calculated.

Results

One hundred forty-seven patients were included in this study. Among them, 37/147 cases were EGFR mutations, and 110/147 cases were EGFR wild-type. The ROC analysis showed that the Stack PET/CT & Clinical model had the best performance (AUC = 0.85 ± 0.09), with 0.76, 0.85 and 0.83 in sensitivity, specificity and accuracy, respectively. Three ResNet PET models had relatively higher AUCs (0.82 ± 0.07, 0.80 ± 0.08 and 0.79 ± 0.07) and outperformed the CT model (AUC = 0.58 ± 0.12).

Conclusion

Using stack generalization, the deep learning model was able to efficiently combine the anatomic and biological imaging information gathered from PET/CT images with clinical data. This stacked deep learning model showed a strong ability to predict EGFR mutations with high accuracy.

Machine Learning for Head and Neck Cancer: A Safe Bet?-A Clinically Oriented Systematic Review for the Radiation Oncologist

BACKGROUND AND PURPOSE

Machine learning (ML) is emerging as a feasible approach to optimize patients' care path in Radiation Oncology. Applications include autosegmentation, treatment planning optimization, and prediction of oncological and toxicity outcomes. The purpose of this clinically oriented systematic review is to illustrate the potential and limitations of the most commonly used ML models in solving everyday clinical issues in head and neck cancer (HNC) radiotherapy (RT).

MATERIALS AND METHODS

Electronic databases were screened up to May 2021. Studies dealing with ML and radiomics were considered eligible. The quality of the included studies was rated by an adapted version of the qualitative checklist originally developed by Luo et al. All statistical analyses were performed using R version 3.6.1.

RESULTS

Forty-eight studies (21 on autosegmentation, four on treatment planning, 12 on oncological outcome prediction, 10 on toxicity prediction, and one on determinants of postoperative RT) were included in the analysis. The most common imaging modality was computed tomography (CT) (40%) followed by magnetic resonance (MR) (10%). Quantitative image features were considered in nine studies (19%). No significant differences were identified in global and methodological scores when works were stratified per their task (i.e., autosegmentation).

DISCUSSION AND CONCLUSION

The range of possible applications of ML in the field of HN Radiation Oncology is wide, albeit this area of research is relatively young. Overall, if not safe yet, ML is most probably a bet worth making.

Application of dual-stream 3D convolutional neural network based on 18F-FDG PET/CT in distinguishing benign and invasive adenocarcinoma in ground-glass lung nodules

Purpose

This work aims to train, validate, and test a dual-stream three-dimensional convolutional neural network (3D-CNN) based on fluorine 18 (¹⁸F)-fluorodeoxyglucose (FDG) PET/CT to distinguish benign lesions and invasive adenocarcinoma (IAC) in ground-glass nodules (GGNs).

Methods

We retrospectively analyzed patients with suspicious GGNs who underwent ¹⁸F-FDG PET/CT in our hospital from November 2011 to November 2020. The patients with benign lesions or IAC were selected for this study. According to the ratio of 7:3, the data were randomly divided into training data and testing data. Partial image feature extraction software was used to segment PET and CT images, and the training data after using the data augmentation were used for the training and validation (fivefold cross-validation) of the three CNNs (PET, CT, and PET/CT networks).

Results

A total of 23 benign nodules and 92 IAC nodules from 106 patients were included in this study. In the training set, the performance of PET network (accuracy, sensitivity, and specificity of 0.92 ± 0.02, 0.97 ± 0.03, and 0.76 ± 0.15) was better than the CT network (accuracy, sensitivity, and specificity of 0.84 ± 0.03, 0.90 ± 0.07, and 0.62 ± 0.16) (especially accuracy was significant, P-value was 0.001); in the testing set, the performance of both networks declined. However, the accuracy and sensitivity of PET network were still higher than that of CT network (0.76 vs. 0.67; 0.85 vs. 0.70). For dual-stream PET/CT network, its performance was almost the same as PET network in the training set (P-value was 0.372–1.000), while in the testing set, although its performance decreased, the accuracy and sensitivity (0.85 and 0.96) were still higher than both CT and PET networks. Moreover, the accuracy of PET/CT network was higher than two nuclear medicine physicians [physician 1 (3-year experience): 0.70 and physician 2 (10-year experience): 0.73].

Conclusion

The 3D-CNN based on ¹⁸F-FDG PET/CT can be used to distinguish benign lesions and IAC in GGNs, and the performance is better when both CT and PET images are used together.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-021-00423-1.

Fully Automated Gross Tumor Volume Delineation From PET in Head and Neck Cancer Using Deep Learning Algorithms

PURPOSE

The availability of automated, accurate, and robust gross tumor volume (GTV) segmentation algorithms is critical for the management of head and neck cancer (HNC) patients. In this work, we evaluated 3 state-of-the-art deep learning algorithms combined with 8 different loss functions for PET image segmentation using a comprehensive training set and evaluated its performance on an external validation set of HNC patients.

PATIENTS AND METHODS

18F-FDG PET/CT images of 470 patients presenting with HNC on which manually defined GTVs serving as standard of reference were used for training (340 patients), evaluation (30 patients), and testing (100 patients from different centers) of these algorithms. PET image intensity was converted to SUVs and normalized in the range (0-1) using the SUVmax of the whole data set. PET images were cropped to 12 × 12 × 12 cm3 subvolumes using isotropic voxel spacing of 3 × 3 × 3 mm3 containing the whole tumor and neighboring background including lymph nodes. We used different approaches for data augmentation, including rotation (-15 degrees, +15 degrees), scaling (-20%, 20%), random flipping (3 axes), and elastic deformation (sigma = 1 and proportion to deform = 0.7) to increase the number of training sets. Three state-of-the-art networks, including Dense-VNet, NN-UNet, and Res-Net, with 8 different loss functions, including Dice, generalized Wasserstein Dice loss, Dice plus XEnt loss, generalized Dice loss, cross-entropy, sensitivity-specificity, and Tversky, were used. Overall, 28 different networks were built. Standard image segmentation metrics, including Dice similarity, image-derived PET metrics, first-order, and shape radiomic features, were used for performance assessment of these algorithms.

RESULTS

The best results in terms of Dice coefficient (mean ± SD) were achieved by cross-entropy for Res-Net (0.86 ± 0.05; 95% confidence interval [CI], 0.85-0.87), Dense-VNet (0.85 ± 0.058; 95% CI, 0.84-0.86), and Dice plus XEnt for NN-UNet (0.87 ± 0.05; 95% CI, 0.86-0.88). The difference between the 3 networks was not statistically significant (P > 0.05). The percent relative error (RE%) of SUVmax quantification was less than 5% in networks with a Dice coefficient more than 0.84, whereas a lower RE% (0.41%) was achieved by Res-Net with cross-entropy loss. For maximum 3-dimensional diameter and sphericity shape features, all networks achieved a RE ≤ 5% and ≤10%, respectively, reflecting a small variability.

CONCLUSIONS

Deep learning algorithms exhibited promising performance for automated GTV delineation on HNC PET images. Different loss functions performed competitively when using different networks and cross-entropy for Res-Net, Dense-VNet, and Dice plus XEnt for NN-UNet emerged as reliable networks for GTV delineation. Caution should be exercised for clinical deployment owing to the occurrence of outliers in deep learning-based algorithms.

The Impact of Artificial Intelligence CNN Based Denoising on FDG PET Radiomics

Background

With a constantly increasing number of diagnostic images performed each year, Artificial Intelligence (AI) denoising methods offer an opportunity to respond to the growing demand. However, it may affect information in the image in an unknown manner. This study quantifies the effect of AI-based denoising on FDG PET textural information in comparison to a convolution with a standard gaussian postfilter (EARL1).

Methods

The study was carried out on 113 patients who underwent a digital FDG PET/CT (VEREOS, Philips Healthcare). 101 FDG avid lesions were segmented semi-automatically by a nuclear medicine physician. VOIs in the liver and lung as reference organs were contoured. PET textural features were extracted with pyradiomics. Texture features from AI denoised and EARL1 versus original PET images were compared with a Concordance Correlation Coefficient (CCC). Features with CCC values ≥ 0.85 threshold were considered concordant. Scatter plots of variable pairs with R2 coefficients of the more relevant features were computed. A Wilcoxon signed rank test to compare the absolute values between AI denoised and original images was performed.

Results

The ratio of concordant features was 90/104 (86.5%) in AI denoised versus 46/104 (44.2%) with EARL1 denoising. In the reference organs, the concordant ratio for AI and EARL1 denoised images was low, respectively 12/104 (11.5%) and 7/104 (6.7%) in the liver, 26/104 (25%) and 24/104 (23.1%) in the lung. SUVpeak was stable after the application of both algorithms in comparison to SUVmax. Scatter plots of variable pairs showed that AI filtering affected more lower versus high intensity regions unlike EARL1 gaussian post filters, affecting both in a similar way. In lesions, the majority of texture features 79/100 (79%) were significantly (p<0.05) different between AI denoised and original PET images.

Conclusions

Applying an AI-based denoising on FDG PET images maintains most of the lesion’s texture information in contrast to EARL1-compatible Gaussian filter. Predictive features of a trained model could be thus the same, however with an adapted threshold. Artificial intelligence based denoising in PET is a very promising approach as it adapts the denoising in function of the tissue type, preserving information where it should.

Impact of total variation regularized expectation maximization reconstruction on the image quality of 68Ga-PSMA PET: a phantom and patient study

OBJECTIVES

To investigate the impact of total variation regularized expectation maximization (TVREM) reconstruction on the image quality of 68Ga-PSMA-11 PET/CT using phantom and patient data.

METHODS

Images of a phantom with small hot sphere inserts and 20 prostate cancer patients were acquired with a digital PET/CT using list-mode and reconstructed with ordered subset expectation maximization (OSEM) and TVREM with seven penalisation factors between 0.01 and 0.42 for 2 and 3 minutes-per-bed (m/b) acquisition. The contrast recovery (CR) and background variability (BV) of the phantom, image noise of the liver, and SUVmax of the lesions were measured. Qualitative image quality was scored by two radiologists using a 5-point scale (1-poor, 5-excellent).

RESULTS

The performance of CR, BV, and image noise, and the gain of SUVmax was higher for TVREM 2 m/b groups with the penalization of 0.07 to 0.28 compared to OSEM 3 m/b group (all p < 0.05). The image noise of OSEM 3 m/b group was equivalent to TVREM 2 and 3 m/b groups with a penalization of 0.14 and 0.07, while lesions' SUVmax increased 15 and 20%. The highest qualitative score was attained at the penalization of 0.21 (3.30 ± 0.66) for TVREM 2 m/b groups and the penalization 0.14 (3.80 ± 0.41) for 3 m/b group that equal to or greater than OSEM 3 m/b group (2.90 ± 0.45, p = 0.2 and p < 0.001).

CONCLUSIONS

TVREM improves lesion contrast and reduces image noise, which allows shorter acquisition with preserved image quality for PSMA PET/CT.

ADVANCES IN KNOWLEDGE

TVREM reconstruction with optimized penalization factors can generate higher quality PSMA-PET images for prostate cancer diagnosis.

Early differences in dynamic uptake of 68Ga-PSMA-11 in primary prostate cancer: A test-retest study

INTRODUCTION

Dynamic PET/CT allows visualization of pharmacokinetics over the time, in contrast to static whole body PET/CT. The objective of this study was to assess 68Ga-PSMA-11 uptake in pathological lesions and benign tissue, within 30 minutes after injection in primary prostate cancer (PCa) patients in test-retest setting.

MATERIALS AND METHODS

Five patients, with biopsy proven PCa, were scanned dynamically in list mode for 30 minutes on a digital PET/CT-scanner directly after an intravenous bolus injection of 100 MBq 68Ga-PSMA-11. Approximately 45 minutes after injection a static whole body scan was acquired, followed by a one bed position scan of the pelvic region. The scans were repeated approximately four weeks later, without any intervention in between. Semi-quantitative assessment was performed using regions-of-interest in the prostate tumor, bladder, gluteal muscle and iliac artery. Time-activity curves were extracted from the counts in these regions and the intra-patient variability between both scans was assessed.

RESULTS

The uptake of the iliac artery and gluteal muscle reached a plateau after 5 and 3 minutes, respectively. The population fell apart in two groups with respect to tumor uptake: in some patients the tumor uptake reached a plateau after 5 minutes, whereas in other patients the uptake kept increasing, which correlated with larger tumor volumes on PET/CT scan. Median intra-patient variation between both scans was 12.2% for artery, 9.7% for tumor, 32.7% for the bladder and 14.1% for the gluteal muscle.

CONCLUSION

Dynamic 68Ga-PSMA-11 PET/CT scans, with a time interval of four weeks, are reproducible with a 10% variation in uptake in the primary prostate tumor. An uptake plateau was reached for the iliac artery and gluteal muscle within 5 minutes post-injection. A larger tumor volume seems to be related to continued tumor uptake. This information might be relevant for both response monitoring and PSMA-based radionuclide therapies.

A CT-based radiomics model to detect prostate cancer lymph node metastases in PSMA radioguided surgery patients

PURPOSE

In recurrent prostate carcinoma, determination of the site of recurrence is crucial to guide personalized therapy. In contrast to prostate-specific membrane antigen (PSMA)-positron emission tomography (PET) imaging, computed tomography (CT) has only limited capacity to detect lymph node metastases (LNM). We sought to develop a CT-based radiomic model to predict LNM status using a PSMA radioguided surgery (RGS) cohort with histological confirmation of all suspected lymph nodes (LNs).

METHODS

Eighty patients that received RGS for resection of PSMA PET/CT-positive LNMs were analyzed. Forty-seven patients (87 LNs) that received inhouse imaging were used as training cohort. Thirty-three patients (62 LNs) that received external imaging were used as testing cohort. As gold standard, histological confirmation was available for all LNs. After preprocessing, 156 radiomic features analyzing texture, shape, intensity, and local binary patterns (LBP) were extracted. The least absolute shrinkage and selection operator (radiomic models) and logistic regression (conventional parameters) were used for modeling.

RESULTS

Texture and shape features were largely correlated to LN volume. A combined radiomic model achieved the best predictive performance with a testing-AUC of 0.95. LBP features showed the highest contribution to model performance. This model significantly outperformed all conventional CT parameters including LN short diameter (AUC 0.84), LN volume (AUC 0.80), and an expert rating (AUC 0.67). In lymph node-specific decision curve analysis, there was a clinical net benefit above LN short diameter.

CONCLUSION

The best radiomic model outperformed conventional measures for detection of LNM demonstrating an incremental value of radiomic features.

A unified framework for simultaneous assessment of accuracy, between-, and within-reader variability of image segmentations

Medical imaging is utilized in a wide range of clinical applications. To enable a detailed quantitative analysis, medical images must often be segmented to label (delineate) structures of interest; for example, a tumor. Frequently, manual segmentation is utilized in clinical practice (e.g., in radiation oncology) to define such structures of interest. However, it can be quite time consuming and subject to substantial between-, and within-reader variability. A more reproducible, less variable, and more time efficient segmentation approach is likely to improve medical treatment. This potential has spurred the development of segmentation algorithms which harness computational power. Segmentation algorithms’ widespread use is limited due to difficulty in quantifying their performance relative to manual segmentation, which itself is subject to variation. This paper presents a statistical model which simultaneously estimates segmentation method accuracy, and between- and within-reader variability. The model is simultaneously fit for multiple segmentation methods within a unified Bayesian framework. The Bayesian model is compared to other methods used in literature via a simulation study, and application to head and neck cancer PET/CT data. The modeling framework is flexible and can be employed in numerous comparison applications. Several alternate applications are discussed in the paper.

Quantitative Imaging Informatics for Cancer Research

PURPOSE

We summarize Quantitative Imaging Informatics for Cancer Research (QIICR; U24 CA180918), one of the first projects funded by the National Cancer Institute (NCI) Informatics Technology for Cancer Research program.

METHODS

QIICR was motivated by the 3 use cases from the NCI Quantitative Imaging Network. 3D Slicer was selected as the platform for implementation of open-source quantitative imaging (QI) tools. Digital Imaging and Communications in Medicine (DICOM) was chosen for standardization of QI analysis outputs. Support of improved integration with community repositories focused on The Cancer Imaging Archive (TCIA). Priorities included improved capabilities of the standard, toolkits and tools, reference datasets, collaborations, and training and outreach.

RESULTS

Fourteen new tools to support head and neck cancer, glioblastoma, and prostate cancer QI research were introduced and downloaded over 100,000 times. DICOM was amended, with over 40 correction proposals addressing QI needs. Reference implementations of the standard in a popular toolkit and standalone tools were introduced. Eight datasets exemplifying the application of the standard and tools were contributed. An open demonstration/connectathon was organized, attracting the participation of academic groups and commercial vendors. Integration of tools with TCIA was improved by implementing programmatic communication interface and by refining best practices for QI analysis results curation.

CONCLUSION

Tools, capabilities of the DICOM standard, and datasets we introduced found adoption and utility within the cancer imaging community. A collaborative approach is critical to addressing challenges in imaging informatics at the national and international levels. Numerous challenges remain in establishing and maintaining the infrastructure of analysis tools and standardized datasets for the imaging community. Ideas and technology developed by the QIICR project are contributing to the NCI Imaging Data Commons currently being developed.

FLT PET Radiomics for Response Prediction to Chemoradiation Therapy in Head and Neck Squamous Cell Cancer

Radiomics is an image analysis approach for extracting large amounts of quantitative information from medical images using a variety of computational methods. Our goal was to evaluate the utility of radiomic feature analysis from ¹⁸F-fluorothymidine positron emission tomography (FLT PET) obtained at baseline in prediction of treatment response in patients with head and neck cancer. Thirty patients with advanced-stage oropharyngeal or laryngeal cancer, treated with definitive chemoradiation therapy, underwent FLT PET imaging before treatment. In total, 377 radiomic features of FLT uptake and feature variants were extracted from volumes of interest; these features variants were defined by either the primary tumor or the total lesion burden, which consisted of the primary tumor and all FLT-avid nodes. Feature variants included normalized measurements of uptake, which were calculated by dividing lesion uptake values by the mean uptake value in the bone marrow. Feature reduction was performed using clustering to remove redundancy, leaving 172 representative features. Effects of these features on progression-free survival were modeled with Cox regression and P-values corrected for multiple comparisons. In total, 9 features were considered significant. Our results suggest that smaller, more homogenous lesions at baseline were associated with better prognosis. In addition, features extracted from total lesion burden had a higher concordance index than primary tumor features for 8 of the 9 significant features. Furthermore, total lesion burden features showed lower interobserver variability.

Gross tumor volume segmentation for head and neck cancer radiotherapy using deep dense multi-modality network

In radiation therapy, the accurate delineation of gross tumor volume (GTV) is crucial for treatment planning. However, it is challenging for head and neck cancer (HNC) due to the morphology complexity of various organs in the head, low targets to background contrast and potential artifacts on conventional planning CT images. Thus, manual delineation of GTV on anatomical images is extremely time consuming and suffers from inter-observer variability that leads to planning uncertainty. With the wide use of PET/CT imaging in oncology, complementary functional and anatomical information can be utilized for tumor contouring and bring a significant advantage for radiation therapy planning. In this study, by taking advantage of multi-modality PET and CT images, we propose an automatic GTV segmentation framework based on deep learning for HNC. The backbone of this segmentation framework is based on 3D convolution with dense connections which enables a better information propagation and takes full advantage of the features extracted from multi-modality input images. We evaluate our proposed framework on a dataset including 250 HNC patients. Each patient receives both planning CT and PET/CT imaging before radiation therapy (RT). Manually delineated GTV contours by radiation oncologists are used as ground truth in this study. To further investigate the advantage of our proposed Dense-Net framework, we also compared with the framework using 3D U-Net which is the state-of-the-art in segmentation tasks. Meanwhile, for each frame, the performance comparison between single modality input (PET or CT image) and multi-modality input (both PET/CT) is conducted. Dice coefficient, mean surface distance (MSD), 95th-percentile Hausdorff distance (HD95) and displacement of mass centroid (DMC) are calculated for quantitative evaluation. The dataset is split into train (140 patients), validation (35 patients) and test (75 patients) groups to optimize the network. Based on the results on independent test group, our proposed multi-modality Dense-Net (Dice 0.73) shows better performance than the compared network (Dice 0.71). Furthermore, the proposed Dense-Net structure has less trainable parameters than the 3D U-Net, which reduces the prediction variability. In conclusion, our proposed multi-modality Dense-Net can enable satisfied GTV segmentation for HNC using multi-modality images and yield superior performance than conventional methods. Our proposed method provides an automatic, fast and consistent solution for GTV segmentation and shows potentials to be generally applied for radiation therapy planning of a variety of cancer (e.g. lung, sarcoma, liver and so on).

FDG PET based prediction of response in head and neck cancer treatment: Assessment of new quantitative imaging features

Introduction

18 F-fluorodeoxyglucose (FDG) positron emission tomography (PET) is now a standard diagnostic imaging test performed in patients with head and neck cancer for staging, re-staging, radiotherapy planning, and outcome assessment. Currently, quantitative analysis of FDG PET scans is limited to simple metrics like maximum standardized uptake value, metabolic tumor volume, or total lesion glycolysis, which have limited predictive value. The goal of this work was to assess the predictive potential of new (i.e., nonstandard) quantitative imaging features on head and neck cancer outcome.

Methods

This retrospective study analyzed fifty-eight pre- and post-treatment FDG PET scans of patients with head and neck squamous cell cancer to calculate five standard and seventeen new features at baseline and post-treatment. Cox survival regression was used to assess the predictive potential of each quantitative imaging feature on disease-free survival.

Results

Analysis showed that the post-treatment change of the average tracer uptake in the rim background region immediately adjacent to the tumor normalized by uptake in the liver represents a novel PET feature that is associated with disease-free survival (HR 1.95; 95% CI 1.27, 2.99) and has good discriminative performance (c index 0.791).

Conclusion

The reported findings define a promising new direction for quantitative imaging biomarker research in head and neck squamous cell cancer and highlight the potential role of new radiomics features in oncology decision making as part of precision medicine.

Joint Tumor Segmentation in PET-CT Images Using Co-Clustering and Fusion Based on Belief Functions

Precise delineation of target tumor is a key factor to ensure the effectiveness of radiation therapy. While hybrid positron emission tomography-computed tomography (PET-CT) has become a standard imaging tool in the practice of radiation oncology, many existing automatic/semi-automatic methods still perform tumor segmentation on mono-modal images. In this paper, a co-clustering algorithm is proposed to concurrently segment 3D tumors in PET-CT images, considering that the two complementary imaging modalities can combine functional and anatomical information to improve segmentation performance. The theory of belief functions is adopted in the proposed method to model, fuse, and reason with uncertain and imprecise knowledge from noisy and blurry PET-CT images. To ensure reliable segmentation for each modality, the distance metric for the quantification of clustering distortions and spatial smoothness is iteratively adapted during the clustering procedure. On the other hand, to encourage consistent segmentation between different modalities, a specific context term is proposed in the clustering objective function. Moreover, during the iterative optimization process, clustering results for the two distinct modalities are further adjusted via a belief-functions-based information fusion strategy. The proposed method has been evaluated on a data set consisting of 21 paired PET-CT images for non-small cell lung cancer patients. The quantitative and qualitative evaluations show that our proposed method performs well compared with the state-of-the-art methods.

Simultaneous cosegmentation of tumors in PET-CT images using deep fully convolutional networks

PURPOSE

To investigate the use and efficiency of 3-D deep learning, fully convolutional networks (DFCN) for simultaneous tumor cosegmentation on dual-modality nonsmall cell lung cancer (NSCLC) and positron emission tomography (PET)-computed tomography (CT) images.

METHODS

We used DFCN cosegmentation for NSCLC tumors in PET-CT images, considering both the CT and PET information. The proposed DFCN-based cosegmentation method consists of two coupled three-dimensional (3D)-UNets with an encoder-decoder architecture, which can communicate with the other in order to share complementary information between PET and CT. The weighted average sensitivity and positive predictive values denoted as Scores, dice similarity coefficients (DSCs), and the average symmetric surface distances were used to assess the performance of the proposed approach on 60 pairs of PET/CTs. A Simultaneous Truth and Performance Level Estimation Algorithm (STAPLE) of 3 expert physicians' delineations were used as a reference. The proposed DFCN framework was compared to 3 graph-based cosegmentation methods.

RESULTS

Strong agreement was observed when using the STAPLE references for the proposed DFCN cosegmentation on the PET-CT images. The average DSCs on CT and PET are 0.861 ± 0.037 and 0.828 ± 0.087, respectively, using DFCN, compared to 0.638 ± 0.165 and 0.643 ± 0.141, respectively, when using the graph-based cosegmentation method. The proposed DFCN cosegmentation using both PET and CT also outperforms the deep learning method using either PET or CT alone.

CONCLUSIONS

The proposed DFCN cosegmentation is able to outperform existing graph-based segmentation methods. The proposed DFCN cosegmentation shows promise for further integration with quantitative multimodality imaging tools in clinical trials.

The Use of Quantitative Imaging in Radiation Oncology: A Quantitative Imaging Network (QIN) Perspective

Modern radiation therapy is delivered with great precision, in part by relying on high-resolution multidimensional anatomic imaging to define targets in space and time. The development of quantitative imaging (QI) modalities capable of monitoring biologic parameters could provide deeper insight into tumor biology and facilitate more personalized clinical decision-making. The Quantitative Imaging Network (QIN) was established by the National Cancer Institute to advance and validate these QI modalities in the context of oncology clinical trials. In particular, the QIN has significant interest in the application of QI to widen the therapeutic window of radiation therapy. QI modalities have great promise in radiation oncology and will help address significant clinical needs, including finer prognostication, more specific target delineation, reduction of normal tissue toxicity, identification of radioresistant disease, and clearer interpretation of treatment response. Patient-specific QI is being incorporated into radiation treatment design in ways such as dose escalation and adaptive replanning, with the intent of improving outcomes while lessening treatment morbidities. This review discusses the current vision of the QIN, current areas of investigation, and how the QIN hopes to enhance the integration of QI into the practice of radiation oncology.

Reproducibility of F18-FDG PET radiomic features for different cervical tumor segmentation methods, gray-level discretization, and reconstruction algorithms

Site‐specific investigations of the role of radiomics in cancer diagnosis and therapy are emerging. We evaluated the reproducibility of radiomic features extracted from ¹⁸Flourine–fluorodeoxyglucose (¹⁸F‐FDG) PET images for three parameters: manual versus computer‐aided segmentation methods, gray‐level discretization, and PET image reconstruction algorithms. Our cohort consisted of pretreatment PET/CT scans from 88 cervical cancer patients. Two board‐certified radiation oncologists manually segmented the metabolic tumor volume (MTV₁ and MTV₂) for each patient. For comparison, we used a graphical‐based method to generate semiautomated segmented volumes (GBSV). To address any perturbations in radiomic feature values, we down‐sampled the tumor volumes into three gray‐levels: 32, 64, and 128 from the original gray‐level of 256. Finally, we analyzed the effect on radiomic features on PET images of eight patients due to four PET 3D‐reconstruction algorithms: maximum likelihood‐ordered subset expectation maximization (OSEM) iterative reconstruction (IR) method, fourier rebinning‐ML‐OSEM (FOREIR), FORE‐filtered back projection (FOREFBP), and 3D‐Reprojection (3DRP) analytical method. We extracted 79 features from all segmentation method, gray‐levels of down‐sampled volumes, and PET reconstruction algorithms. The features were extracted using gray‐level co‐occurrence matrices (GLCM), gray‐level size zone matrices (GLSZM), gray‐level run‐length matrices (GLRLM), neighborhood gray‐tone difference matrices (NGTDM), shape‐based features (SF), and intensity histogram features (IHF). We computed the Dice coefficient between each MTV and GBSV to measure segmentation accuracy. Coefficient values close to one indicate high agreement, and values close to zero indicate low agreement. We evaluated the effect on radiomic features by calculating the mean percentage differences (d¯) between feature values measured from each pair of parameter elements (i.e. segmentation methods: MTV₁‐MTV₂, MTV₁‐GBSV, MTV₂‐GBSV; gray‐levels: 64‐32, 64‐128, and 64‐256; reconstruction algorithms: OSEM‐FORE‐OSEM, OSEM‐FOREFBP, and OSEM‐3DRP). We used |d¯| as a measure of radiomic feature reproducibility level, where any feature scored |d¯| ±SD ≤ |25|% ± 35% was considered reproducible. We used Bland–Altman analysis to evaluate the mean, standard deviation (SD), and upper/lower reproducibility limits (U/LRL) for radiomic features in response to variation in each testing parameter. Furthermore, we proposed U/LRL as a method to classify the level of reproducibility: High— ±1% ≤ U/LRL ≤ ±30%; Intermediate— ±30% < U/LRL ≤ ±45%; Low— ±45 < U/LRL ≤ ±50%. We considered any feature below the low level as nonreproducible (NR). Finally, we calculated the interclass correlation coefficient (ICC) to evaluate the reliability of radiomic feature measurements for each parameter. The segmented volumes of 65 patients (81.3%) scored Dice coefficient >0.75 for all three volumes. The result outcomes revealed a tendency of higher radiomic feature reproducibility among segmentation pair MTV₁‐GBSV than MTV₂‐GBSV, gray‐level pairs of 64‐32 and 64‐128 than 64‐256, and reconstruction algorithm pairs of OSEM‐FOREIR and OSEM‐FOREFBP than OSEM‐3DRP. Although the choice of cervical tumor segmentation method, gray‐level value, and reconstruction algorithm may affect radiomic features, some features were characterized by high reproducibility through all testing parameters. The number of radiomic features that showed insensitivity to variations in segmentation methods, gray‐level discretization, and reconstruction algorithms was 10 (13%), 4 (5%), and 1 (1%), respectively. These results suggest that a careful analysis of the effects of these parameters is essential prior to any radiomics clinical application.

Development of a radiobiological evaluation tool to assess the expected clinical impacts of contouring accuracy between manual and semi-automated segmentation algorithms

RADEval is a tool developed to assess the expected clinical impact of contouring accuracy when comparing manual contouring and semi-automated segmentation. The RADEval tool, designed to process large scale datasets, imported a total of 2,760 segmentation datasets, along with a Simultaneous Truth and Performance Level Estimation (STAPLE) to act as ground truth tumor segmentations. Virtual dose-maps were created within RADEval and two different tumor control probability (TCP) values using a Logistic and a Poisson TCP models were calculated in RADEval using each STAPLE and each dose-map. RADEval also virtually generated a ring of normal tissue. To evaluate clinical impact, two different uncomplicated TCP (UTCP) values were calculated in RADEval by using two TCP-NTCP correlation parameters (δ = 0 and 1). NTCP values showed that semi-automatic segmentation resulted in lower NTCP with an average 1.5 - 1.6 % regardless of STAPLE design. This was true even though each normal tissue was created from each STAPLE (p <; 0.00001). TCP and UTCP presented no statistically significant differences (p ≥ 0.1884). The intra-operator standard deviations (SDs) for TCP, NTCP and UTCP were significantly lower for the semi-automatic segmentation method regardless of STAPLE design (p <; 0.0331). Both intra-and inter-operator SDs of TCP, NTCP and UTCP were significantly lower for semi-automatic segmentation for the STAPLE 1 design (p <;0.0331). RADEval was able to efficiently process 4,920 datasets of two STAPLE designs and successfully assess the expected clinical impact of contouring accuracy.

Classification and evaluation strategies of auto-segmentation approaches for PET: Report of AAPM task group No. 211

PURPOSE

The purpose of this educational report is to provide an overview of the present state-of-the-art PET auto-segmentation (PET-AS) algorithms and their respective validation, with an emphasis on providing the user with help in understanding the challenges and pitfalls associated with selecting and implementing a PET-AS algorithm for a particular application.

APPROACH

A brief description of the different types of PET-AS algorithms is provided using a classification based on method complexity and type. The advantages and the limitations of the current PET-AS algorithms are highlighted based on current publications and existing comparison studies. A review of the available image datasets and contour evaluation metrics in terms of their applicability for establishing a standardized evaluation of PET-AS algorithms is provided. The performance requirements for the algorithms and their dependence on the application, the radiotracer used and the evaluation criteria are described and discussed. Finally, a procedure for algorithm acceptance and implementation, as well as the complementary role of manual and auto-segmentation are addressed.

FINDINGS

A large number of PET-AS algorithms have been developed within the last 20 years. Many of the proposed algorithms are based on either fixed or adaptively selected thresholds. More recently, numerous papers have proposed the use of more advanced image analysis paradigms to perform semi-automated delineation of the PET images. However, the level of algorithm validation is variable and for most published algorithms is either insufficient or inconsistent which prevents recommending a single algorithm. This is compounded by the fact that realistic image configurations with low signal-to-noise ratios (SNR) and heterogeneous tracer distributions have rarely been used. Large variations in the evaluation methods used in the literature point to the need for a standardized evaluation protocol.

CONCLUSIONS

Available comparison studies suggest that PET-AS algorithms relying on advanced image analysis paradigms provide generally more accurate segmentation than approaches based on PET activity thresholds, particularly for realistic configurations. However, this may not be the case for simple shape lesions in situations with a narrower range of parameters, where simpler methods may also perform well. Recent algorithms which employ some type of consensus or automatic selection between several PET-AS methods have potential to overcome the limitations of the individual methods when appropriately trained. In either case, accuracy evaluation is required for each different PET scanner and scanning and image reconstruction protocol. For the simpler, less robust approaches, adaptation to scanning conditions, tumor type, and tumor location by optimization of parameters is necessary. The results from the method evaluation stage can be used to estimate the contouring uncertainty. All PET-AS contours should be critically verified by a physician. A standard test, i.e., a benchmark dedicated to evaluating both existing and future PET-AS algorithms needs to be designed, to aid clinicians in evaluating and selecting PET-AS algorithms and to establish performance limits for their acceptance for clinical use. The initial steps toward designing and building such a standard are undertaken by the task group members.

Multi-site quality and variability analysis of 3D FDG PET segmentations based on phantom and clinical image data

PURPOSE

Radiomics utilizes a large number of image-derived features for quantifying tumor characteristics that can in turn be correlated with response and prognosis. Unfortunately, extraction and analysis of such image-based features is subject to measurement variability and bias. The challenge for radiomics is particularly acute in Positron Emission Tomography (PET) where limited resolution, a high noise component related to the limited stochastic nature of the raw data, and the wide variety of reconstruction options confound quantitative feature metrics. Extracted feature quality is also affected by tumor segmentation methods used to define regions over which to calculate features, making it challenging to produce consistent radiomics analysis results across multiple institutions that use different segmentation algorithms in their PET image analysis. Understanding each element contributing to these inconsistencies in quantitative image feature and metric generation is paramount for ultimate utilization of these methods in multi-institutional trials and clinical oncology decision making.

METHODS

To assess segmentation quality and consistency at the multi-institutional level, we conducted a study of seven institutional members of the National Cancer Institute Quantitative Imaging Network. For the study, members were asked to segment a common set of phantom PET scans acquired over a range of imaging conditions as well as a second set of head and neck cancer (HNC) PET scans. Segmentations were generated at each institution using their preferred approach. In addition, participants were asked to repeat segmentations with a time interval between initial and repeat segmentation. This procedure resulted in overall 806 phantom insert and 641 lesion segmentations. Subsequently, the volume was computed from the segmentations and compared to the corresponding reference volume by means of statistical analysis.

RESULTS

On the two test sets (phantom and HNC PET scans), the performance of the seven segmentation approaches was as follows. On the phantom test set, the mean relative volume errors ranged from 29.9 to 87.8% of the ground truth reference volumes, and the repeat difference for each institution ranged between -36.4 to 39.9%. On the HNC test set, the mean relative volume error ranged between -50.5 to 701.5%, and the repeat difference for each institution ranged between -37.7 to 31.5%. In addition, performance measures per phantom insert/lesion size categories are given in the paper. On phantom data, regression analysis resulted in coefficient of variation (CV) components of 42.5% for scanners, 26.8% for institutional approaches, 21.1% for repeated segmentations, 14.3% for relative contrasts, 5.3% for count statistics (acquisition times), and 0.0% for repeated scans. Analysis showed that the CV components for approaches and repeated segmentations were significantly larger on the HNC test set with increases by 112.7% and 102.4%, respectively.

CONCLUSION

Analysis results underline the importance of PET scanner reconstruction harmonization and imaging protocol standardization for quantification of lesion volumes. In addition, to enable a distributed multi-site analysis of FDG PET images, harmonization of analysis approaches and operator training in combination with highly automated segmentation methods seems to be advisable. Future work will focus on quantifying the impact of segmentation variation on radiomics system performance.

Graph-based segmentation of abnormal nuclei in cervical cytology

A general method is reported for improving the segmentation of abnormal cell nuclei in cervical cytology images. In automation-assisted reading of cervical cytology, one of the essential steps is the segmentation of nuclei. Despite some progress, there is a need to improve the sensitivity, particularly the segmentation of abnormal nuclei. Our method starts with pre-segmenting the nucleus to define the coarse center and size of nucleus, which is used to construct a graph by image unfolding that maps ellipse-like border in the Cartesian coordinate system to lines in the polar coordinate system. The cost function jointly reflects properties of nucleus border and nucleus region. The prior constraints regarding the context of nucleus-cytoplasm position are utilized to modify the local cost functions. The globally optimal path in the constructed graph is then identified by dynamic programming with an iterative approach ensuring an optimal closed contour. Validation of our method was performed on abnormal nuclei from two cervical cell image datasets, Herlev and H&E stained manual liquid-based cytology (HEMLBC). Compared with five state-of-the-art approaches, our graph-search based method shows superior performance.

DICOM for quantitative imaging biomarker development: a standards based approach to sharing clinical data and structured PET/CT analysis results in head and neck cancer research

Background. Imaging biomarkers hold tremendous promise for precision medicine clinical applications. Development of such biomarkers relies heavily on image post-processing tools for automated image quantitation. Their deployment in the context of clinical research necessitates interoperability with the clinical systems. Comparison with the established outcomes and evaluation tasks motivate integration of the clinical and imaging data, and the use of standardized approaches to support annotation and sharing of the analysis results and semantics. We developed the methodology and tools to support these tasks in Positron Emission Tomography and Computed Tomography (PET/CT) quantitative imaging (QI) biomarker development applied to head and neck cancer (HNC) treatment response assessment, using the Digital Imaging and Communications in Medicine (DICOM(®)) international standard and free open-source software. Methods. Quantitative analysis of PET/CT imaging data collected on patients undergoing treatment for HNC was conducted. Processing steps included Standardized Uptake Value (SUV) normalization of the images, segmentation of the tumor using manual and semi-automatic approaches, automatic segmentation of the reference regions, and extraction of the volumetric segmentation-based measurements. Suitable components of the DICOM standard were identified to model the various types of data produced by the analysis. A developer toolkit of conversion routines and an Application Programming Interface (API) were contributed and applied to create a standards-based representation of the data. Results. DICOM Real World Value Mapping, Segmentation and Structured Reporting objects were utilized for standards-compliant representation of the PET/CT QI analysis results and relevant clinical data. A number of correction proposals to the standard were developed. The open-source DICOM toolkit (DCMTK) was improved to simplify the task of DICOM encoding by introducing new API abstractions. Conversion and visualization tools utilizing this toolkit were developed. The encoded objects were validated for consistency and interoperability. The resulting dataset was deposited in the QIN-HEADNECK collection of The Cancer Imaging Archive (TCIA). Supporting tools for data analysis and DICOM conversion were made available as free open-source software. Discussion. We presented a detailed investigation of the development and application of the DICOM model, as well as the supporting open-source tools and toolkits, to accommodate representation of the research data in QI biomarker development. We demonstrated that the DICOM standard can be used to represent the types of data relevant in HNC QI biomarker development, and encode their complex relationships. The resulting annotated objects are amenable to data mining applications, and are interoperable with a variety of systems that support the DICOM standard.