A distributed feature selection pipeline for survival analysis using radiomics in non-small cell lung cancer patients

Predictive modelling of cancer outcomes using radiomics faces dimensionality problems and data limitations, as radiomics features often number in the hundreds, and multi-institutional data sharing is ()often unfeasible. Federated learning (FL) and feature selection (FS) techniques combined can help overcome these issues, as one provides the means of training models without exchanging sensitive data, while the other identifies the most informative features, reduces overfitting, and improves model interpretability. Our proposed FS pipeline based on FL principles targets data-driven radiomics FS in a multivariate survival study of non-small cell lung cancer patients. The pipeline was run across datasets from three institutions without patient-level data exchange. It includes two FS techniques, Correlation-based Feature Selection and LASSO regularization, and Cox Proportional-Hazard regression with Overall Survival as endpoint. Trained and validated on 828 patients overall, our pipeline yielded a radiomic signature comprising "intensity-based energy" and "mean discretised intensity". Validation resulted in a mean Harrell C-index of 0.59, showcasing fair efficacy in risk stratification. In conclusion, we suggest a distributed radiomics approach that incorporates preliminary feature selection to systematically decrease the feature set based on data-driven considerations. This aims to address dimensionality challenges beyond those associated with data constraints and interpretability concerns.

SwinHR: Hemodynamic-powered hierarchical vision transformer for breast tumor segmentation

Accurate and automated segmentation of breast tumors in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) plays a critical role in computer-aided diagnosis and treatment of breast cancer. However, this task is challenging, due to random variation in tumor sizes, shapes, appearances, and blurred boundaries of tumors caused by inherent heterogeneity of breast cancer. Moreover, the presence of ill-posed artifacts in DCE-MRI further complicate the process of tumor region annotation. To address the challenges above, we propose a scheme (named SwinHR) integrating prior DCE-MRI knowledge and temporal-spatial information of breast tumors. The prior DCE-MRI knowledge refers to hemodynamic information extracted from multiple DCE-MRI phases, which can provide pharmacokinetics information to describe metabolic changes of the tumor cells over the scanning time. The Swin Transformer with hierarchical re-parameterization large kernel architecture (H-RLK) can capture long-range dependencies within DCE-MRI while maintaining computational efficiency by a shifted window-based self-attention mechanism. The use of H-RLK can extract high-level features with a wider receptive field, which can make the model capture contextual information at different levels of abstraction. Extensive experiments are conducted in large-scale datasets to validate the effectiveness of our proposed SwinHR scheme, demonstrating its superiority over recent state-of-the-art segmentation methods. Also, a subgroup analysis split by MRI scanners, field strength, and tumor size is conducted to verify its generalization. The source code is released on (https://github.com/GDPHMediaLab/SwinHR).

A PET/CT radiomics model for predicting distant metastasis in early-stage non-small cell lung cancer patients treated with stereotactic body radiotherapy: a multicentric study

OBJECTIVES

Stereotactic body radiotherapy (SBRT) is a treatment option for patients with early-stage non-small cell lung cancer (NSCLC) who are unfit for surgery. Some patients may experience distant metastasis. This study aimed to develop and validate a radiomics model for predicting distant metastasis in patients with early-stage NSCLC treated with SBRT.

METHODS

Patients at five institutions were enrolled in this study. Radiomics features were extracted based on the PET/CT images. After feature selection in the training set (from Tianjin), CT-based and PET-based radiomics signatures were built. Models based on CT and PET signatures were built and validated using external datasets (from Zhejiang, Zhengzhou, Shandong, and Shanghai). An integrated model that included CT and PET radiomic signatures was developed. The performance of the proposed model was evaluated in terms of its discrimination, calibration, and clinical utility. Multivariate logistic regression was used to calculate the probability of distant metastases. The cutoff value was obtained using the receiver operator characteristic curve (ROC), and the patients were divided into high- and low-risk groups. Kaplan-Meier analysis was used to evaluate the distant metastasis-free survival (DMFS) of different risk groups.

RESULTS

In total, 228 patients were enrolled. The median follow-up time was 31.4 (2.0-111.4) months. The model based on CT radiomics signatures had an area under the curve (AUC) of 0.819 in the training set (n = 139) and 0.786 in the external dataset (n = 89). The PET radiomics model had an AUC of 0.763 for the training set and 0.804 for the external dataset. The model combining CT and PET radiomics had an AUC of 0.835 for the training set and 0.819 for the external dataset. The combined model showed a moderate calibration and a positive net benefit. When the probability of distant metastasis was greater than 0.19, the patient was considered to be at high risk. The DMFS of patients with high- and low-risk was significantly stratified (P < 0.001).

CONCLUSIONS

The proposed PET/CT radiomics model can be used to predict distant metastasis in patients with early-stage NSCLC treated with SBRT and provide a reference for clinical decision-making. In this study, the model was established by combining CT and PET radiomics signatures in a moderate-quantity training cohort of early-stage NSCLC patients treated with SBRT and was successfully validated in independent cohorts. Physicians could use this easy-to-use model to assess the risk of distant metastasis after SBRT. Identifying subgroups of patients with different risk factors for distant metastasis is useful for guiding personalized treatment approaches.

External validation of a lung cancer-based prediction model for two-year mortality in esophageal cancer patient cohorts

PURPOSE/OBJECTIVE

Chemo-radiotherapy can improve the oncological outcome of esophageal cancer (EC) patients, but may cause long term radiation-induced toxicity, including an increased risk of non-cancer related death. For lung cancer patients, a model to predict 2-year total mortality using mean heart dose (MHD) and gross tumor volume (GTV) has previously been developed and validated. This project aimed to externally validate this model in EC patients.

METHODS

Five EC patient cohorts from 3 different Dutch centres were used for model validation. External validity of the model was assessed separately in definitive (n = 170) and neo-adjuvant (n = 568) chemoradiotherapy (dCRT and nCRT) patients. External validity was assessed in terms of calibration by calibration plots, calibration-in-the-large (CITL) and calibration slope (CS), and discrimination by assessment of the c-statistic. If suboptimal model performance was observed, the model was further updated accordingly.

RESULTS

For the dCRT patients, good calibration was found after adjustment of the intercept (CITL 0.00; CS 1.08). The c-statistic of the adjusted model was 0.67 (95%CI: 0.58 to 0.75). For nCRT patients the model needed adjustment of both the slope and the intercept because of initial miscalibration in the validation population (CITL 0.00; CS 1.72). After recalibration, the model showed perfect calibration (i.e., CITL 0, CS 1), as is common after recalibration. The c-statistic of the recalibrated model equaled 0.62 (95%CI: 0.57 to 0.67).

CONCLUSION

The existing model for 2-year mortality prediction in lung cancer patients, based on the predictive factors MHD and GTV, showed good performance in EC patients after updating the intercept and/or slope of the original model.

Comparing the performance of a deep learning-based lung gross tumour volume segmentation algorithm before and after transfer learning in a new hospital

Objectives

Radiation therapy for lung cancer requires a gross tumour volume (GTV) to be carefully outlined by a skilled radiation oncologist (RO) to accurately pinpoint high radiation dose to a malignant mass while simultaneously minimizing radiation damage to adjacent normal tissues. This is manually intensive and tedious however, it is feasible to train a deep learning (DL) neural network that could assist ROs to delineate the GTV. However, DL trained on large openly accessible data sets might not perform well when applied to a superficially similar task but in a different clinical setting. In this work, we tested the performance of DL automatic lung GTV segmentation model trained on open-access Dutch data when used on Indian patients from a large public tertiary hospital, and hypothesized that generic DL performance could be improved for a specific local clinical context, by means of modest transfer-learning on a small representative local subset.

Methods

X-ray computed tomography (CT) series in a public data set called "NSCLC-Radiomics" from The Cancer Imaging Archive was first used to train a DL-based lung GTV segmentation model (Model 1). Its performance was assessed using a different open access data set (Interobserver1) of Dutch subjects plus a private Indian data set from a local tertiary hospital (Test Set 2). Another Indian data set (Retrain Set 1) was used to fine-tune the former DL model using a transfer learning method. The Indian data sets were taken from CT of a hybrid scanner based in nuclear medicine, but the GTV was drawn by skilled Indian ROs. The final (after fine-tuning) model (Model 2) was then re-evaluated in "Interobserver1" and "Test Set 2." Dice similarity coefficient (DSC), precision, and recall were used as geometric segmentation performance metrics.

Results

Model 1 trained exclusively on Dutch scans showed a significant fall in performance when tested on "Test Set 2." However, the DSC of Model 2 recovered by 14 percentage points when evaluated in the same test set. Precision and recall showed a similar rebound of performance after transfer learning, in spite of using a comparatively small sample size. The performance of both models, before and after the fine-tuning, did not significantly change the segmentation performance in "Interobserver1."

Conclusions

A large public open-access data set was used to train a generic DL model for lung GTV segmentation, but this did not perform well initially in the Indian clinical context. Using transfer learning methods, it was feasible to efficiently and easily fine-tune the generic model using only a small number of local examples from the Indian hospital. This led to a recovery of some of the geometric segmentation performance, but the tuning did not appear to affect the performance of the model in another open-access data set.

Advances in knowledge

Caution is needed when using models trained on large volumes of international data in a local clinical setting, even when that training data set is of good quality. Minor differences in scan acquisition and clinician delineation preferences may result in an apparent drop in performance. However, DL models have the advantage of being efficiently "adapted" from a generic to a locally specific context, with only a small amount of fine-tuning by means of transfer learning on a small local institutional data set.

External Validation of Robust Radiomic Signature to Predict 2-Year Overall Survival in Non-Small-Cell Lung Cancer

Lung cancer is the second most fatal disease worldwide. In the last few years, radiomics is being explored to develop prediction models for various clinical endpoints in lung cancer. However, the robustness of radiomic features is under question and has been identified as one of the roadblocks in the implementation of a radiomic-based prediction model in the clinic. Many past studies have suggested identifying the robust radiomic feature to develop a prediction model. In our earlier study, we identified robust radiomic features for prediction model development. The objective of this study was to develop and validate the robust radiomic signatures for predicting 2-year overall survival in non-small cell lung cancer (NSCLC). This retrospective study included a cohort of 300 stage I-IV NSCLC patients. Institutional 200 patients' data were included for training and internal validation and 100 patients' data from The Cancer Image Archive (TCIA) open-source image repository for external validation. Radiomic features were extracted from the CT images of both cohorts. The feature selection was performed using hierarchical clustering, a Chi-squared test, and recursive feature elimination (RFE). In total, six prediction models were developed using random forest (RF-Model-O, RF-Model-B), gradient boosting (GB-Model-O, GB-Model-B), and support vector(SV-Model-O, SV-Model-B) classifiers to predict 2-year overall survival (OS) on original data as well as balanced data. Model validation was performed using 10-fold cross-validation, internal validation, and external validation. Using a multistep feature selection method, the overall top 10 features were chosen. On internal validation, the two random forest models (RF-Model-O, RF-Model-B) displayed the highest accuracy; their scores on the original and balanced datasets were 0.81 and 0.77 respectively. During external validation, both the random forest models' accuracy was 0.68. In our study, robust radiomic features showed promising predictive performance to predict 2-year overall survival in NSCLC.

Image based prognosis in head and neck cancer using convolutional neural networks: a case study in reproducibility and optimization

In the past decade, there has been a sharp increase in publications describing applications of convolutional neural networks (CNNs) in medical image analysis. However, recent reviews have warned of the lack of reproducibility of most such studies, which has impeded closer examination of the models and, in turn, their implementation in healthcare. On the other hand, the performance of these models is highly dependent on decisions on architecture and image pre-processing. In this work, we assess the reproducibility of three studies that use CNNs for head and neck cancer outcome prediction by attempting to reproduce the published results. In addition, we propose a new network structure and assess the impact of image pre-processing and model selection criteria on performance. We used two publicly available datasets: one with 298 patients for training and validation and another with 137 patients from a different institute for testing. All three studies failed to report elements required to reproduce their results thoroughly, mainly the image pre-processing steps and the random seed. Our model either outperforms or achieves similar performance to the existing models with considerably fewer parameters. We also observed that the pre-processing efforts significantly impact the model's performance and that some model selection criteria may lead to suboptimal models. Although there have been improvements in the reproducibility of deep learning models, our work suggests that wider implementation of reporting standards is required to avoid a reproducibility crisis.

Emerging role of quantitative imaging (radiomics) and artificial intelligence in precision oncology

Cancer is a fatal disease and the second most cause of death worldwide. Treatment of cancer is a complex process and requires a multi-modality-based approach. Cancer detection and treatment starts with screening/diagnosis and continues till the patient is alive. Screening/diagnosis of the disease is the beginning of cancer management and continued with the staging of the disease, planning and delivery of treatment, treatment monitoring, and ongoing monitoring and follow-up. Imaging plays an important role in all stages of cancer management. Conventional oncology practice considers that all patients are similar in a disease type, whereas biomarkers subgroup the patients in a disease type which leads to the development of precision oncology. The utilization of the radiomic process has facilitated the advancement of diverse imaging biomarkers that find application in precision oncology. The role of imaging biomarkers and artificial intelligence (AI) in oncology has been investigated by many researchers in the past. The existing literature is suggestive of the increasing role of imaging biomarkers and AI in oncology. However, the stability of radiomic features has also been questioned. The radiomic community has recognized that the instability of radiomic features poses a danger to the global generalization of radiomic-based prediction models. In order to establish radiomic-based imaging biomarkers in oncology, the robustness of radiomic features needs to be established on a priority basis. This is because radiomic models developed in one institution frequently perform poorly in other institutions, most likely due to radiomic feature instability. To generalize radiomic-based prediction models in oncology, a number of initiatives, including Quantitative Imaging Network (QIN), Quantitative Imaging Biomarkers Alliance (QIBA), and Image Biomarker Standardisation Initiative (IBSI), have been launched to stabilize the radiomic features.

Identifying radiomics signatures in body composition imaging for the prediction of outcome following pancreatic cancer resection

Background

Computerized radiological image analysis (radiomics) enables the investigation of image-derived phenotypes by extracting large numbers of quantitative features. We hypothesized that radiomics features may contain prognostic information that enhances conventional body composition analysis. We aimed to investigate whether body composition-associated radiomics features hold additional value over conventional body composition analysis and clinical patient characteristics used to predict survival of pancreatic ductal adenocarcinoma (PDAC) patients.

Methods

Computed tomography images of 304 patients undergoing elective pancreatic cancer resection were analysed. 2D radiomics features were extracted from skeletal muscle and subcutaneous and visceral adipose tissue (SAT and VAT) compartments from a single slice at the third lumbar vertebra. The study population was randomly split (80:20) into training and holdout subsets. Feature ranking with Least Absolute Shrinkage Selection Operator (LASSO) followed by multivariable stepwise Cox regression in 1000 bootstrapped re-samples of the training data was performed and tested on the holdout data. The fitted regression predictors were used as "scores" for a clinical (C-Score), body composition (B-Score), and radiomics (R-Score) model. To stratify patients into the highest 25% and lowest 25% risk of mortality compared to the middle 50%, the Harrell Concordance Index was used.

Results

Based on LASSO and stepwise cox regression for overall survival, ASA ≥3 and age were the most important clinical variables and constituted the C-score, and VAT-index (VATI) was the most important body composition variable and constituted the B-score. Three radiomics features (SATI_original_shape2D_Perimeter, VATI_original_glszm_SmallAreaEmphasis, and VATI_original_firstorder_Maximum) emerged as the most frequent set of features and yielded an R-Score. Of the mean concordance indices of C-, B-, and R-scores, R-score performed best (0.61, 95% CI 0.56-0.65, p<0.001), followed by the C-score (0.59, 95% CI 0.55-0.63, p<0.001) and B-score (0.55, 95% CI 0.50-0.60, p=0.03). Kaplan-Meier projection revealed that C-, B, and R-scores showed a clear split in the survival curves in the training set, although none remained significant in the holdout set.

Conclusion

It is feasible to implement a data-driven radiomics approach to body composition imaging. Radiomics features provided improved predictive performance compared to conventional body composition variables for the prediction of overall survival of PDAC patients undergoing primary resection.

Clinical Concept-Based Radiology Reports Classification Pipeline for Lung Carcinoma

Rising incidence and mortality of cancer have led to an incremental amount of research in the field. To learn from preexisting data, it has become important to capture maximum information related to disease type, stage, treatment, and outcomes. Medical imaging reports are rich in this kind of information but are only present as free text. The extraction of information from such unstructured text reports is labor-intensive. The use of Natural Language Processing (NLP) tools to extract information from radiology reports can make it less time-consuming as well as more effective. In this study, we have developed and compared different models for the classification of lung carcinoma reports using clinical concepts. This study was approved by the institutional ethics committee as a retrospective study with a waiver of informed consent. A clinical concept-based classification pipeline for lung carcinoma radiology reports was developed using rule-based as well as machine learning models and compared. The machine learning models used were XGBoost and two more deep learning model architectures with bidirectional long short-term neural networks. A corpus consisting of 1700 radiology reports including computed tomography (CT) and positron emission tomography/computed tomography (PET/CT) reports were used for development and testing. Five hundred one radiology reports from MIMIC-III Clinical Database version 1.4 was used for external validation. The pipeline achieved an overall F1 score of 0.94 on the internal set and 0.74 on external validation with the rule-based algorithm using expert input giving the best performance. Among the machine learning models, the Bi-LSTM_dropout model performed better than the ML model using XGBoost and the Bi-LSTM_simple model on internal set, whereas on external validation, the Bi-LSTM_simple model performed relatively better than other 2. This pipeline can be used for clinical concept-based classification of radiology reports related to lung carcinoma from a huge corpus and also for automated annotation of these reports.

Systematic review and meta-analysis of prediction models used in cervical cancer

BACKGROUND

Cervical cancer is one of the most common cancers in women with an incidence of around 6.5 % of all the cancer in women worldwide. Early detection and adequate treatment according to staging improve the patient's life expectancy. Outcome prediction models might aid treatment decisions, but a systematic review on prediction models for cervical cancer patients is not available.

DESIGN

We performed a systematic review for prediction models in cervical cancer following PRISMA guidelines. Key features that were used for model training and validation, the endpoints were extracted from the article and data were analyzed. Selected articles were grouped based on prediction endpoints i.e. Group1: Overall survival, Group2: progression-free survival; Group3: recurrence or distant metastasis; Group4: treatment response; Group5: toxicity or quality of life. We developed a scoring system to evaluate the manuscript. As per our criteria, studies were divided into four groups based on scores obtained in our scoring system, the Most significant study (Score > 60 %); Significant study (60 % > Score > 50 %); Moderately Significant study (50 % > Score > 40 %); least significant study (score < 40 %). A meta-analysis was performed for all the groups separately.

RESULTS

The first line of search selected 1358 articles and finally 39 articles were selected as eligible for inclusion in the review. As per our assessment criteria, 16, 13 and 10 studies were found to be the most significant, significant and moderately significant respectively. The intra-group pooled correlation coefficient for Group1, Group2, Group3, Group4, and Group5 were 0.76 [0.72, 0.79], 0.80 [0.73, 0.86], 0.87 [0.83, 0.90], 0.85 [0.77, 0.90], 0.88 [0.85, 0.90] respectively. All the models were found to be good (prediction accuracy [c-index/AUC/R²] >0.7) in endpoint prediction.

CONCLUSIONS

Prediction models of cervical cancer toxicity, local or distant recurrence and survival prediction show promising results with reasonable prediction accuracy [c-index/AUC/R² > 0.7]. These models should also be validated on external data and evaluated in prospective clinical studies.

Using 3D deep features from CT scans for cancer prognosis based on a video classification model: A multi-dataset feasibility study

BACKGROUND

Cancer prognosis before and after treatment is key for patient management and decision making. Handcrafted imaging biomarkers-radiomics-have shown potential in predicting prognosis.

PURPOSE

However, given the recent progress in deep learning, it is timely and relevant to pose the question: could deep learning based 3D imaging features be used as imaging biomarkers and outperform radiomics?

METHODS

Effectiveness, reproducibility in test/retest, across modalities, and correlation of deep features with clinical features such as tumor volume and TNM staging were tested in this study. Radiomics was introduced as the reference image biomarker. For deep feature extraction, we transformed the CT scans into videos, and we adopted the pre-trained Inflated 3D ConvNet (I3D) video classification network as the architecture. We used four datasets-LUNG 1 (n = 422), LUNG 4 (n = 106), OPC (n = 605), and H&N 1 (n = 89)-with 1270 samples from different centers and cancer types-lung and head and neck cancer-to test deep features' predictiveness and two additional datasets to assess the reproducibility of deep features.

RESULTS

Support Vector Machine-Recursive Feature Elimination (SVM-RFE) selected top 100 deep features achieved a concordance index (CI) of 0.67 in survival prediction in LUNG 1, 0.87 in LUNG 4, 0.76 in OPC, and 0.87 in H&N 1, while SVM-RFE selected top 100 radiomics achieved CIs of 0.64, 0.77, 0.73, and 0.74, respectively, all statistically significant differences (p < 0.01, Wilcoxon's test). Most selected deep features are not correlated with tumor volume and TNM staging. However, full radiomics features show higher reproducibility than full deep features in a test/retest setting (0.89 vs. 0.62, concordance correlation coefficient).

CONCLUSION

The results show that deep features can outperform radiomics while providing different views for tumor prognosis compared to tumor volume and TNM staging. However, deep features suffer from lower reproducibility than radiomic features and lack the interpretability of the latter.

Multi-centre radiomics for prediction of recurrence following radical radiotherapy for head and neck cancers: Consequences of feature selection, machine learning classifiers and batch-effect harmonization

Background and purpose

Radiomics models trained with limited single institution data are often not reproducible and generalisable. We developed radiomics models that predict loco-regional recurrence within two years of radiotherapy with private and public datasets and their combinations, to simulate small and multi-institutional studies and study the responsiveness of the models to feature selection, machine learning algorithms, centre-effect harmonization and increased dataset sizes.

Materials and methods

562 patients histologically confirmed and treated for locally advanced head-and-neck cancer (LA-HNC) from two public and two private datasets; one private dataset exclusively reserved for validation. Clinical contours of primary tumours were not recontoured and were used for Pyradiomics based feature extraction. ComBat harmonization was applied, and LASSO-Logistic Regression (LR) and Support Vector Machine (SVM) models were built. 95% confidence interval (CI) of 1000 bootstrapped area-under-the-Receiver-operating-curves (AUC) provided predictive performance. Responsiveness of the models' performance to the choice of feature selection methods, ComBat harmonization, machine learning classifier, single and pooled data was evaluated.

Results

LASSO and SelectKBest selected 14 and 16 features, respectively; three were overlapping. Without ComBat, the LR and SVM models for three institutional data showed AUCs (CI) of 0.513 (0.481-0.559) and 0.632 (0.586-0.665), respectively. Performances following ComBat revealed AUCs of 0.559 (0.536-0.590) and 0.662 (0.606-0.690), respectively. Compared to single cohort AUCs (0.562-0.629), SVM models from pooled data performed significantly better at AUC = 0.680.

Conclusions

Multi-institutional retrospective data accentuates the existing variabilities that affect radiomics. Carefully designed prospective, multi-institutional studies and data sharing are necessary for clinically relevant head-and-neck cancer prognostication models.

Radiomics and Dosiomics Signature From Whole Lung Predicts Radiation Pneumonitis: A Model Development Study With Prospective External Validation and Decision-curve Analysis

PURPOSE

Radiation pneumonitis (RP) is one of the common side effects of radiation therapy in the thoracic region. Radiomics and dosiomics quantify information implicit within medical images and radiation therapy dose distributions. In this study we demonstrate the prognostic potential of radiomics, dosiomics, and clinical features for RP prediction.

METHODS AND MATERIALS

Radiomics, dosiomics, dose-volume histogram (DVH) metrics, and clinical parameters were obtained on 314 retrospectively collected and 35 prospectively enrolled patients diagnosed with lung cancer between 2013 to 2019. A radiomics risk score (R score) and dosiomics risk score (D score), as well as a DVH-score, were calculated based on logistic regression after feature selection. Six models were built using different combinations of R score, D score, DVH score, and clinical parameters to evaluate their added prognostic power. Overoptimism was evaluated by bootstrap resampling from the training set, and the prospectively collected cohort was used as the external test set. Model calibration and decision-curve characteristics of the best-performing models were evaluated. For ease of further evaluation, nomograms were constructed for selected models.

RESULTS

A model built by integrating all of the R score, D score, and clinical parameters had the best discriminative ability with areas under the curve of 0.793 (95% confidence interval [CI], 0.735-0.851), 0.774 (95% CI, 0.762-0.786), and 0.855 (95% CI, 0.719-0.990) in the training, bootstrapping, and external test sets, respectively. The calibration curve image showed good agreement between the predicted and actual values, with a slope of 1.21 and intercept of -0.04. The decision curve image showed a positive net benefit for the final model based on the nomogram.

CONCLUSIONS

Radiomic and dosiomic features have the potential to assist with the prediction of RP, and the combination of radiomics, dosiomics, and clinical parameters led to the best prognostic model in the present study.

Computed tomography and radiation dose images-based deep-learning model for predicting radiation pneumonitis in lung cancer patients after radiation therapy

PURPOSE

To develop a deep learning model that combines CT and radiation dose (RD) images to predict the occurrence of radiation pneumonitis (RP) in lung cancer patients who received radical (chemo)radiotherapy.

METHODS

CT, RD images and clinical parameters were obtained from 314 retrospectively-collected patients (training set) and 35 prospectively-collected patients (test-set-1) who were diagnosed with lung cancer and received radical radiotherapy in the dose range of 50 Gy and 70 Gy. Another 194 (60 Gy group, test-set-2) and 158 (74 Gy group, test-set-3) patients from the clinical trial RTOG 0617 were used for external validation. A ResNet architecture was used to develop a prediction model that combines CT and RD features. Thereafter, the CT and RD weights were adjusted by using 40 patients from test-set-2 or 3 to accommodate cohorts with different clinical settings or dose delivery patterns. Visual interpretation was implemented using a gradient-weighted class activation map (grad-CAM) to observe the area of model attention during the prediction process. To improve the usability, ready-to-use online software was developed.

RESULTS

The discriminative ability of a baseline trained model had an AUC of 0.83 for test-set-1, 0.55 for test-set-2, and 0.63 for test-set-3. After adjusting CT and RD weights of the model using a subset of the RTOG-0617 subjects, the discriminatory power of test-set-2 and 3 improved to AUC 0.65 and AUC 0.70, respectively. Grad-CAM showed the regions of interest to the model that contribute to the prediction of RP.

CONCLUSION

A novel deep learning approach combining CT and RD images can effectively and accurately predict the occurrence of RP, and this model can be adjusted easily to fit new cohorts.

Deep learning based unpaired image-to-image translation applications for medical physics: a systematic review

Purpose. There is a growing number of publications on the application of unpaired image-to-image (I2I) translation in medical imaging. However, a systematic review covering the current state of this topic for medical physicists is lacking. The aim of this article is to provide a comprehensive review of current challenges and opportunities for medical physicists and engineers to apply I2I translation in practice.Methods and materials. The PubMed electronic database was searched using terms referring to unpaired (unsupervised), I2I translation, and medical imaging. This review has been reported in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. From each full-text article, we extracted information extracted regarding technical and clinical applications of methods, Transparent Reporting for Individual Prognosis Or Diagnosis (TRIPOD) study type, performance of algorithm and accessibility of source code and pre-trained models.Results. Among 461 unique records, 55 full-text articles were included in the review. The major technical applications described in the selected literature are segmentation (26 studies), unpaired domain adaptation (18 studies), and denoising (8 studies). In terms of clinical applications, unpaired I2I translation has been used for automatic contouring of regions of interest in MRI, CT, x-ray and ultrasound images, fast MRI or low dose CT imaging, CT or MRI only based radiotherapy planning, etc Only 5 studies validated their models using an independent test set and none were externally validated by independent researchers. Finally, 12 articles published their source code and only one study published their pre-trained models.Conclusion. I2I translation of medical images offers a range of valuable applications for medical physicists. However, the scarcity of external validation studies of I2I models and the shortage of publicly available pre-trained models limits the immediate applicability of the proposed methods in practice.

Longitudinal Doppler Assessments in Late Preterm Fetal Growth Restriction

PURPOSE

 To assess the longitudinal variation of the ratio of umbilical and cerebral artery pulsatility index (UCR) in late preterm fetal growth restriction (FGR).

MATERIALS AND METHODS

 A prospective European multicenter observational study included women with a singleton pregnancy, 32^+ 0-36^+ 6, at risk of FGR (estimated fetal weight [EFW] or abdominal circumference [AC] < 10^th percentile, abnormal arterial Doppler or fall in AC from 20-week scan of > 40 percentile points). The primary outcome was a composite of abnormal condition at birth or major neonatal morbidity. UCR was categorized as normal (< 0.9) or abnormal (≥ 0.9). UCR was assessed by gestational age at measurement interval to delivery, and by individual linear regression coefficient in women with two or more measurements.

RESULTS

 856 women had 2770 measurements; 696 (81 %) had more than one measurement (median 3 (IQR 2-4). At inclusion, 63 (7 %) a UCR ≥ 0.9. These delivered earlier and had a lower birth weight and higher incidence of adverse outcome (30 % vs. 9 %, relative risk 3.2; 95 %CI 2.1-5.0) than women with a normal UCR at inclusion. Repeated measurements after an abnormal UCR at inclusion were abnormal again in 67 % (95 %CI 55-80), but after a normal UCR the chance of finding an abnormal UCR was 6 % (95 %CI 5-7 %). The risk of composite adverse outcome was similar using the first or subsequent UCR values.

CONCLUSION

 An abnormal UCR is likely to be abnormal again at a later measurement, while after a normal UCR the chance of an abnormal UCR is 5-7 % when repeated weekly. Repeated measurements do not predict outcome better than the first measurement, most likely due to the most compromised fetuses being delivered after an abnormal UCR.

Generation of synthetic ground glass nodules using generative adversarial networks (GANs)

BACKGROUND

Data shortage is a common challenge in developing computer-aided diagnosis systems. We developed a generative adversarial network (GAN) model to generate synthetic lung lesions mimicking ground glass nodules (GGNs).

METHODS

We used 216 computed tomography images with 340 GGNs from the Lung Image Database Consortium and Image Database Resource Initiative database. A GAN model retrieving information from the whole image and the GGN region was built. The generated samples were evaluated with visual Turing test performed by four experienced radiologists or pulmonologists. Radiomic features were compared between real and synthetic nodules. Performances were evaluated by area under the curve (AUC) at receiver operating characteristic analysis. In addition, we trained a classification model (ResNet) to investigate whether the synthetic GGNs can improve the performances algorithm and how performances changed as a function of labelled data used in training.

RESULTS

Of 51 synthetic GGNs, 19 (37%) were classified as real by clinicians. Of 93 radiomic features, 58 (62.4%) showed no significant difference between synthetic and real GGNs (p ≥ 0.052). The discrimination performances of physicians (AUC 0.68) and radiomics (AUC 0.66) were similar, with no-significantly different (p = 0.23), but clinicians achieved a better accuracy (AUC 0.74) than radiomics (AUC 0.62) (p < 0.001). The classification model trained on datasets with synthetic data performed better than models without the addition of synthetic data.

CONCLUSIONS

GAN has promising potential for generating GGNs. Through similar AUC, clinicians achieved better ability to diagnose whether the data is synthetic than radiomics.

Clinical evaluation of automated segmentation for body composition analysis on abdominal L3 CT slices in polytrauma patients

INTRODUCTION

Sarcopenia is a muscle disease that involves loss of muscle strength and physical function and is associated with adverse health effects. Even though sarcopenia has attracted increasing attention in the literature, many research findings have not yet been translated into clinical practice. In this article, we aim to validate a deep learning neural network for automated segmentation of L3 CT slices and aim to explore the potential for clinical utilization of such a tool for clinical practice.

MATERIALS AND METHODS

A deep learning neural network was trained on a multi-centre collection of 3413 abdominal cancer surgery subjects to automatically segment muscle, subcutaneous and visceral adipose tissue at the L3 lumbar vertebral level. 536 Polytrauma subjects were used as an independent test set to show generalizability. The Dice Similarity Coefficient was calculated to validate the geometric similarity. Quantitative agreement was quantified using Bland-Altman's Limits of Agreement interval and Lin's Concordance Correlation Coefficient. To determine the potential clinical usability, randomly selected segmentation images were presented to a panel of experienced clinicians to rate on a Likert scale.

RESULTS

Deep learning results gave excellent agreement versus a human expert operator for all of the body composition indices, with Concordance Correlation Coefficient for skeletal muscle index of 0.92, Skeletal muscle radiation attenuation 0.94, Visceral Adipose Tissue index 0.99 and Subcutaneous Adipose Tissue Index 0.99. Triple-blinded visual assessment of segmentation by clinicians correlated only to the Dice coefficient, but had no association to quantitative body composition metrics which were accurate irrespective of clinicians' visual rating.

CONCLUSION

A deep learning method for automatic segmentation of truncal muscle, visceral and subcutaneous adipose tissue on individual L3 CT slices has been independently validated against expert human-generated results for an enlarged polytrauma registry dataset. Time efficiency, consistency and high accuracy relative to human experts suggest that quantitative body composition analysis with deep learning should is a promising tool for clinical application in a hospital setting.

Two-Compartment or 4-Compartment Fasciotomy for Lower Leg Chronic Exertional Compartment Syndrome: A Systematic Review

Patients with lower leg chronic exertional compartment syndrome are impaired due to exercise-related pain. Fasciotomy is the surgical gold standard. However, it is unknown whether number of simultaneously opened compartments affects outcome. The purpose of this systematic review was to compare patient-reported outcomes of a 2-compartment fasciotomy with a 4-compartment fasciotomy. Controlled clinical trials (randomized/nonrandomized), cohort studies and case series reporting on outcome following either 2-compartment or 4-compartment fasciotomy for lower leg chronic exertional compartment syndrome were searched until May 31, 2021 in PubMed, EMBASE, and Cochrane. Results were qualitatively synthesized. Risk of bias and levels of evidence were determined. Seven studies reporting on altogether 194 athletes and military personnel (mean age 24 y) were included. Quality assessment revealed a high risk of bias in all studies. Both 2-compartment and 4-compartment fasciotomy were associated with a 50% to 100% "return to activity" rate (in studies reporting group results separately: 2-compartment 90%-100%; 4-compartment 50%-100%) and a 41% to 100% "return to previous activity" rate (in studies reporting group results separately: 2-compartment 82-100%; 4-compartment 50%-100%) without significant differences. Mean Marx activity score of 1 study found a small significant standardized mean difference (0.196 [0.524,0.916]) favoring 4-compartment fasciotomy. Rate of satisfaction (2-compartment 74%-89%; 4-compartment 75%-100%) and residual symptoms (2-compartment 0%-36%; 4-compartment 0%-50%) indicated no group differences. In conclusion, a 2-compartment fasciotomy or a 4-compartment fasciotomy for lower leg chronic exertional compartment syndrome appears to be equally successful. However, included studies were hampered by methodological shortcomings (low sample size, selection bias, heterogeneity and no uniform outcome measures).

Development and validation of prognostic models for anal cancer outcomes using distributed learning: protocol for the international multi-centre atomCAT2 study

BACKGROUND

Anal cancer is a rare cancer with rising incidence. Despite the relatively good outcomes conferred by state-of-the-art chemoradiotherapy, further improving disease control and reducing toxicity has proven challenging. Developing and validating prognostic models using routinely collected data may provide new insights for treatment development and selection. However, due to the rarity of the cancer, it can be difficult to obtain sufficient data, especially from single centres, to develop and validate robust models. Moreover, multi-centre model development is hampered by ethical barriers and data protection regulations that often limit accessibility to patient data. Distributed (or federated) learning allows models to be developed using data from multiple centres without any individual-level patient data leaving the originating centre, therefore preserving patient data privacy. This work builds on the proof-of-concept three-centre atomCAT1 study and describes the protocol for the multi-centre atomCAT2 study, which aims to develop and validate robust prognostic models for three clinically important outcomes in anal cancer following chemoradiotherapy.

METHODS

This is a retrospective multi-centre cohort study, investigating overall survival, locoregional control and freedom from distant metastasis after primary chemoradiotherapy for anal squamous cell carcinoma. Patient data will be extracted and organised at each participating radiotherapy centre (n = 18). Candidate prognostic factors have been identified through literature review and expert opinion. Summary statistics will be calculated and exchanged between centres prior to modelling. The primary analysis will involve developing and validating Cox proportional hazards models across centres for each outcome through distributed learning. Outcomes at specific timepoints of interest and factor effect estimates will be reported, allowing for outcome prediction for future patients.

DISCUSSION

The atomCAT2 study will analyse one of the largest available cross-institutional cohorts of patients with anal cancer treated with chemoradiotherapy. The analysis aims to provide information on current international clinical practice outcomes and may aid the personalisation and design of future anal cancer clinical trials through contributing to a better understanding of patient risk stratification.

Improving reproducibility and performance of radiomics in low-dose CT using cycle GANs

Background

As a means to extract biomarkers from medical imaging, radiomics has attracted increased attention from researchers. However, reproducibility and performance of radiomics in low‐dose CT scans are still poor, mostly due to noise. Deep learning generative models can be used to denoise these images and in turn improve radiomics’ reproducibility and performance. However, most generative models are trained on paired data, which can be difficult or impossible to collect.

Purpose

In this article, we investigate the possibility of denoising low‐dose CTs using cycle generative adversarial networks (GANs) to improve radiomics reproducibility and performance based on unpaired datasets.

Methods and materials

Two cycle GANs were trained: (1) from paired data, by simulating low‐dose CTs (i.e., introducing noise) from high‐dose CTs and (2) from unpaired real low dose CTs. To accelerate convergence, during GAN training, a slice‐paired training strategy was introduced. The trained GANs were applied to three scenarios: (1) improving radiomics reproducibility in simulated low‐dose CT images and (2) same‐day repeat low dose CTs (RIDER dataset), and (3) improving radiomics performance in survival prediction. Cycle GAN results were compared with a conditional GAN (CGAN) and an encoder–decoder network (EDN) trained on simulated paired data.

Results

The cycle GAN trained on simulated data improved concordance correlation coefficients (CCC) of radiomic features from 0.87 (95%CI, [0.833,0.901]) to 0.93 (95%CI, [0.916,0.949]) on simulated noise CT and from 0.89 (95%CI, [0.881,0.914]) to 0.92 (95%CI, [0.908,0.937]) on the RIDER dataset, as well improving the area under the receiver operating characteristic curve (AUC) of survival prediction from 0.52 (95%CI, [0.511,0.538]) to 0.59 (95%CI, [0.578,0.602]). The cycle GAN trained on real data increased the CCCs of features in RIDER to 0.95 (95%CI, [0.933,0.961]) and the AUC of survival prediction to 0.58 (95%CI, [0.576,0.596]).

Conclusion

The results show that cycle GANs trained on both simulated and real data can improve radiomics’ reproducibility and performance in low‐dose CT and achieve similar results compared to CGANs and EDNs.

The impact of inter-observer variation in delineation on robustness of radiomics features in non-small cell lung cancer

Artificial intelligence and radiomics have the potential to revolutionise cancer prognostication and personalised treatment. Manual outlining of the tumour volume for extraction of radiomics features (RF) is a subjective process. This study investigates robustness of RF to inter-observer variation (IOV) in contouring in lung cancer. We utilised two public imaging datasets: ‘NSCLC-Radiomics’ and ‘NSCLC-Radiomics-Interobserver1’ (‘Interobserver’). For ‘NSCLC-Radiomics’, we created an additional set of manual contours for 92 patients, and for ‘Interobserver’, there were five manual and five semi-automated contours available for 20 patients. Dice coefficients (DC) were calculated for contours. 1113 RF were extracted including shape, first order and texture features. Intraclass correlation coefficient (ICC) was computed to assess robustness of RF to IOV. Cox regression analysis for overall survival (OS) was performed with a previously published radiomics signature. The median DC ranged from 0.81 (‘NSCLC-Radiomics’) to 0.85 (‘Interobserver’—semi-automated). The median ICC for the ‘NSCLC-Radiomics’, ‘Interobserver’ (manual) and ‘Interobserver’ (semi-automated) were 0.90, 0.88 and 0.93 respectively. The ICC varied by feature type and was lower for first order and gray level co-occurrence matrix (GLCM) features. Shape features had a lower median ICC in the ‘NSCLC-Radiomics’ dataset compared to the ‘Interobserver’ dataset. Survival analysis showed similar separation of curves for three of four RF apart from ‘original_shape_Compactness2’, a feature with low ICC (0.61). The majority of RF are robust to IOV, with first order, GLCM and shape features being the least robust. Semi-automated contouring improves feature stability. Decreased robustness of a feature is significant as it may impact upon the features’ prognostic capability.

Automatic Tumor Grading on Colorectal Cancer Whole-Slide Images: Semi-Quantitative Gland Formation Percentage and New Indicator Exploration

Tumor grading is an essential factor for cancer staging and survival prognostication. The widely used the WHO grading system defines the histological grade of CRC adenocarcinoma based on the density of glandular formation on whole-slide images (WSIs). We developed a fully automated approach for stratifying colorectal cancer (CRC) patients' risk of mortality directly from histology WSI relating to gland formation. A tissue classifier was trained to categorize regions on WSI as glands, stroma, immune cells, background, and other tissues. A gland formation classifier was trained on expert annotations to categorize regions as different degrees of tumor gland formation versus normal tissues. The glandular formation density can thus be estimated using the aforementioned tissue categorization and gland formation information. This estimation was called a semi-quantitative gland formation ratio (SGFR), which was used as a prognostic factor in survival analysis. We evaluated gland formation percentage and validated it by comparing it against the WHO cutoff point. Survival data and gland formation maps were then used to train a spatial pyramid pooling survival network (SPPSN) as a deep survival model. We compared the survival prediction performance of estimated gland formation percentage and the SPPSN deep survival grade and found that the deep survival grade had improved discrimination. A univariable Cox model for survival yielded moderate discrimination with SGFR (c-index 0.62) and deep survival grade (c-index 0.64) in an independent institutional test set. Deep survival grade also showed better discrimination performance in multivariable Cox regression. The deep survival grade significantly increased the c-index of the baseline Cox model in both validation set and external test set, but the inclusion of SGFR can only improve the Cox model less in external test and is unable to improve the Cox model in the validation set.

Radiomics: a quantitative imaging biomarker in precision oncology

Cancer treatment is heading towards precision medicine driven by genetic and biochemical markers. Various genetic and biochemical markers are utilized to render personalized treatment in cancer. In the last decade, noninvasive imaging biomarkers have also been developed to assist personalized decision support systems in oncology. The imaging biomarkers i.e., radiomics is being researched to develop specific digital phenotype of tumor in cancer. Radiomics is a process to extract high throughput data from medical images by using advanced mathematical and statistical algorithms. The radiomics process involves various steps i.e., image generation, segmentation of region of interest (e.g. a tumor), image preprocessing, radiomic feature extraction, feature analysis and selection and finally prediction model development. Radiomics process explores the heterogeneity, irregularity and size parameters of the tumor to calculate thousands of advanced features. Our study investigates the role of radiomics in precision oncology. Radiomics research has witnessed a rapid growth in the last decade with several studies published that show the potential of radiomics in diagnosis and treatment outcome prediction in oncology. Several radiomics based prediction models have been developed and reported in the literature to predict various prediction endpoints i.e., overall survival, progression-free survival and recurrence in various cancer i.e., brain tumor, head and neck cancer, lung cancer and several other cancer types. Radiomics based digital phenotypes have shown promising results in diagnosis and treatment outcome prediction in oncology. In the coming years, radiomics is going to play a significant role in precision oncology.

Generative models improve radiomics performance in different tasks and different datasets: An experimental study

PURPOSE

Radiomics is an active area of research focusing on high throughput feature extraction from medical images with a wide array of applications in clinical practice, such as clinical decision support in oncology. However, noise in low dose computed tomography (CT) scans can impair the accurate extraction of radiomic features. In this article, we investigate the possibility of using deep learning generative models to improve the performance of radiomics from low dose CTs.

METHODS

We used two datasets of low dose CT scans - NSCLC Radiogenomics and LIDC-IDRI - as test datasets for two tasks - pre-treatment survival prediction and lung cancer diagnosis. We used encoder-decoder networks and conditional generative adversarial networks (CGANs) trained in a previous study as generative models to transform low dose CT images into full dose CT images. Radiomic features extracted from the original and improved CT scans were used to build two classifiers - a support vector machine (SVM) and a deep attention based multiple instance learning model - for survival prediction and lung cancer diagnosis respectively. Finally, we compared the performance of the models derived from the original and improved CT scans.

RESULTS

Denoising with the encoder-decoder network and the CGAN improved the area under the curve (AUC) of survival prediction from 0.52 to 0.57 (p-value < 0.01). On the other hand, the encoder-decoder network and the CGAN improved the AUC of lung cancer diagnosis from 0.84 to 0.88 and 0.89 respectively (p-value < 0.01). Finally, there are no statistically significant improvements in AUC using encoder-decoder networks and CGAN (p-value = 0.34) when networks trained at 75 and 100 epochs.

CONCLUSION

Generative models can improve the performance of low dose CT-based radiomics in different tasks. Hence, denoising using generative models seems to be a necessary pre-processing step for calculating radiomic features from low dose CTs.

Radiomics biopsy signature for predicting survival in patients with spinal bone metastases (SBMs)

•

Prediction of survival is crucial for guiding patient-tailored treatment.

•

Radiomics can be described as the next era of possibilities in precision medicine.

•

Radiomics model had an inferior performance with no added predictive power to the clinical predictive model.

Study design

Retrospective analysis of a registered cohort of patients treated and irradiated for metastases in the spinal column in a single institute.

Objective

This is the first study to develop and internally validate radiomics features for predicting six-month survival probability for patients with spinal bone metastases (SBM).

Background data

Extracted radiomics features from routine clinical CT images can be used to identify textural and intensity-based features unperceivable to human observers and associate them with a patient survival probability or disease progression.

Methods

A study was conducted on 250 patients treated for metastases in the spinal column irradiated for the first time between 2014 and 2016, at the MAASTRO clinic in Maastricht, the Netherlands. The first 150 available patients were used to develop the model and the subsequent 100 patient were considered as a test set for the model. A bootstrap (B = 400) stepwise model selection, which combines both the forward and backward variable elimination procedure, was used to select the most useful predictive features from the training data based on the Akaike information criterion (AIC). The stepwise selection procedure was applied to the 400 bootstrap samples, and the results were plotted as a histogram to visualize how often each variable was selected. Only variables selected more than 90 % of the time over the bootstrap runs were used to build the final model.

A prognostic index (PI) called radiomics score (radscore) and clinical score (clinscore) was calculated for each patient. The prognostic index was not scaled, the original values were used which can be extracted from the model directly or calculated as a linear combination of the variables in the model multiplied by the respective beta value for each patient.

Results

The clinical model had a good discrimination power. The radiomics model, on the other hand, had an inferior performance with no added predictive power to the clinical model. The internal imaging characteristics do not seem to have a value in the prediction of survival. However, the Shape features were excluded from further analyses in our study since all biopsies had a standard shape hence no variability.

Lung cancer diagnosis using deep attention based multiple instance learning and radiomics

Background

Early diagnosis of lung cancer is a key intervention for the treatment of lung cancer in which computer‐aided diagnosis (CAD) can play a crucial role. Most published CAD methods perform lung cancer diagnosis by classifying each lung nodule in isolation. However, this does not reflect clinical practice, where clinicians diagnose a patient based on a set of images of nodules, instead of looking at one nodule at a time. Besides, the low interpretability of the output provided by these methods presents an important barrier for their adoption.

Method

In this article, we treat lung cancer diagnosis as a multiple instance learning (MIL) problem, which better reflects the diagnosis process in the clinical setting and provides higher interpretability of the output. We selected radiomics as the source of input features and deep attention‐based MIL as the classification algorithm. The attention mechanism provides higher interpretability by estimating the importance of each instance in the set for the final diagnosis. To improve the model's performance in a small imbalanced dataset, we propose a new bag simulation method for MIL.

Results and conclusion

The results show that our method can achieve a mean accuracy of 0.807 with a standard error of the mean (SEM) of 0.069, a recall of 0.870 (SEM 0.061), a positive predictive value of 0.928 (SEM 0.078), a negative predictive value of 0.591 (SEM 0.155), and an area under the curve (AUC) of 0.842 (SEM 0.074), outperforming other MIL methods. Additional experiments show that the proposed oversampling strategy significantly improves the model's performance. In addition, experiments show that our method provides a good indication of the importance of each nodule in determining the diagnosis, which combined with the well‐defined radiomic features, to make the results more interpretable and acceptable for doctors and patients.

Artificial Intelligence in Radiation Therapy

Artificial intelligence (AI) has great potential to transform the clinical workflow of radiotherapy. Since the introduction of deep neural networks, many AI-based methods have been proposed to address challenges in different aspects of radiotherapy. Commercial vendors have started to release AI-based tools that can be readily integrated to the established clinical workflow. To show the recent progress in AI-aided radiotherapy, we have reviewed AI-based studies in five major aspects of radiotherapy including image reconstruction, image registration, image segmentation, image synthesis, and automatic treatment planning. In each section, we summarized and categorized the recently published methods, followed by a discussion of the challenges, concerns, and future development. Given the rapid development of AI-aided radiotherapy, the efficiency and effectiveness of radiotherapy in the future could be substantially improved through intelligent automation of various aspects of radiotherapy.

A systematic review and quality of reporting checklist for repeatability and reproducibility of radiomic features

•

Main factors impacting feature stability: Image acquisition, reconstruction, tumor segmentation, and interpolation.

•

Textural features are less robust than morphological or statistical features.

•

A checklist is provided including items that should be reported in a radiomic study.

Purpose

Although quantitative image biomarkers (radiomics) show promising value for cancer diagnosis, prognosis, and treatment assessment, these biomarkers still lack reproducibility. In this systematic review, we aimed to assess the progress in radiomics reproducibility and repeatability in the recent years.

Methods and materials

Four hundred fifty-one abstracts were retrieved according to the original PubMed search pattern with the publication dates ranging from 2017/05/01 to 2020/12/01. Each abstract including the keywords was independently screened by four observers. Forty-two full-text articles were selected for further analysis. Patient population data, radiomic feature classes, feature extraction software, image preprocessing, and reproducibility results were extracted from each article. To support the community with a standardized reporting strategy, we propose a specific reporting checklist to evaluate the feasibility to reproduce each study.

Results

Many studies continue to under-report essential reproducibility information: all but one clinical and all but two phantom studies missed to report at least one important item reporting image acquisition. The studies included in this review indicate that all radiomic features are sensitive to image acquisition, reconstruction, tumor segmentation, and interpolation. However, the amount of sensitivity is feature dependent, for instance, textural features were, in general, less robust than statistical features.

Conclusions

Radiomics repeatability, reproducibility, and reporting quality can substantially be improved regarding feature extraction software and settings, image preprocessing and acquisition, cutoff values for stable feature selection. Our proposed radiomics reporting checklist can serve to simplify and improve the reporting and, eventually, guarantee the possibility to fully replicate and validate radiomic studies.

Practitioners' views on shared decision-making implementation: A qualitative study

INTRODUCTION

Shared decision-making (SDM) refers to the collaboration between patients and their healthcare providers to make clinical decisions based on evidence and patient preferences, often supported by patient decision aids (PDAs). This study explored practitioner experiences of SDM in a context where SDM has been successfully implemented. Specifically, we focused on practitioners' perceptions of SDM as a paradigm, factors influencing implementation success, and outcomes.

METHODS

We used a qualitative approach to examine the experiences and perceptions of 10 Danish practitioners at a cancer hospital experienced in SDM implementation. A semi-structured interview format was used and interviews were audio-recorded and transcribed. Data was analyzed through thematic analysis.

RESULTS

Prior to SDM implementation, participants had a range of attitudes from skeptical to receptive. Those with more direct long-term contact with patients (such as nurses) were more positive about the need for SDM. We identified four main factors that influenced SDM implementation success: raising awareness of SDM behaviors among clinicians through concrete measurements, supporting the formation of new habits through reinforcement mechanisms, increasing the flexibility of PDA delivery, and strong leadership. According to our participants, these factors were instrumental in overcoming initial skepticism and solidifying new SDM behaviors. Improvements to the clinical process were reported. Sustaining and transferring the knowledge gained to other contexts will require adapting measurement tools.

CONCLUSIONS

Applying SDM in clinical practice represents a major shift in mindset for clinicians. Designing SDM initiatives with an understanding of the underlying behavioral mechanisms may increase the probability of successful and sustained implementation.

Quantification of the spatial distribution of primary tumors in the lung to develop new prognostic biomarkers for locally advanced NSCLC

The anatomical location and extent of primary lung tumors have shown prognostic value for overall survival (OS). However, its manual assessment is prone to interobserver variability. This study aims to use data driven identification of image characteristics for OS in locally advanced non-small cell lung cancer (NSCLC) patients. Five stage IIIA/IIIB NSCLC patient cohorts were retrospectively collected. Patients were treated either with radiochemotherapy (RCT): RCT1* (n = 107), RCT2 (n = 95), RCT3 (n = 37) or with surgery combined with radiotherapy or chemotherapy: S1* (n = 135), S2 (n = 55). Based on a deformable image registration (MIM Vista, 6.9.2.), an in-house developed software transferred each primary tumor to the CT scan of a reference patient while maintaining the original tumor shape. A frequency-weighted cumulative status map was created for both exploratory cohorts (indicated with an asterisk), where the spatial extent of the tumor was uni-labeled with 2 years OS. For the exploratory cohorts, a permutation test with random assignment of patient status was performed to identify regions with statistically significant worse OS, referred to as decreased survival areas (DSA). The minimal Euclidean distance between primary tumor to DSA was extracted from the independent cohorts (negative distance in case of overlap). To account for the tumor volume, the distance was scaled with the radius of the volume-equivalent sphere. For the S1 cohort, DSA were located at the right main bronchus whereas for the RCT1 cohort they further extended in cranio-caudal direction. In the independent cohorts, the model based on distance to DSA achieved performance: AUC_RCT2 [95% CI] = 0.67 [0.55–0.78] and AUC_RCT3 = 0.59 [0.39–0.79] for RCT patients, but showed bad performance for surgery cohort (AUC_S2 = 0.52 [0.30–0.74]). Shorter distance to DSA was associated with worse outcome (p = 0.0074). In conclusion, this explanatory analysis quantifies the value of primary tumor location for OS prediction based on cumulative status maps. Shorter distance of primary tumor to a high-risk region was associated with worse prognosis in the RCT cohort.

Generative models improve radiomics reproducibility in low dose CTs: a simulation study

Radiomics is an active area of research in medical image analysis, however poor reproducibility of radiomics has hampered its application in clinical practice. This issue is especially prominent when radiomic features are calculated from noisy images, such as low dose computed tomography (CT) scans. In this article, we investigate the possibility of improving the reproducibility of radiomic features calculated on noisy CTs by using generative models for denoising. Our work concerns two types of generative models-encoder-decoder network (EDN) and conditional generative adversarial network (CGAN). We then compared their performance against a more traditional 'non-local means' denoising algorithm. We added noise to sinograms of full dose CTs to mimic low dose CTs with two levels of noise: low-noise CT and high-noise CT. Models were trained on high-noise CTs and used to denoise low-noise CTs without re-training. We tested the performance of our model in real data, using a dataset of same-day repeated low dose CTs in order to assess the reproducibility of radiomic features in denoised images. EDN and the CGAN achieved similar improvements on the concordance correlation coefficients (CCC) of radiomic features for low-noise images from 0.87 [95%CI, (0.833, 0.901)] to 0.92 [95%CI, (0.909, 0.935)] and for high-noise images from 0.68 [95%CI, (0.617, 0.745)] to 0.92 [95%CI, (0.909, 0.936)], respectively. The EDN and the CGAN improved the test-retest reliability of radiomic features (mean CCC increased from 0.89 [95%CI, (0.881, 0.914)] to 0.94 [95%CI, (0.927, 0.951)]) based on real low dose CTs. These results show that denoising using EDN and CGANs could be used to improve the reproducibility of radiomic features calculated from noisy CTs. Moreover, images at different noise levels can be denoised to improve the reproducibility using the above models without need for re-training, provided the noise intensity is not excessively greater that of the high-noise CTs. To the authors' knowledge, this is the first effort to improve the reproducibility of radiomic features calculated on low dose CT scans by applying generative models.

Predicting outcomes in anal cancer patients using multi-centre data and distributed learning - A proof-of-concept study

BACKGROUND AND PURPOSE

Predicting outcomes is challenging in rare cancers. Single-institutional datasets are often small and multi-institutional data sharing is complex. Distributed learning allows machine learning models to use data from multiple institutions without exchanging individual patient-level data. We demonstrate this technique in a proof-of-concept study of anal cancer patients treated with chemoradiotherapy across multiple European countries.

MATERIALS AND METHODS

atomCAT is a three-centre collaboration between Leeds Cancer Centre (UK), MAASTRO Clinic (The Netherlands) and Oslo University Hospital (Norway). We trained and validated a Cox proportional hazards regression model in a distributed fashion using data from 281 patients treated with radical, conformal chemoradiotherapy for anal cancer in three institutions. Our primary endpoint was overall survival. We selected disease stage, sex, age, primary tumour size, and planned radiotherapy dose (in EQD2) a priori as predictor variables.

RESULTS

The Cox regression model trained across all three centres found worse overall survival for high risk disease stage (HR = 2.02), male sex (HR = 3.06), older age (HR = 1.33 per 10 years), larger primary tumour volume (HR = 1.05 per 10 cm3) and lower radiotherapy dose (HR = 1.20 per 5 Gy). A mean concordance index of 0.72 was achieved during validation, with limited variation between centres (Leeds = 0.72, MAASTRO = 0.74, Oslo = 0.70). The global model performed well for risk stratification for two out of three centres.

CONCLUSIONS

Using distributed learning, we accessed and analysed one of the largest available multi-institutional cohorts of anal cancer patients treated with modern radiotherapy techniques. This demonstrates the value of distributed learning in outcome modelling for rare cancers.

Emerging role of artificial intelligence in nuclear medicine

The role of artificial intelligence is increasing in all branches of medicine. The emerging role of artificial intelligence applications in nuclear medicine is going to improve the nuclear medicine clinical workflow in the coming years. Initial research outcomes are suggestive of increasing role of artificial intelligence in nuclear medicine workflow, particularly where selective automation tasks are of concern. Artificial intelligence-assisted planning, dosimetry and procedure execution appear to be areas for rapid and significant development. The role of artificial intelligence in more directly imaging-related tasks, such as dose optimization, image corrections and image reconstruction, have been particularly strong points of artificial intelligence research in nuclear medicine. Natural Language Processing (NLP)-based text processing task is another area of interest of artificial intelligence implementation in nuclear medicine.

Artificial Intelligence Applications to Improve the Treatment of Locally Advanced Non-Small Cell Lung Cancers

Locally advanced non-small cell lung cancer patients represent around one third of newly diagnosed lung cancer patients. There remains a large unmet need to find treatment strategies that can improve the survival of these patients while minimizing therapeutical side effects. Increasing the availability of patients' data (imaging, electronic health records, patients' reported outcomes, and genomics) will enable the application of AI algorithms to improve therapy selections. In this review, we discuss how artificial intelligence (AI) can be integral to improving clinical decision support systems. To realize this, a roadmap for AI must be defined. We define six milestones involving a broad spectrum of stakeholders, from physicians to patients, that we feel are necessary for an optimal transition of AI into the clinic.