Machine learning helps identifying volume-confounding effects in radiomics

Measured intrapatient radiomic variability as a predictor of treatment response in multi-metastatic soft tissue sarcoma patients

Radiomics offers a non-invasive approach to tumor characterization, yet its application in metastatic cancers is limited by intertumor heterogeneity-variability in radiomic phenotypes across lesions within the same patient. We introduce Measured Intrapatient Radiomic Variability (MIRV), a novel metric quantifying heterogeneity using standard-of-care imaging. Applied to 397 metastatic soft-tissue sarcoma (STS) patients from the SARC021 trial, MIRV was calculated from pretreatment CT scans using pairwise Euclidean distance and cosine dissimilarity between lesions. Euclidean distance captures absolute differences in radiomic features, while cosine dissimilarity assesses variation in feature patterns independent of magnitude. Higher MIRV correlated with greater variability in tumor-specific response classification (TSRC) and volumetric response, independent of baseline tumor volume. In a subset with liquid biopsy data, MIRV showed a moderate association with ctDNA positivity, suggesting links to molecular heterogeneity. While MIRV was not prognostic for overall survival (OS) in the full cohort, higher MIRV was significantly associated with worse survival in leiomyosarcoma patients (n=165, p=0.007). These findings establish MIRV as a biomarker for intertumor heterogeneity, with potential to predict mixed treatment responses and guide personalized therapy in metastatic STS. Future studies should assess its relevance across other tumor types and therapeutic settings.

Radiomic-based approaches in the multi-metastatic setting: a quantitative review

BACKGROUND

Radiomics traditionally focuses on analyzing a single lesion within a patient to extract tumor characteristics, yet this process may overlook inter-lesion heterogeneity, particularly in the multi-metastatic setting. There is currently no established method for combining radiomic features in such settings, leading to diverse approaches with varying strengths and limitations. Our quantitative review aims to illuminate these methodologies, assess their replicability, and guide future research toward establishing best practices, offering insights into the challenges of multi-lesion radiomic analysis across diverse datasets.

METHODS

We conducted a comprehensive literature search to identify methods for integrating data from multiple lesions in radiomic analyses. We replicated these methods using either the author's code or by reconstructing them based on the information provided in the papers. Subsequently, we applied these identified methods to three distinct datasets, each depicting a different metastatic scenario.

RESULTS

We compared ten mathematical methods for combining radiomic features across three distinct datasets, encompassing 16,894 lesions in 3,930 patients. Performance was evaluated using the Cox proportional hazards model and benchmarked against univariable analysis of total tumor volume. Results varied by dataset and lesion burden, with no single method consistently outperforming others. In colorectal liver metastases (TCIA-CRLM, 494 lesions in 197 patients), averaging methods showed the highest median performance. In soft tissue sarcoma (TH CR-406/SARC021, 1255 lesions in 545 patients), concatenating radiomic features from multiple lesions exhibited the best performance. In head and neck cancers (TCIA-RADCURE, 15,145 lesions in 3188 patients), total tumor volume remained a strong predictor. These findings highlight dataset-specific influences, including tumor type and lesion burden, on the effectiveness of radiomic feature aggregation methods.

CONCLUSIONS

Radiomic features can be effectively selected or combined to estimate patient-level outcomes in multi-metastatic patients, though the approach varies by metastatic setting. Our study fills a critical gap in radiomics research by examining the challenges of radiomic-based analysis in this setting. Through a comprehensive review and rigorous testing of different methods across diverse datasets representing unique metastatic scenarios, we provide valuable insights into effective radiomic analysis strategies.

Preoperative assessment of tumor consistency and gross total resection in pituitary adenoma: Radiomic analysis of T2-weighted MRI and interpretation of contributing radiomic features

Background

Preoperative knowledge of tumor consistency and the likelihood of gross total resection (GTR) would greatly benefit planning of pituitary adenoma surgery, however, no reliable methods currently exist.

Objectives

To evaluate the utility of radiomic analysis of MRI for predicting tumor consistency and GTR. To explore the interpretability of contributing radiomic features.

Methods

Patients undergoing first endoscopic surgery for pituitary macroadenomas were included. Tumor consistency was assessed intraoperatively, GTR was assessed based on postoperative MRI. Radiomic features were extracted from axial T2-weighted MRI. Low-variability and highly intercorrelated features were removed. Random Forest Classifiers were optimized using 70 % of patient data and evaluated on the remaining 30 %. Relative feature importance was assessed using the Gini-Simpson index.

Results

542 patients were included. GTR was achieved in 325 (60.0 %) cases, firm tumors were encountered in 122 (22.5 %) cases. There was a significant correlation between GTR and tumor consistency (67.1 % vs. 35.2 %, p < 0.001). 1688 radiomic variables were extracted, 442 were removed due to low variance and 699 due to high intercorrelation. The consistency prediction model achieved an accuracy of 81.6 % and utilized 32 features, GTR prediction model achieved 79.1 % accuracy using 73 features.

Conclusions

Radiomic analysis demonstrated significant potential for preoperative evaluation of pituitary adenomas. Texture and intensity-based features were the primary contributors to consistency prediction. However, the explanation of these features was insufficient. GTR prediction was predominantly driven by shape-related features. Our findings highlight the challenges of linking radiomic features to underlying tissue properties and emphasize the need for cautious interpretation.

Correlation of T2-Weighted Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Cone Beam Computed Tomography (CBCT) Radiomic Features for Prostate Cancer

Radiomics extracted from cone beam computed tomography (CBCT) can be assessed at time points during treatment and may provide an advantage over assessments in a pre-treatment setting using diagnostic images, like magnetic resonance imaging (MRI) or computed tomography (CT), for prostate cancer (pCa) patients receiving radiotherapy (RT). The purpose of this study was to analyze correlations between prostate radiomic features (RFs) derived from T2-weighted (T2w) MRI, CT, and first fraction CBCT for patients receiving RT for pCa. Forty-seven patients were analyzed. The prostate volumes were manually segmented, and 42 radiomic features were extracted, of which seven volume-normalized RFs were considered. The absolute Spearman correlation was calculated among the RFs of the aforementioned imaging modalities (RM) and prostate volume (RV) since the motivation of this paper was to analyze the strength of the correlation. The Benjamini-Hochberg adjustment was applied to p-values to account for multiple comparisons. No high correlations were found between CT/CBCT vs. T2w. The intramodality RM demonstrated that CT RFs were much higher than the other modalities. For example, intramodality RM≥0.95 the percentage of RFs was 17% for CT, 9% for CBCT, and 4.5% for T2w. The differences in RFs across different modalities can be viewed positively: the lack of correlation between RFs across T2w and CT/CBCT could indicate a fundamental difference in the extractable image information. It could also indicate that some RFs did not have any extractable information. A future study will include evaluating the predictive performance of patient outcomes using radiomic features from CT, CBCT, and T2w, which could help in answering such questions.

Repeatability, reproducibility, and the effects of radiotherapy on radiomic features of lowfield MR-LINAC images of the prostate

Definitive radiotherapy (RT) has been shown to be a successful method of treating prostate cancer (PCa) patients. Through radiomics, a quantitative analysis of medical images, it is possible to adapt treatment early on, which may prevent or mitigate future adverse events. During RT of PCa, low-field magnetic resonance (MR) images, taken with a LINAC onboard imaging system in a process known as magnetic resonance-guided radiotherapy (MRgRT), are used to improve treatment accuracy via superior setup compared to x-ray methods. This work investigated baseline repeatability of radiomic features (RFs) by comparing planning MR images (pMR) with first-fraction setup images (FX1) taken with onboard MRI. The changes in RFs following RT were also looked at with the use of last-fraction setup images (FX5). Earlier research has investigated the use of planning images from cone beam CT (CBCT), but to our knowledge no research has previously shown the relationship with onboard MRI. The correlation between FX1 images and 3T diagnostic MR (dT2) images was also studied. Forty-three first and second order radiomic features extracted from these images were compared by calculating Lin's concordance correlation coefficient (with Benjamini-Hochberg correction for multiple comparisons) between the modalities. FX1 and pMR images were correlated (p<0.05) for all but one RF. 12 RFs correlated between pMR and dT2 images. There was a noticeable change in correlation values for RFs when looking at FX1 and FX5 images, with only 15 correlating significantly. The change in correlation values between pMR and FX5 images was comparable to that between FX1 and FX5 images, with 33 features having a CCC value deviation of less than 0.1. These results demonstrate that RF features are repeatable across different images of the same modality without treatment intervention. This study has also shown a noticeable, reproducible change in RFs as RT goes on. Reproducibility of RFs between different modalities was not strong. This study demonstrated that we can reliably use onboard MRI to observe day-to-day feature changes as a result of RT.

Diagnostic performance of radiomics in prediction of Ki-67 index status in non-small cell lung cancer: A systematic review and meta-analysis

BACKGROUND

Lung cancer's high prevalence and invasiveness make it a major global health concern. The Ki-67 index, which indicates cellular proliferation, is crucial for assessing lung cancer aggressiveness. Radiomics, which extracts quantifiable features from medical images using algorithms, may provide insights into tumor behavior. This systematic review and meta-analysis evaluate the effectiveness of radiomics in predicting Ki-67 status in Non-Small Cell Lung Cancer (NSCLC) using CT scans.

METHODS AND MATERIALS

A comprehensive search was conducted in PubMed/MEDLINE, Embase, Scopus, and Web of Science databases from inception until April 19, 2024. Original studies discussing the performance of CT-based radiomics for predicting Ki-67 status in NSCLC cohorts were included. The quality assessment involved quality assessment of diagnostic accuracy studies (QUADAS-2), radiomics quality score (RQS) and METhodological RadiomICs Score (METRICS). Quantitative meta-analysis, using R, assessed pooled diagnostic odds ratio, sensitivity, and specificity in NSCLC cohorts.

RESULTS

We identified 10 studies that met the inclusion criteria, involving 2279 participants, with 9 of these studies included in quantitative meta-analysis. The pooled sensitivity and specificity of radiomics-based models for predicting Ki-67 status in NSCLC were 0.783 (95 % CI: 0.732 - 0.827) and 0.796 (95 % CI: 0.707 - 0.864) in training cohorts, and 0.803 (95 % CI: 0.744 - 0.851) and 0.696 (95 % CI: 0.613 - 0.768) in validation cohorts. It was identified in subgroup analysis that utilizing ITK-SNAP as a segmentation software contributed to a significantly higher pooled sensitivity.

CONCLUSION

This meta-analysis indicates promising diagnostic accuracy of radiomics in predicting Ki-67 in NSCLC.

Robust machine learning challenge: An AIFM multicentric competition to spread knowledge, identify common pitfalls and recommend best practice

PURPOSE

A novel and unconventional approach to a machine learning challenge was designed to spread knowledge, identify robust methods and highlight potential pitfalls about machine learning within the Medical Physics community.

METHODS

A public dataset comprising 41 radiomic features and 535 patients was employed to assess the potential of radiomics in distinguishing between primary lung tumors and metastases. Each participant developed two classification models using: (i) all features (base model); (ii) only robust features (robust model). Both models were validated with cross-validation and on unseen data. The population stability index (PSI) was used as diagnostic metric for implementation issues. Performance was compared to reference. Base and robust models were compared in terms of performance and stability (coefficient of variation (CoV) of prediction probabilities).

RESULTS

PSI detected potential implementation errors in 70 % of models. The dataset exhibited strong imbalance. The average Gmean (i.e. an appropriate metric for imbalance) among all participants was 0.67 ± 0.01, significantly higher than reference Gmean of 0.50 ± 0.04. Robust models performances were slightly worse than base models (p < 0.05). Regarding stability, robust models exhibited lower median CoV on training set only.

CONCLUSION

AI4MP-Challenge models overperformed the reference, significantly improving the Gmean. Exclusion of less-robust features did not improve model robustness and it should be avoided when confounding effects are absent. Other methods, like harmonization or data augmentation, should be evaluated. This study demonstrated how the collaborative effort to foster knowledge on machine learning among medical physicists, through interactive sessions and exchange of information among participants, can result in improved models.

radMLBench: A dataset collection for benchmarking in radiomics

BACKGROUND

New machine learning methods and techniques are frequently introduced in radiomics, but they are often tested on a single dataset, which makes it challenging to assess their true benefit. Currently, there is a lack of a larger, publicly accessible dataset collection on which such assessments could be performed. In this study, a collection of radiomics datasets with binary outcomes in tabular form was curated to allow benchmarking of machine learning methods and techniques.

METHODS

A variety of journals and online sources were searched to identify tabular radiomics data with binary outcomes, which were then compiled into a homogeneous data collection that is easily accessible via Python. To illustrate the utility of the dataset collection, it was applied to investigate whether feature decorrelation prior to feature selection could improve predictive performance in a radiomics pipeline.

RESULTS

A total of 50 radiomic datasets were collected, with sample sizes ranging from 51 to 969 and 101 to 11165 features. Using this data, it was observed that decorrelating features did not yield any significant improvement on average.

CONCLUSIONS

A large collection of datasets, easily accessible via Python, suitable for benchmarking and evaluating new machine learning techniques and methods was curated. Its utility was exemplified by demonstrating that feature decorrelation prior to feature selection does not, on average, lead to significant performance gains and could be omitted, thereby increasing the robustness and reliability of the radiomics pipeline.

Radiomics reproducibility in computed tomography through changes of ROI size, resolution, and hounsfield unit: A phantom study

INTRODUCTION

Although radiomics has revealed an intriguing perspective for quantitative radiology, the impact of scanning parameters on its outcomes must be considered. In this study, the effects of changes in the region of interest (ROI) sizes, Hounsfield Unit (HU), and resolution of computed tomography (CT) on feature reproducibility have been investigated.

METHODS

The GAMMEX 464 phantom was used to evaluate the reproducibility of radiomics features across different ROI sizes, HU, and resolution. Data were acquired using a consistent system setup, with the phantom repositioned for each scan. The first acquisition series examined the effects of different ROI sizes and resolutions (1, 3, and 5 mm) on feature reproducibility. The second series assessed the impact of different HU and resolution. Segmentation and feature extraction were performed using LIFEx 7.1.0 software, focusing on textural radiomics features. Statistical analysis involved calculating the coefficient of variation (COV) to categorize feature variability. COV <5 % was considered highly stable.

RESULTS

Out of the 32 textural features studied, the analysis of changes in ROI size with a resolution of 1 mm, 3 mm, and 5 mm revealed that 16, 17, and 18 features had high reproducibility, with a COV<5 %. Polyethylene, acrylic, and water also demonstrated stable textural features across changes in scan parameters and image resolutions, with 4 reproducible features in all resolutions. The grey-level run length matrix (GLRLM) and grey-level zone length matrix (GLZLM) radiomics groups were highly stable in the context of variations in scan parameters and different materials.

CONCLUSION

The results of this study highlight the importance of standardizing radiomics studies to reduce the influence of pre-analysis steps on feature values. This standardization is crucial for guaranteeing the consistency of radiomics features under various imaging conditions. Additional research is required to enhance these results.

IMPLICATIONS FOR PRACTICE

To ensure the reproducibility and reliability of radiomics features, it is imperative to standardize scanning parameters and pre-analysis protocols. This standardization will enhance the consistency of radiomics applications in both clinical and research environments.

Combined CT-Based Radiomics and Clinic-Radiological Characteristics for Preoperative Differentiation of Solitary-Type Invasive Mucinous and Non-Mucinous Lung Adenocarcinoma

Purpose

The clinical, pathological, gene expression, and prognosis of invasive mucinous adenocarcinoma (IMA) differ from those of invasive non-mucinous adenocarcinoma (INMA), but it is not easy to distinguish these two. This study aims to explore the value of combining CT-based radiomics features with clinic-radiological characteristics for preoperative diagnosis of solitary-type IMA and to establish an optimal diagnostic model.

Methods

In this retrospective study, a total of 220 patients were enrolled and randomly assigned to a training cohort (n = 154; 73 IMA and 81 INMA) and a testing cohort (n = 66; 31 IMA and 35 INMA). Radiomics features and clinic-radiological characteristics were extracted from plain CT images. The radiomics models for predicting solitary-type IMA were developed by three classifiers: linear discriminant analysis (LDA), logistic regression-least absolute shrinkage and selection operator (LR-LASSO), and support vector machine (SVM). The combined model was constructed by integrating radiomics and clinic-radiological features with the best performing classifier. Receiver operating characteristic (ROC) curves were used to evaluate models' performance, and the area under the curve (AUC) were compared by the DeLong test. Decision curve analysis (DCA) was conducted to assess the clinical utility.

Results

Regarding CT characteristics, tumor lung interface, and pleural retraction were the independent risk factors of solitary-type IMA. The radiomics model using the SVM classifier outperformed the other two classifiers in the testing cohort, with an AUC of 0.776 (95% CI: 0.664-0.888). The combined model incorporating radiomics features and clinic-radiological factors was the optimal model, with AUCs of 0.843 (95% CI: 0.781-0.906) and 0.836 (95% CI: 0.732-0.940) in the training and testing cohorts, respectively.

Conclusion

The combined model showed good ability in predicting solitary-type IMA and can provide a non-invasive and efficient approach to clinical decision-making.

Radiomic-based prediction of lesion-specific systemic treatment response in metastatic disease

Despite sharing the same histologic classification, individual tumors in multi metastatic patients may present with different characteristics and varying sensitivities to anticancer therapies. In this study, we investigate the utility of radiomic biomarkers for prediction of lesion-specific treatment resistance in multi metastatic leiomyosarcoma patients. Using a dataset of n=202 lung metastases (LM) from n=80 patients with 1648 pre-treatment computed tomography (CT) radiomics features and LM progression determined from follow-up CT, we developed a radiomic model to predict the progression of each lesion. Repeat experiments assessed the relative predictive performance across LM volume groups. Lesion-specific radiomic models indicate up to a 4.5-fold increase in predictive capacity compared with a no-skill classifier, with an area under the precision-recall curve of 0.70 for the most precise model (FDR = 0.05). Precision varied by administered drug and LM volume. The effect of LM volume was controlled by removing radiomic features at a volume-correlation coefficient threshold of 0.20. Predicting lesion-specific responses using radiomic features represents a novel strategy by which to assess treatment response that acknowledges biological diversity within metastatic subclones, which could facilitate management strategies involving selective ablation of resistant clones in the setting of systemic therapy.

Radiomic-Based Prediction of Lesion-Specific Systemic Treatment Response in Metastatic Disease

Despite sharing the same histologic classification, individual tumors in multi metastatic patients may present with different characteristics and varying sensitivities to anticancer therapies. In this study, we investigate the utility of radiomic biomarkers for prediction of lesion-specific treatment resistance in multi metastatic leiomyosarcoma patients. Using a dataset of n=202 lung metastases (LM) from n=80 patients with 1648 pre-treatment computed tomography (CT) radiomics features and LM progression determined from follow-up CT, we developed a radiomic model to predict the progression of each lesion. Repeat experiments assessed the relative predictive performance across LM volume groups. Lesion-specific radiomic models indicate up to a 4.5-fold increase in predictive capacity compared with a no-skill classifier, with an area under the precision-recall curve of 0.70 for the most precise model (FDR = 0.05). Precision varied by administered drug and LM volume. The effect of LM volume was controlled by removing radiomic features at a volume-correlation coefficient threshold of 0.20. Predicting lesion-specific responses using radiomic features represents a novel strategy by which to assess treatment response that acknowledges biological diversity within metastatic subclones, which could facilitate management strategies involving selective ablation of resistant clones in the setting of systemic therapy.

Mitigating the impact of image processing variations on tumour [18F]-FDG-PET radiomic feature robustness

Radiomics analysis of [18F]-fluorodeoxyglucose ([18F]-FDG) PET images could be leveraged for personalised cancer medicine. However, the inherent sensitivity of radiomic features to intensity discretisation and voxel interpolation complicates its clinical translation. In this work, we evaluated the robustness of tumour [18F]-FDG-PET radiomic features to 174 different variations in intensity resolution or voxel size, and determined whether implementing parameter range conditions or dependency corrections could improve their robustness. Using 485 patient images spanning three cancer types: non-small cell lung cancer (NSCLC), melanoma, and lymphoma, we observed features were more sensitive to intensity discretisation than voxel interpolation, especially texture features. In most of our investigations, the majority of non-robust features could be made robust by applying parameter range conditions. Correctable features, which were generally fewer than conditionally robust, showed systematic dependence on bin configuration or voxel size that could be minimised by applying corrections based on simple mathematical equations. Melanoma images exhibited limited robustness and correctability relative to NSCLC and lymphoma. Our study provides an in-depth characterisation of the sensitivity of [18F]-FDG-PET features to image processing variations and reinforces the need for careful selection of imaging biomarkers prior to any clinical application.

Radiomic-Based Approaches in the Multi-metastatic Setting: A Quantitative Review

Background

Radiomics traditionally focuses on analyzing a single lesion within a patient to extract tumor characteristics, yet this process may overlook inter-lesion heterogeneity, particularly in the multi-metastatic setting. There is currently no established method for combining radiomic features in such settings, leading to diverse approaches with varying strengths and limitations. Our quantitative review aims to illuminate these methodologies, assess their replicability, and guide future research toward establishing best practices, offering insights into the challenges of multi-lesion radiomic analysis across diverse datasets.

Methods

We conducted a comprehensive literature search to identify methods for integrating data from multiple lesions in radiomic analyses. We replicated these methods using either the author's code or by reconstructing them based on the information provided in the papers. Subsequently, we applied these identified methods to three distinct datasets, each depicting a different metastatic scenario.

Results

We compared ten mathematical methods for combining radiomic features across three distinct datasets, encompassing a total of 16,850 lesions in 3,930 patients. Performance of these methods was evaluated using the Cox proportional hazards model and benchmarked against univariable analysis of total tumor volume. We observed variable performance in methods across datasets. However, no single method consistently outperformed others across all datasets. Notably, while some methods surpassed total tumor volume analysis in certain datasets, others did not. Averaging methods showed higher median performance in patients with colorectal liver metastases, and in soft tissue sarcoma, concatenation of radiomic features from different lesions exhibited the highest median performance among tested methods.

Conclusions

Radiomic features can be effectively selected or combined to estimate patient-level outcomes in multi-metastatic patients, though the approach varies by metastatic setting. Our study fills a critical gap in radiomics research by examining the challenges of radiomic-based analysis in this setting. Through a comprehensive review and rigorous testing of different methods across diverse datasets representing unique metastatic scenarios, we provide valuable insights into effective radiomic analysis strategies.

Increasing the efficiency of cone-beam CT based delta-radiomics using automated contours to predict radiotherapy-related toxicities in prostate cancer

Extracting longitudinal image quantitative data, known as delta-radiomics, has the potential to capture changes in a patient's anatomy throughout the course of radiation treatment for prostate cancer. Some of the major challenges of delta-radiomics studies are contouring the structures for individual fractions and accruing patients' data in an efficient manner. The manual contouring process is often time consuming and would limit the efficiency of accruing larger sample sizes for future studies. The problem is amplified because the contours are often made by highly trained radiation oncologists with limited time to dedicate to research studies of this nature. This work compares the use of automated prostate contours generated using a deformable image-based algorithm to make predictive models of genitourinary and changes in total international prostate symptom score in comparison to manually contours for a cohort of fifty patients. Area under the curve of manual and automated models were compared using the Delong test. This study demonstrated that the delta-radiomics models were similar for both automated and manual delta-radiomics models.

Distinguishing recurrence from radiation-induced lung injury at the time of RECIST progressive disease on post-SABR CT scans using radiomics

Stereotactic ablative radiotherapy (SABR) is a highly effective treatment for patients with early-stage lung cancer who are inoperable. However, SABR causes benign radiation-induced lung injury (RILI) which appears as lesion growth on follow-up CT scans. This triggers the standard definition of progressive disease, yet cancer recurrence is not usually present, and distinguishing RILI from recurrence when a lesion appears to grow in size is critical but challenging. In this study, we developed a tool to do this using scans with apparent lesion growth after SABR from 68 patients. We performed bootstrapped experiments using radiomics and explored the use of multiple regions of interest (ROIs). The best model had an area under the receiver operating characteristic curve of 0.66 and used a sphere with a diameter equal to the lesion's longest axial measurement as the ROI. We also investigated the effect of using inter-feature and volume correlation filters and found that the former was detrimental to performance and that the latter had no effect. We also found that the radiomics features ranked as highly important by the model were significantly correlated with outcomes. These findings represent a key step in developing a tool that can help determine who would benefit from follow-up invasive interventions when a SABR-treated lesion increases in size, which could help provide better treatment for patients.

Assessing robustness of quantitative susceptibility-based MRI radiomic features in patients with multiple sclerosis

Multiple Sclerosis (MS) is an autoimmune demyelinating disease characterised by changes in iron and myelin content. These biomarkers are detectable by Quantitative Susceptibility Mapping (QSM), an advanced Magnetic Resonance Imaging technique detecting magnetic properties. When analysed with radiomic techniques that exploit its intrinsic quantitative nature, QSM may furnish biomarkers to facilitate early diagnosis of MS and timely assessment of progression. In this work, we explore the robustness of QSM radiomic features by varying the number of grey levels (GLs) and echo times (TEs), in a sample of healthy controls and patients with MS. We analysed the white matter in total and within six clinically relevant tracts, including the cortico-spinal tract and the optic radiation. After optimising the number of GLs (n = 64), at least 65% of features were robust for each Volume of Interest (VOI), with no difference (p > .05) between left and right hemispheres. Different outcomes in feature robustness among the VOIs depend on their characteristics, such as volume and variance of susceptibility values. This study validated the processing pipeline for robustness analysis and established the reliability of QSM-based radiomics features against GLs and TEs. Our results provide important insights for future radiomics studies using QSM in clinical applications.

The role of parametric feature maps to correct different volume of interest sizes: an in vivo liver MRI study

BACKGROUND

Different volume of interest (VOI) sizes influence radiomic features. This study examined if translating images into feature maps before feature sampling could compensate for these effects in liver magnetic resonance imaging (MRI).

METHODS

T1- and T2-weighted sequences from three different scanners (two 3-T scanners, one 1.5-T scanner) of 66 patients with normal abdominal MRI were included retrospectively. Three differently sized VOIs (10, 20, and 30 mm in diameter) were drawn in the liver parenchyma (right lobe), excluding adjacent structures. Ninety-three features were extracted conventionally using PyRadiomics. All images were also converted to 93 parametric feature maps using a pretested software. Agreement between the three VOI sizes was assessed with overall concordance correlation coefficients (OCCCs), while OCCCs > 0.85 were rated reproducible. OCCCs were calculated twice: for the VOI sizes of 10, 20, and 30 mm and for those of 20 and 30 mm.

RESULTS

When extracted from original images, only 4 out of the 93 features were reproducible across all VOI sizes in T1- and T2-weighted images. When the smallest VOI was excluded, 5 features (T1-weighted) and 7 features (T2-weighted) were reproducible. Extraction from parametric maps increased the number of reproducible features to 9 (T1- and T2-weighted) across all VOIs. Excluding the 10-mm VOI, reproducibility improved to 16 (T1-weighted) and 55 features (T2-weighted). The stability of all other features also increased in feature maps.

CONCLUSIONS

Translating images into parametric maps before feature extraction improves reproducibility across different VOI sizes in normal liver MRI.

RELEVANCE STATEMENT

The size of the segmented VOI influences the feature quantity of radiomics, while software-based conversion of images into parametric feature maps before feature sampling improves reproducibility across different VOI sizes in MRI of normal liver tissue.

KEY POINTS

• Parametric feature maps can compensate for different VOI sizes. • The effect seems dependent on the VOI sizes and the MRI sequence. • Feature maps can visualize features throughout the entire image stack.

Multi-institutional Prognostic Modeling in Head and Neck Cancer: Evaluating Impact and Generalizability of Deep Learning and Radiomics

Artificial intelligence (AI) and machine learning (ML) are becoming critical in developing and deploying personalized medicine and targeted clinical trials. Recent advances in ML have enabled the integration of wider ranges of data including both medical records and imaging (radiomics). However, the development of prognostic models is complex as no modeling strategy is universally superior to others and validation of developed models requires large and diverse datasets to demonstrate that prognostic models developed (regardless of method) from one dataset are applicable to other datasets both internally and externally. Using a retrospective dataset of 2,552 patients from a single institution and a strict evaluation framework that included external validation on three external patient cohorts (873 patients), we crowdsourced the development of ML models to predict overall survival in head and neck cancer (HNC) using electronic medical records (EMR) and pretreatment radiological images. To assess the relative contributions of radiomics in predicting HNC prognosis, we compared 12 different models using imaging and/or EMR data. The model with the highest accuracy used multitask learning on clinical data and tumor volume, achieving high prognostic accuracy for 2-year and lifetime survival prediction, outperforming models relying on clinical data only, engineered radiomics, or complex deep neural network architecture. However, when we attempted to extend the best performing models from this large training dataset to other institutions, we observed significant reductions in the performance of the model in those datasets, highlighting the importance of detailed population-based reporting for AI/ML model utility and stronger validation frameworks. We have developed highly prognostic models for overall survival in HNC using EMRs and pretreatment radiological images based on a large, retrospective dataset of 2,552 patients from our institution.Diverse ML approaches were used by independent investigators. The model with the highest accuracy used multitask learning on clinical data and tumor volume.External validation of the top three performing models on three datasets (873 patients) with significant differences in the distributions of clinical and demographic variables demonstrated significant decreases in model performance.

Significance

ML combined with simple prognostic factors outperformed multiple advanced CT radiomics and deep learning methods. ML models provided diverse solutions for prognosis of patients with HNC but their prognostic value is affected by differences in patient populations and require extensive validation.

Recognition of liver tumors by predicted hyperspectral features based on patient's Computed Tomography radiomics features

BACKGROUND

Primary liver tumors have posed a serious threat to human life and health, and their early diagnosis is urgent. Therefore, enhancing the accuracy of non-invasive early detection of liver tumors is imperative.

METHODS

Firstly, image enhancement was applied to augment the dataset, resulting in a total of 464 samples after employing seven data augmentation methods. Subsequently, the XGBoost model was utilized to construct and learn the mapping relationship between Computed Tomography (CT) and corresponding hyperspectral imaging (HSI) data. This model enables the prediction of HSI features corresponding to CT features, thereby enriching CT with more comprehensive hyperspectral information.

RESULTS

Four classifiers were employed to discern the presence of tumors in patients. The results demonstrated exceptional performance, with a classification accuracy exceeding 90%.

CONCLUSIONS

This study proposes an artificial intelligence-based methodology that utilizes early CT radiomics features to predict HSI features. Subsequently, the results are utilized for non-invasive tumor prediction and early screening, thereby enhancing the accuracy of non-invasive liver tumor detection.

Blood- and Imaging-Derived Biomarkers for Oncological Outcome Modelling in Oropharyngeal Cancer: Exploring the Low-Hanging Fruit

Aims: To assess whether CT-based radiomics and blood-derived biomarkers could improve the prediction of overall survival (OS) and locoregional progression-free survival (LRPFS) in patients with oropharyngeal cancer (OPC) treated with curative-intent RT. Methods: Consecutive OPC patients with primary tumors treated between 2005 and 2021 were included. Analyzed clinical variables included gender, age, smoking history, staging, subsite, HPV status, and blood parameters (baseline hemoglobin levels, neutrophils, monocytes, and platelets, and derived measurements). Radiomic features were extracted from the gross tumor volumes (GTVs) of the primary tumor using pyradiomics. Outcomes of interest were LRPFS and OS. Following feature selection, a radiomic score (RS) was calculated for each patient. Significant variables, along with age and gender, were included in multivariable analysis, and models were retained if statistically significant. The models’ performance was compared by the C-index. Results: One hundred and five patients, predominately male (71%), were included in the analysis. The median age was 59 (IQR: 52–66) years, and stage IVA was the most represented (70%). HPV status was positive in 63 patients, negative in 7, and missing in 35 patients. The median OS follow-up was 6.3 (IQR: 5.5–7.9) years. A statistically significant association between low Hb levels and poorer LRPFS in the HPV-positive subgroup (p = 0.038) was identified. The calculation of the RS successfully stratified patients according to both OS (log-rank p < 0.0001) and LRPFS (log-rank p = 0.0002). The C-index of the clinical and radiomic model resulted in 0.82 [CI: 0.80–0.84] for OS and 0.77 [CI: 0.75–0.79] for LRPFS. Conclusions: Our results show that radiomics could provide clinically significant informative content in this scenario. The best performances were obtained by combining clinical and quantitative imaging variables, thus suggesting the potential of integrative modeling for outcome predictions in this setting of patients.

The effect of hippocampal radiomic features and functional connectivity on the relationship between hippocampal volume and cognitive function in Alzheimer's disease

Hippocampal volume is associated with cognitive function in Alzheimer's disease (AD). Hippocampal radiomic features and resting-state functional connectivity (rs-FC) are promising biomarkers and correlate with AD pathology. However, few studies have been conducted on how hippocampal biomarkers affect the cognition-structure relationship. Therefore, we aimed to investigate the effects of hippocampal radiomic features and resting-state functional connectivity (rs-FC) on this relationship in AD. We enrolled 70 AD patients and 65 healthy controls (HCs). The FreeSurfer software was used to measure hippocampal volume. We selected hippocampal radiomic features to build a model to distinguish AD patients from HCs and used a seed-based approach to calculate the hippocampal rs-FC. Furthermore, we conducted mediation and moderation analyses to investigate the effect of hippocampal radiomic features and rs-FC on the relationship between hippocampal volume and cognition in AD. The results suggested that hippocampal radiomic features mediated the association between bilateral hippocampal volume and cognition in AD. Additionally, patients with AD showed weaker rs-FC between the bilateral hippocampus and right ventral posterior cingulate cortex and stronger rs-FC between the left hippocampus and left insula than HCs. The rs-FC between the hippocampus and insula moderated the relationship between hippocampal volume and cognition in AD, suggesting that this rs-FC could exacerbate or ameliorate the effects of hippocampal volume on cognition and may be essential in improving cognitive function in AD. Our findings may not only expand existing biological knowledge of the interrelationships among hippocampal biomarkers and cognition but also provide potential targets for treatment strategies for AD.

Comprehensive prognostic modeling of locoregional recurrence after radiotherapy for patients with locoregionally advanced hypopharyngeal squamous cell carcinoma

Purpose

To propose and evaluate a comprehensive modeling approach combing radiomics, dosiomics and clinical components, for more accurate prediction of locoregional recurrence risk after radiotherapy for patients with locoregionally advanced HPSCC.

Materials and methods

Clinical data of 77 HPSCC patients were retrospectively investigated, whose median follow-up duration was 23.27 (4.83-81.40) months. From the planning CT and dose distribution, 1321 radiomics and dosiomics features were extracted respectively from planning gross tumor volume (PGTV) region each patient. After stability test, feature dimension was further reduced by Principal Component Analysis (PCA), yielding Radiomic and Dosiomic Principal Components (RPCs and DPCs) respectively. Multiple Cox regression models were constructed using various combinations of RPC, DPC and clinical variables as the predictors. Akaike information criterion (AIC) and C-index were used to evaluate the performance of Cox regression models.

Results

PCA was performed on 338 radiomic and 873 dosiomic features that were tested as stable (ICC₁ > 0.7 and ICC₂ > 0.95), yielding 5 RPCs and DPCs respectively. Three comprehensive features (RPC0, P<0.01, DPC0, P<0.01 and DPC3, P<0.05) were found to be significant in the individual Radiomic or Dosiomic Cox regression models. The model combining the above features and clinical variable (total stage IVB) provided best risk stratification of locoregional recurrence (C-index, 0.815; 95%CI, 0.770-0.859) and prevailing balance between predictive accuracy and complexity (AIC, 143.65) than any other investigated models using either single factors or two combined components.

Conclusion

This study provided quantitative tools and additional evidence for the personalized treatment selection and protocol optimization for HPSCC, a relatively rare cancer. By combining complementary information from radiomics, dosiomics, and clinical variables, the proposed comprehensive model provided more accurate prediction of locoregional recurrence risk after radiotherapy.

Radiogenomic markers enable risk stratification and inference of mutational pathway states in head and neck cancer

PURPOSE

Head and neck squamous cell carcinomas (HNSCCs) are a molecularly, histologically, and clinically heterogeneous set of tumors originating from the mucosal epithelium of the oral cavity, pharynx, and larynx. This heterogeneous nature of HNSCC is one of the main contributing factors to the lack of prognostic markers for personalized treatment. The aim of this study was to develop and identify multi-omics markers capable of improved risk stratification in this highly heterogeneous patient population.

METHODS

In this retrospective study, we approached this issue by establishing radiogenomics markers to identify high-risk individuals in a cohort of 127 HNSCC patients. Hybrid in vivo imaging and whole-exome sequencing were employed to identify quantitative imaging markers as well as genetic markers on pathway-level prognostic in HNSCC. We investigated the deductibility of the prognostic genetic markers using anatomical and metabolic imaging using positron emission tomography combined with computed tomography. Moreover, we used statistical and machine learning modeling to investigate whether a multi-omics approach can be used to derive prognostic markers for HNSCC.

RESULTS

Radiogenomic analysis revealed a significant influence of genetic pathway alterations on imaging markers. A highly prognostic radiogenomic marker based on cellular senescence was identified. Furthermore, the radiogenomic biomarkers designed in this study vastly outperformed the prognostic value of markers derived from genetics and imaging alone.

CONCLUSION

Using the identified markers, a clinically meaningful stratification of patients is possible, guiding the identification of high-risk patients and potentially aiding in the development of effective targeted therapies.

External validation of a CT-based radiomics signature in oropharyngeal cancer: Assessing sources of variation

BACKGROUND AND PURPOSE

Radiomics is a high-throughput approach that allows for quantitative analysis of imaging data for prognostic applications. Medical images are used in oropharyngeal cancer (OPC) diagnosis and treatment planning and these images may contain prognostic information allowing for treatment personalization. However, the lack of validated models has been a barrier to the translation of radiomic research to the clinic. We hypothesize that a previously developed radiomics model for risk stratification in OPC can be validated in a local dataset.

MATERIALS AND METHODS

The radiomics signature predicting overall survival incorporates features derived from the primary gross tumor volume of OPC patients treated with radiation +/- chemotherapy at a single institution (n = 343). Model fit, calibration, discrimination, and utility were evaluated. The signature was compared with a clinical model using overall stage and a model incorporating both radiomics and clinical data. A model detecting dental artifacts on computed tomography images was also validated.

RESULTS

The radiomics signature had a Concordance index (C-index) of 0.66 comparable to the clinical model's C-index of 0.65. The combined model significantly outperformed (C-index of 0.69, p = 0.024) the clinical model, suggesting that radiomics provides added value. The dental artifact model demonstrated strong ability in detecting dental artifacts with an area under the curve of 0.87.

CONCLUSION

This work demonstrates model performance comparable to previous validation work and provides a framework for future independent and multi-center validation efforts. With sufficient validation, radiomic models have the potential to improve traditional systems of risk stratification, treatment personalization and patient outcomes.

Nomogram Based on Clinical and Radiomics Data for Predicting Radiation-induced Temporal Lobe Injury in Patients with Non-metastatic Stage T4 Nasopharyngeal Carcinoma

AIMS

To use pre-treatment magnetic resonance imaging-based radiomics data with clinical data to predict radiation-induced temporal lobe injury (RTLI) in nasopharyngeal carcinoma (NPC) patients with stage T4/N0-3/M0 within 5 years after radiotherapy.

MATERIALS AND METHODS

This study retrospectively examined 98 patients (198 temporal lobes) with stage T4/N0-3/M0 NPC. Participants were enrolled into a training cohort or a validation cohort in a ratio of 7:3. Radiomics features were extracted from pre-treatment magnetic resonance imaging that were T1-and T2-weighted. Spearman rank correlation, the t-test and the least absolute shrinkage and selection operator (LASSO) algorithm were used to select significant radiomics features; machine-learning models were used to generate radiomics signatures (Rad-Scores). Rad-Scores and clinical factors were integrated into a nomogram for prediction of RTLI. Nomogram discrimination was evaluated using receiver operating characteristic analysis and clinical benefits were evaluated using decision curve analysis.

RESULTS

Participants were enrolled into a training cohort (n = 139) or a validation cohort (n = 59). In total, 3568 radiomics features were initially extracted from T1-and T2-weighted images. Age, D_max, D_1cc and 16 stable radiomics features (six from T1-weighted and 10 from T2-weighted images) were identified as independent predictive factors. A greater Rad-Score was associated with a greater risk of RTLI. The nomogram showed good discrimination, with a C-index of 0.85 (95% confidence interval 0.79-0.92) in the training cohort and 0.82 (95% confidence interval 0.71-0.92) in the validation cohort.

CONCLUSION

We developed models for the prediction of RTLI in patients with stage T4/N0-3/M0 NPC using pre-treatment radiomics data and clinical data. Nomograms from these pre-treatment data improved the prediction of RTLI. These results may allow the selection of patients for earlier clinical interventions.

Cone-beam CT delta-radiomics to predict genitourinary toxicities and international prostate symptom of prostate cancer patients: a pilot study

For prostate cancer (PCa) patients treated with definitive radiotherapy (RT), acute and late RT-related genitourinary (GU) toxicities adversely impact disease-specific quality of life. Early warning of potential RT toxicities can prompt interventions that may prevent or mitigate future adverse events. During intensity modulated RT (IMRT) of PCa, daily cone-beam computed tomography (CBCT) images are used to improve treatment accuracy through image guidance. This work investigated the performance of CBCT-based delta-radiomic features (DRF) models to predict acute and sub-acute International Prostate Symptom Scores (IPSS) and Common Terminology Criteria for Adverse Events (CTCAE) version 5 GU toxicity grades for 50 PCa patients treated with definitive RT. Delta-radiomics models were built using logistic regression, random forest for feature selection, and a 1000 iteration bootstrapping leave one analysis for cross validation. To our knowledge, no prior studies of PCa have used DRF models based on daily CBCT images. AUC of 0.83 for IPSS and greater than 0.7 for CTCAE grades were achieved as early as week 1 of treatment. DRF extracted from CBCT images showed promise for the development of models predictive of RT outcomes. Future studies will include using artificial intelligence and machine learning to expand CBCT sample sizes available for radiomics analysis.

Prostate cancer classification using radiomics and machine learning on mp-MRI validated using co-registered histology

BACKGROUND

Multi-parametric magnetic resonance imaging (mp-MRI) is emerging as a useful tool for prostate cancer (PCa) detection but currently has unaddressed limitations. Computer aided diagnosis (CAD) systems have been developed to address these needs, but many approaches used to generate and validate the models have inherent biases.

METHOD

All clinically significant PCa on histology was mapped to mp-MRI using a previously validated registration algorithm. Shape and size matched non-PCa regions were selected using a proposed sampling algorithm to eliminate biases towards shape and size. Further analysis was performed to assess biases regarding inter-zonal variability.

RESULTS

A 5-feature Naïve-Bayes classifier produced an area under the receiver operating characteristic curve (AUC) of 0.80 validated using leave-one-patient-out cross-validation. As mean inter-class area mismatch increased, median AUC trended towards positively biasing classifiers to producing higher AUCs. Classifiers were invariant to differences in shape between PCa and non-PCa lesions (AUC: 0.82 vs 0.82). Performance for models trained and tested only in the peripheral zone was found to be lower than in the central gland (AUC: 0.75 vs 0.95).

CONCLUSION

We developed a radiomics based machine learning system to classify PCa vs non-PCa tissue on mp-MRI validated on accurately co-registered mid-gland histology with a measured target registration error. Potential biases involved in model development were interrogated to provide considerations for future work in this area.

Enhancing the stability of CT radiomics across different volume of interest sizes using parametric feature maps: a phantom study

BACKGROUND

In radiomics studies, differences in the volume of interest (VOI) are often inevitable and may confound the extracted features. We aimed to correct this confounding effect of VOI variability by applying parametric maps with a fixed voxel size.

METHODS

Ten scans of a cup filled with sodium chloride solution were scanned using a multislice computed tomography (CT) unit. Sphere-shaped VOIs with different diameters (4, 8, or 16 mm) were drawn centrally into the phantom. A total of 93 features were extracted conventionally from the original images using PyRadiomics. Using a self-designed and pretested software tool, parametric maps for the same 93 features with a fixed voxel size of 4 mm3 were created. To retrieve the feature values from the maps, VOIs were copied from the original images to preserve the position. Differences in feature quantities between the VOI sizes were tested with the Mann-Whitney U-test and agreement with overall concordance correlation coefficients (OCCC).

RESULTS

Fifty-five conventionally extracted features were significantly different between the VOI sizes, and none of the features showed excellent agreement in terms of OCCCs. When read from the parametric maps, only 8 features showed significant differences, and 3 features showed an excellent OCCC (≥ 0.85). The OCCCs for 89 features substantially increased using the parametric maps.

CONCLUSIONS

This phantom study shows that converting CT images into parametric maps resolves the confounding effect of VOI variability and increases feature reproducibility across VOI sizes.

Radiation and immune checkpoint inhibitor-mediated pneumonitis risk stratification in patients with locally advanced non-small cell lung cancer: role of functional lung radiomics?

BACKGROUND

Patients undergoing chemoradiation and immune checkpoint inhibitor (ICI) therapy for locally advanced non-small cell lung cancer (NSCLC) experience pulmonary toxicity at higher rates than historical reports. Identifying biomarkers beyond conventional clinical factors and radiation dosimetry is especially relevant in the modern cancer immunotherapy era. We investigated the role of novel functional lung radiomics, relative to functional lung dosimetry and clinical characteristics, for pneumonitis risk stratification in locally advanced NSCLC.

METHODS

Patients with locally advanced NSCLC were prospectively enrolled on the FLARE-RT trial (NCT02773238). All received concurrent chemoradiation using functional lung avoidance planning, while approximately half received consolidation durvalumab ICI. Within tumour-subtracted lung regions, 110 radiomics features (size, shape, intensity, texture) were extracted on pre-treatment [99mTc]MAA SPECT/CT perfusion images using fixed-bin-width discretization. The performance of functional lung radiomics for pneumonitis (CTCAE v4 grade 2 or higher) risk stratification was benchmarked against previously reported lung dosimetric parameters and clinical risk factors. Multivariate least absolute shrinkage and selection operator Cox models of time-varying pneumonitis risk were constructed, and prediction performance was evaluated using optimism-adjusted concordance index (c-index) with 95% confidence interval reporting throughout.

RESULTS

Thirty-nine patients were included in the study and pneumonitis occurred in 16/39 (41%) patients. Among clinical characteristics and anatomic/functional lung dosimetry variables, only the presence of baseline chronic obstructive pulmonary disease (COPD) was significantly associated with the development of pneumonitis (HR 4.59 [1.69-12.49]) and served as the primary prediction benchmark model (c-index 0.69 [0.59-0.80]). Discrimination of time-varying pneumonitis risk was numerically higher when combining COPD with perfused lung radiomics size (c-index 0.77 [0.65-0.88]) or shape feature classes (c-index 0.79 [0.66-0.91]) but did not reach statistical significance compared to benchmark models (p > 0.26). COPD was associated with perfused lung radiomics size features, including patients with larger lung volumes (AUC 0.75 [0.59-0.91]). Perfused lung radiomic texture features were correlated with lung volume (adj R2 = 0.84-1.00), representing surrogates rather than independent predictors of pneumonitis risk.

CONCLUSIONS

In patients undergoing chemoradiation with functional lung avoidance therapy and optional consolidative immune checkpoint inhibitor therapy for locally advanced NSCLC, the strongest predictor of pneumonitis was the presence of baseline chronic obstructive pulmonary disease. Results from this novel functional lung radiomics exploratory study can inform future validation studies to refine pneumonitis risk models following combinations of radiation and immunotherapy. Our results support functional lung radiomics as surrogates of COPD for non-invasive monitoring during and after treatment. Further study of clinical, dosimetric, and radiomic feature combinations for radiation and immune-mediated pneumonitis risk stratification in a larger patient population is warranted.

Radiomics software comparison using digital phantom and patient data: IBSI-compliance does not guarantee concordance of feature values

Introduction Radiomics is a promising imaging-based tool which could enhance clinical observation and identify representative features. To avoid different interpretations, the Image Biomarker Standardisation Initiative (IBSI) imposed conditions for harmonisation. This study evaluates IBSI-compliant radiomics applications against a known benchmark and clinical datasets for agreements. Materials and methods The applications compared were RadiomiX Research Toolbox, LIFEx v7.0.0, and syngo.via Frontier Radiomics v1.2.5 (based on PyRadiomics v2.1). Basic assessment included comparing feature names and their formulas. The IBSI digital phantom was used for evaluation against reference values. For agreement evaluation (including same software but different versions), two clinical datasets were used: 27 contrast-enhanced computed tomography (CECT) of colorectal liver metastases and 39 magnetic resonance imaging (MRI) of breast cancer, including intravoxel incoherent motion and dynamic contrast-enhanced (DCE) MRI. The lower 95% confidence interval of intraclass correlation coefficient was used to define excellent agreement (>0.9). Results The three radiomics applications share 41 (3 shape, 8 intensity, 30 texture) out of 172, 84 and 110 features for RadiomiX, LIFEx and syngo.via, respectively, as well as wavelet filtering. The naming convention is, however, different between them. Syngo.via had excellent agreement with the IBSI benchmark, while LIFEx and RadiomiX showed slightly worse agreement. Excellent reproducibility was achieved for shape features only, while intensity and texture features varied considerably with the imaging type. For intensity, excellent agreement ranged from 46% for the DCE maps to 100% for CECT, while this lowered to 44% and 73% for texture features, respectively. Wavelet features produced the greatest variation between applications, with an excellent agreement for only 3% to 11% features. Conclusion Even with IBSI-compliance, the reproducibility of features between radiomics applications is not guaranteed. To evaluate variation, quality assurance of radiomics applications should be performed and repeated when updating to a new version or adding a new modality.

HeLLePhant: A phantom mimicking non-small cell lung cancer for texture analysis in CT images

PURPOSE

Phantoms mimicking human tissue heterogeneity and intensity are required to establish radiomic features robustness in Computed Tomography (CT) images. We developed inserts with two different techniques for the radiomic study of Non-Small Cell Lung Cancer (NSCLC) lesions.

METHODS

We developed two insert prototypes: two 3D-printed made of glycol-modified polyethylene terephthalate (PET-G), and nine with sodium polyacrylate plus iodinated contrast medium. The inserts were put in a handcraft phantom (HeLLePhant). We also analysed four materials of a commercial homogeneous phantom (Catphan® 424) and collected 29 NSCLC patients for comparison. All the CT acquisitions were performed with the same clinical protocol and scanner at 120kVp. The HeLLePhant phantom was scanned ten times in fixed condition at 120kVp and 100kVp for repeatability investigation. We extracted 153 radiomic features using Pyradiomics. To compare the features between phantoms and patients, we computed how many phantom features fell in the range between 10th and 90th percentile of the corresponding patient values. We deemed repeatable the features with a coefficient of variation (CV) less than or equal to 0.10.

RESULTS

The best similarity with the patients was obtained with the polyacrylate inserts (55.6-90.2%), the worst with Catphan (15.7-19.0%). For the PET-G inserts 35.3% and 36.6% of the features match the patient range. We found high repeatability for all the inserts of the HeLLePhant phantom (74.3-100% at 120kVp, 75.7-97.9% at 100kVp), and observed a texture dependency in repeatability.

CONCLUSIONS

Our study shows a promising way to construct heterogeneous inserts mimicking a target tissue for radiomic studies.

A Complex Radiomic Signature in Luminal Breast Cancer from a Weighted Statistical Framework: A Pilot Study

Radiomics is rapidly advancing in precision diagnostics and cancer treatment. However, there are several challenges that need to be addressed before translation to clinical use. This study presents an ad-hoc weighted statistical framework to explore radiomic biomarkers for a better characterization of the radiogenomic phenotypes in breast cancer. Thirty-six female patients with breast cancer were enrolled in this study. Radiomic features were extracted from MRI and PET imaging techniques for malignant and healthy lesions in each patient. To reduce within-subject bias, the ratio of radiomic features extracted from both lesions was calculated for each patient. Radiomic features were further normalized, comparing the z-score, quantile, and whitening normalization methods to reduce between-subjects bias. After feature reduction by Spearman’s correlation, a methodological approach based on a principal component analysis (PCA) was applied. The results were compared and validated on twenty-seven patients to investigate the tumor grade, Ki-67 index, and molecular cancer subtypes using classification methods (LogitBoost, random forest, and linear discriminant analysis). The classification techniques achieved high area-under-the-curve values with one PC that was calculated by normalizing the radiomic features via the quantile method. This pilot study helped us to establish a robust framework of analysis to generate a combined radiomic signature, which may lead to more precise breast cancer prognosis.

Cardiac Magnetic Resonance Radiomics Reveal Differential Impact of Sex, Age, and Vascular Risk Factors on Cardiac Structure and Myocardial Tissue

Background: Cardiovascular magnetic resonance (CMR) radiomics analysis provides multiple quantifiers of ventricular shape and myocardial texture, which may be used for detailed cardiovascular phenotyping. Objectives: We studied variation in CMR radiomics phenotypes by age and sex in healthy UK Biobank participants. Then, we examined independent associations of classical vascular risk factors (VRFs: smoking, diabetes, hypertension, high cholesterol) with CMR radiomics features, considering potential sex and age differential relationships. Design: Image acquisition was with 1.5 Tesla scanners (MAGNETOM Aera, Siemens). Three regions of interest were segmented from short axis stack images using an automated pipeline: right ventricle, left ventricle, myocardium. We extracted 237 radiomics features from each study using Pyradiomics. In a healthy subset of participants (n = 14,902) without cardiovascular disease or VRFs, we estimated independent associations of age and sex with each radiomics feature using linear regression models adjusted for body size. We then created a sample comprising individuals with at least one VRF matched to an equal number of healthy participants (n = 27,400). We linearly modelled each radiomics feature against age, sex, body size, and all the VRFs. Bonferroni adjustment for multiple testing was applied to all p-values. To aid interpretation, we organised the results into six feature clusters. Results: Amongst the healthy subset, men had larger ventricles with dimmer and less texturally complex myocardium than women. Increasing age was associated with smaller ventricles and greater variation in myocardial intensities. Broadly, all the VRFs were associated with dimmer, less varied signal intensities, greater uniformity of local intensity levels, and greater relative presence of low signal intensity areas within the myocardium. Diabetes and high cholesterol were also associated with smaller ventricular size, this association was of greater magnitude in men than women. The pattern of alteration of radiomics features with the VRFs was broadly consistent in men and women. However, the associations between intensity based radiomics features with both diabetes and hypertension were more prominent in women than men. Conclusions: We demonstrate novel independent associations of sex, age, and major VRFs with CMR radiomics phenotypes. Further studies into the nature and clinical significance of these phenotypes are needed.

Stability of Liver Radiomics across Different 3D ROI Sizes-An MRI In Vivo Study

We aimed to evaluate the stability of radiomic features in the liver of healthy individuals across different three-dimensional regions of interest (3D ROI) sizes in T1-weighted (T1w) and T2-weighted (T2w) images from different MR scanners. We retrospectively included 66 examinations of patients without known diseases or pathological imaging findings acquired on three MRI scanners (3 Tesla I: 25 patients, 3 Tesla II: 19 patients, 1.5 Tesla: 22 patients). 3D ROIs of different diameters (10, 20, 30 mm) were drawn on T1w GRE and T2w TSE images into the liver parenchyma (segment V–VIII). We extracted 93 radiomic features from the different ROIs and tested features for significant differences with the Mann–Whitney-U (MWU)-test. The MWU-test revealed significant differences for most second- and higher-order features, indicating a systematic difference dependent on the ROI size. The features mean, median, root mean squared (RMS), 10th percentile, and 90th percentile were not significantly different. We also assessed feature robustness to ROI size variation with overall concordance correlation coefficients (OCCCs). OCCCs across the different ROI-sizes for mean, median, and RMS were excellent (>0.90) in both sequences on all three scanners. These features, therefore, seem robust to ROI-size variation and suitable for radiomic studies of liver MRI.

On dose cube pixel spacing pre-processing for features extraction stability in dosiomic studies

PURPOSE

Dosiomics allows to parameterize regions of interest (ROIs) and to produce quantitative dose features encoding the spatial and statistical distribution of radiotherapy dose. The stability of dosiomics features extraction on dose cube pixel spacing variation has been investigated in this study.

MATERIAL AND METHODS

Based on 17 clinical delivered dose distributions (P_n), dataset has been generated considering all the possible combinations of four dose grid resolutions and two calculation algorithms. Each dose voxel cube has been post-processed considering 4 different dose cube pixel spacing values: 1x1x1, 2x2x2, 3x3x3 mm³ and the one equal to the planning CT. Dosiomics features extraction has been performed from four different ROIs. The stability of each extracted dosiomic feature has been analyzed in terms of coefficient of variation (CV) intraclass correlation coefficient (ICC).

RESULTS

The highest CV mean values were observed for PTV ROI and for the grey level size zone matrix features family. On the other hand, the lowest CV mean values have been found for RING ROI for the grey level co-occurrence matrix features family. P₃ showed the highest percentage of CV >1 (1.14%) followed by P₁₅ (0.41%), P₁ (0.29%) and P13 (0.19%). ICC analysis leads to identify features with an ICC >0.95 that could be considered stable to use in dosiomic studies when different dose cube pixel spacing are considered, especially the features in common among the seventeen plans.

CONCLUSION

Considering the observed variability, dosiomic studies should always provide a report not only on grid resolution and algorithm dose calculation, but also on dose cube pixel spacing.

The Challenge of Choosing the Best Classification Method in Radiomic Analyses: Recommendations and Applications to Lung Cancer CT Images

Radiomics uses high-dimensional sets of imaging features to predict biological characteristics of tumors and clinical outcomes. The choice of the algorithm used to analyze radiomic features and perform predictions has a high impact on the results, thus the identification of adequate machine learning methods for radiomic applications is crucial. In this study we aim to identify suitable approaches of analysis for radiomic-based binary predictions, according to sample size, outcome balancing and the features-outcome association strength. Simulated data were obtained reproducing the correlation structure among 168 radiomic features extracted from Computed Tomography images of 270 Non-Small-Cell Lung Cancer (NSCLC) patients and the associated to lymph node status. Performances of six classifiers combined with six feature selection (FS) methods were assessed on the simulated data using AUC (Area Under the Receiver Operating Characteristics Curves), sensitivity, and specificity. For all the FS methods and regardless of the association strength, the tree-based classifiers Random Forest and Extreme Gradient Boosting obtained good performances (AUC ≥ 0.73), showing the best trade-off between sensitivity and specificity. On small samples, performances were generally lower than in large-medium samples and with larger variations. FS methods generally did not improve performances. Thus, in radiomic studies, we suggest evaluating the choice of FS and classifiers, considering specific sample size, balancing, and association strength.

Stability of Radiomic Features across Different Region of Interest Sizes-A CT and MR Phantom Study

We aimed to evaluate radiomic features' stability across different region of interest (ROI) sizes in CT and MR images. We chose a phantom with a homogenous internal structure so no differences for a feature extracted from ROIs of different sizes would be expected. For this, we scanned a plastic cup filled with sodium chloride solution ten times in CT and per MR sequence (T1-weighted-gradient-echo and T2-weighted-turbo-inversion-recovery-magnitude). We placed sphere-shaped ROIs of different diameters (4, 8, and 16 mm, and 4, 8, and 16 pixels) into the phantom's center. Features were extracted using PyRadiomics. We assessed feature stability across ROI sizes with overall concordance correlation coefficients (OCCCs). Differences were tested for significance with the Mann-Whitney U-test. Of 93 features, 87 T1w-derived, 87 TIRM-derived, and 70 CT-derived features were significantly different between ROI sizes. Among MR-derived features, OCCCs showed excellent (>0.90) agreement for mean, median, and root mean squared for ROI sizes between 4 and 16 mm and pixels. We further observed excellent agreement for 10th and 90th percentile in T1w and 10th percentile in T2w TIRM images. There was no excellent agreement among the OCCCs of CT-derived features. In summary, many features indicated significant differences and only few showed excellent agreement across varying ROI sizes, although we examined a homogenous phantom. Since we considered a small phantom in an experimental setting, further studies to investigate this size effect would be necessary for a generalization. Nevertheless, we believe knowledge about this effect is crucial in interpreting radiomics studies, as features that supposedly discriminate disease entities may only indicate a systematic difference in ROI size.

Supervised machine learning enables non-invasive lesion characterization in primary prostate cancer with [68Ga]Ga-PSMA-11 PET/MRI

PURPOSE

Risk classification of primary prostate cancer in clinical routine is mainly based on prostate-specific antigen (PSA) levels, Gleason scores from biopsy samples, and tumor-nodes-metastasis (TNM) staging. This study aimed to investigate the diagnostic performance of positron emission tomography/magnetic resonance imaging (PET/MRI) in vivo models for predicting low-vs-high lesion risk (LH) as well as biochemical recurrence (BCR) and overall patient risk (OPR) with machine learning.

METHODS

Fifty-two patients who underwent multi-parametric dual-tracer [18F]FMC and [68Ga]Ga-PSMA-11 PET/MRI as well as radical prostatectomy between 2014 and 2015 were included as part of a single-center pilot to a randomized prospective trial (NCT02659527). Radiomics in combination with ensemble machine learning was applied including the [68Ga]Ga-PSMA-11 PET, the apparent diffusion coefficient, and the transverse relaxation time-weighted MRI scans of each patient to establish a low-vs-high risk lesion prediction model (MLH). Furthermore, MBCR and MOPR predictive model schemes were built by combining MLH, PSA, and clinical stage values of patients. Performance evaluation of the established models was performed with 1000-fold Monte Carlo (MC) cross-validation. Results were additionally compared to conventional [68Ga]Ga-PSMA-11 standardized uptake value (SUV) analyses.

RESULTS

The area under the receiver operator characteristic curve (AUC) of the MLH model (0.86) was higher than the AUC of the [68Ga]Ga-PSMA-11 SUVmax analysis (0.80). MC cross-validation revealed 89% and 91% accuracies with 0.90 and 0.94 AUCs for the MBCR and MOPR models respectively, while standard routine analysis based on PSA, biopsy Gleason score, and TNM staging resulted in 69% and 70% accuracies to predict BCR and OPR respectively.

CONCLUSION

Our results demonstrate the potential to enhance risk classification in primary prostate cancer patients built on PET/MRI radiomics and machine learning without biopsy sampling.

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.

Whole-tumor 3D volumetric MRI-based radiomics approach for distinguishing between benign and malignant soft tissue tumors

OBJECTIVES

Our purpose was to differentiate between malignant from benign soft tissue neoplasms using a combination of MRI-based radiomics metrics and machine learning.

METHODS

Our retrospective study identified 128 histologically diagnosed benign (n = 36) and malignant (n = 92) soft tissue lesions. 3D ROIs were manually drawn on 1 sequence of interest and co-registered to other sequences obtained during the same study. One thousand seven hundred eight radiomics features were extracted from each ROI. Univariate analyses with supportive ROC analyses were conducted to evaluate the discriminative power of predictive models constructed using Real Adaptive Boosting (Adaboost) and Random Forest (RF) machine learning approaches.

RESULTS

Univariate analyses demonstrated that 36.89% of individual radiomics varied significantly between benign and malignant lesions at the p ≤ 0.05 level. Adaboost and RF performed similarly well, with AUCs of 0.77 (95% CI 0.68-0.85) and 0.72 (95% CI 0.63-0.81), respectively, after 10-fold cross-validation. Restricting the machine learning models to only sequences extracted from T2FS and STIR sequences maintained comparable performance, with AUCs of 0.73 (95% CI 0.64-0.82) and 0.75 (95% CI 0.65-0.84), respectively.

CONCLUSION

Machine learning decision classifiers constructed from MRI-based radiomics features show promising ability to preoperatively discriminate between benign and malignant soft tissue masses. Our approach maintains applicability even when the dataset is restricted to T2FS and STIR fluid-sensitive sequences, which may bolster practicality in clinical application scenarios by eliminating the need for complex co-registrations for multisequence analysis.

KEY POINTS

• Predictive models constructed from MRI-based radiomics data and machine learning-augmented approaches yielded good discriminative power to correctly classify benign and malignant lesions on preoperative scans, with AUCs of 0.77 (95% CI 0.68-0.85) and 0.72 (95% CI 0.63-0.81) for Real Adaptive Boosting (Adaboost) and Random Forest (RF), respectively. • Restricting the models to only use metrics extracted from T2 fat-saturated (T2FS) and Short-Tau Inversion Recovery (STIR) sequences yielded similar performance, with AUCs of 0.73 (95% CI 0.64-0.82) and 0.75 (95% CI 0.65-0.84) for Adaboost and RF, respectively. • Radiomics-based machine learning decision classifiers constructed from multicentric data more closely mimic the real-world practice environment and warrant additional validation ahead of prospective implementation into clinical workflows.

Voxel-Wise Feature Selection Method for CNN Binary Classification of Neuroimaging Data

Voxel-wise group analysis is presented as a novel feature selection (FS) technique for a deep learning (DL) approach to brain imaging data classification. The method, based on a voxel-wise two-sample t-test and denoted as t-masking, is integrated into the learning procedure as a data-driven FS strategy. t-Masking has been introduced in a convolutional neural network (CNN) for the test bench of binary classification of very-mild Alzheimer’s disease vs. normal control, using a structural magnetic resonance imaging dataset of 180 subjects. To better characterize the t-masking impact on CNN classification performance, six different experimental configurations were designed. Moreover, the performances of the presented FS method were compared to those of similar machine learning (ML) models that relied on different FS approaches. Overall, our results show an enhancement of about 6% in performance when t-masking was applied. Moreover, the reported performance enhancement was higher with respect to similar FS-based ML models. In addition, evaluation of the impact of t-masking on various selection rates has been provided, serving as a useful characterization for future insights. The proposed approach is also highly generalizable to other DL architectures, neuroimaging modalities, and brain pathologies.

Exploring the variability of radiomic features of lung cancer lesions on unenhanced and contrast-enhanced chest CT imaging

PURPOSE

The aim of this methods work is to explore the different behavior of radiomic features resulting by using or not the contrast medium in chest CT imaging of non-small cell lung cancer.

METHODS

Chest CT scans, unenhanced and contrast-enhanced, of 17 patients were selected from images collected as part of the staging process. The major T1-T3 lesion was contoured through a semi-automatic approach. These lesions formed the lesion phantoms to study features behavior. The stability of 94 features of the 3D-Slicer package Radiomics was analyzed. Feature discrimination power was quantified by means of Gini's coefficient. Correlation between distance matrices was evaluated through Mantel statistic. Heatmap, cluster and silhouette plots were applied to find well-structured partitions of lesions.

RESULTS

The Gini's coefficient evidenced a low discrimination power, <0.05, for four features and a large discrimination power, around 0.8, for five features. About 90% of features was affected by the contrast medium, masking tumor lesions variability; thirteen features only were found stable. On 8178 combinations of stable features, only one group of four features produced the same partition of lesions with the silhouette width greater than 0.51, both on unenhanced and contrast-enhanced images.

CONCLUSIONS

Gini's coefficient highlighted the features discrimination power in both CT series. Many features were sensitive to the use of the contrast medium, masking the lesions intrinsic variability. Four stable features produced, on both series, the same partition of cancer lesions with reasonable structure; this may merit being objects of further validation studies and interpretative investigations.

Plausibility and redundancy analysis to select FDG-PET textural features in non-small cell lung cancer

Background

Radiomics refers to the extraction of a large number of image biomarker describing the tumor phenotype displayed in a medical image. Extracted from positron emission tomography (PET) images, radiomics showed diagnostic and prognostic value for several cancer types. However, a large number of radiomic features are nonreproducible or highly correlated with conventional PET metrics. Moreover, radiomic features used in the clinic should yield relevant information about tumor texture. In this study, we propose a framework to identify technical and clinical meaningful features and exemplify our results using a PET non‐small cell lung cancer (NSCLC) dataset.

Materials and methods

The proposed selection procedure consists of several steps. A priori, we only include features that were found to be reproducible in a multicenter setting. Next, we apply a voxel randomization step to identify features that reflect actual textural information, that is, that yield in 90% of the patient scans a value significantly different from random texture. Finally, the remaining features were correlated with standard PET metrics to further remove redundancy with common PET metrics. The selection procedure was performed for different volume ranges, that is, excluding lesions with smaller volumes in order to assess the effect of tumor size on the results. To exemplify our procedure, the selected features were used to predict 1‐yr survival in a dataset of 150 NSCLC patients. A predictive model was built using volume as predictive factor for smaller, and one of the selected features as predictive factor for bigger lesions. The prediction accuracy of the both models were compared with the prediction accuracy of volume.

Results

The number of selected features depended on the lesion size included in the analysis. When including the whole dataset, from 19 features reflecting actual texture only two were found to be not strongly correlated with conventional PET metrics. When excluding lesions smaller than 11.49 and 33.10 mL (25 and 50 percentile of the dataset), four out of 27 features and 13 out of 29 features remained after eliminating features highly correlated with standard PET metrics. When excluding lesions smaller than 103.9 mL (75 percentile), 33 out of 53 features remained. For larger lesions, some of these features outperformed volume in terms of classification accuracy (increase of 4–10%). The combination of using volume as predictor for smaller and one of the selected features for larger lesions also improved the accuracy when compared with volume only (increase from 72% to 76%).

Conclusion

When performing radiomic analysis for smaller lesions, it should be first carefully investigated if a textural feature reflects actual heterogeneity information. Next, verification of the absence of correlation with all conventional PET metrics is essential in order to assess the additional value of radiomic features. Radiomic analysis with lesions larger than 11.4 mL might give additional information to conventional metrics while at the same time reflecting actual tumor texture. Using a combination of volume and one of the selected features for prediction yields promise to increase accuracy and reliability of a radiomic model.

A machine learning-based prediction of the micropapillary/solid growth pattern in invasive lung adenocarcinoma with radiomics

BACKGROUND

Micropapillary/solid (MP/S) growth patterns of lung adenocarcinoma are vital for making clinical decisions regarding surgical intervention. This study aimed to predict the presence of a MP/S component in lung adenocarcinoma using radiomics analysis.

METHODS

Between January 2011 and December 2013, patients undergoing curative invasive lung adenocarcinoma resection were included. Using the "PyRadiomics" package, we extracted 90 radiomics features from the preoperative computed tomography (CT) images. Subsequently, four prediction models were built by utilizing conventional machine learning approaches fitting into radiomics analysis: a generalized linear model (GLM), Naïve Bayes, support vector machine (SVM), and random forest classifiers. The models' accuracy was assessed using a receiver operating curve (ROC) analysis, and the models' stability was validated both internally and externally.

RESULTS

A total of 268 patients were included as a primary cohort, and 36.6% (98/268) of them had lung adenocarcinoma with an MP/S component. Patients with an MP/S component had a higher rate of lymph node metastasis (18.4% versus 5.3%) and worse recurrence-free and overall survival. Five radiomics features were selected for model building, and in the internal validation, the four models achieved comparable performance of MP/S prediction in terms of area under the curve (AUC): GLM, 0.74 [95% confidence interval (CI): 0.65-0.83]; Naïve Bayes, 0.75 (95% CI: 0.65-0.85); SVM, 0.73 (95% CI: 0.61-0.83); and random forest, 0.72 (95% CI: 0.63-0.81). External validation was performed using a test cohort with 193 patients, and the AUC values were 0.70, 0.72, 0.73, and 0.69 for Naïve Bayes, SVM, random forest, and GLM, respectively.

CONCLUSIONS

Radiomics-based machine learning approach is a very strong tool for preoperatively predicting the presence of MP/S growth patterns in lung adenocarcinoma, and can help customize treatment and surveillance strategies.

A deep survival interpretable radiomics model of hepatocellular carcinoma patients

This work aims to identify a new radiomics signature using imaging phenotypes and clinical variables for risk prediction of overall survival (OS) in hepatocellular carcinoma (HCC) patients treated with stereotactic body radiation therapy (SBRT). 167 patients were retrospectively analyzed with repeated nested cross-validation to mitigate overfitting issues. 56 radiomic features were extracted from pre-treatment contrast-enhanced (CE) CT images. 37 clinical factors were obtained from patients' electronic records. Variational autoencoders (VAE) based survival models were designed for radiomics and clinical features and a convolutional neural network (CNN) survival model was used for the CECT. Finally, radiomics, clinical and raw image deep learning network (DNN) models were combined to predict the risk probability for OS. The final models yielded c-indices of 0.579 (95%CI: 0.544-0.621), 0.629 (95%CI: 0.601-0.643), 0.581 (95%CI: 0.553-0.613) and 0.650 (95%CI: 0.635-0.683) for radiomics, clinical, image input and combined models on nested cross validation scheme, respectively. Integrated gradients method was used to interpret the trained models. Our interpretability analysis of the DNN showed that the top ranked features were clinical liver function and liver exclusive of tumor radiomics features, which suggests a prominent role of side effects and toxicities in liver outside the tumor region in determining the survival rate of these patients. In summary, novel deep radiomic analysis provides improved performance for risk assessment of HCC prognosis compared with Cox survival models and may facilitate stratification of HCC patients and personalization of their treatment strategies. Liver function was found to contribute most to the OS for these HCC patients and radiomics can aid in their management.

Prediction of multi-criteria optimization (MCO) parameter efficiency in volumetric modulated arc therapy (VMAT) treatment planning using machine learning (ML)

PURPOSE

To predict the impact of optimization parameter changes on dosimetric plan quality criteria in multi-criteria optimized volumetric-modulated-arc therapy (VMAT) planning prior to optimization using machine learning (ML).

METHODS

A data base comprising a total of 21,266 VMAT treatment plans for 44 cranial and 18 spinal patient geometries was generated. The underlying optimization algorithm is governed by three highly composite parameters which model a combination of important aspects of the solution. Patient geometries were parametrized via volume- and shape properties of the voxel objects and overlap-volume histograms (OVH) of the planning-target-volume (PTV) and a relevant organ-at-risk (OAR). The impact of changes in one of the three optimization parameters on the maximally achievable value range of five dosimetric properties of the resulting dose distributions was studied. To predict the extent of this impact based on patient geometry, treatment site, and current parameter settings prior to optimization, three different ML-models were trained and tested. Precision-recall curves, as well as the area-under-curve (AUC) of the resulting receiver-operator-characteristic (ROC) curves were analyzed for model assessment.

RESULTS

Successful identification of parameter regions resulting in a high variability of dosimetric plan properties depended on the choice of geometry features, the treatment indication and the plan property under investigation. AUC values between 0.82 and 0.99 could be achieved. The best average-precision (AP) values obtained from the corresponding precision/recall curves ranged from 0.71 to 0.99.

CONCLUSIONS

Machine learning models trained on a database of pre-optimized treatment plans can help finding relevant optimization parameter ranges prior to optimization.

Radiomics in medical imaging-"how-to" guide and critical reflection

Radiomics is a quantitative approach to medical imaging, which aims at enhancing the existing data available to clinicians by means of advanced mathematical analysis. Through mathematical extraction of the spatial distribution of signal intensities and pixel interrelationships, radiomics quantifies textural information by using analysis methods from the field of artificial intelligence. Various studies from different fields in imaging have been published so far, highlighting the potential of radiomics to enhance clinical decision-making. However, the field faces several important challenges, which are mainly caused by the various technical factors influencing the extracted radiomic features.The aim of the present review is twofold: first, we present the typical workflow of a radiomics analysis and deliver a practical "how-to" guide for a typical radiomics analysis. Second, we discuss the current limitations of radiomics, suggest potential improvements, and summarize relevant literature on the subject.