To systematically identify radiomics features on CT enterography (CTE) scans which can accurately diagnose active Crohn's disease across multiple sources of variation. Retrospective study of CTE scans curated between 2013 and 2015, comprising 164 subjects (65 male, 99 female; all patients were over the age of 18) with endoscopic confirmation for the presence or absence of active Crohn's disease. All patients had three distinct sets of scans available (full and reduced dose, where the latter had been reconstructed via two different methods), acquired on a single scanner at a single institution. Radiomics descriptors from annotated terminal ileum regions were individually and systematically evaluated for resilience to different imaging variations (changes in dose/reconstruction, batch effects, and simulated annotation differences) via multiple reproducibility measures. Multiple radiomics models (by accounting for each source of variation) were evaluated in terms of classifier area under the ROC curve (AUC) for identifying patients with active Crohn's disease, across separate discovery and hold-out validation cohorts. Radiomics descriptors selected based on resiliency to multiple sources of imaging variation yielded the highest overall classification performance in the discovery cohort (AUC = 0.79 ± 0.04) which also best generalized in hold-out validation (AUC = 0.81). Performance was maintained across multiple doses and reconstructions while also being significantly better (p < 0.001) than non-resilient descriptors or descriptors only resilient to a single source of variation. Radiomics features can accurately diagnose active Crohn's disease on CTE scans across multiple sources of imaging variation via systematic analysis of reproducibility measures. Clinical utility and translatability of radiomics features for diagnosis and characterization of Crohn's disease on CTE scans will be contingent on their reproducibility across multiple types and sources of imaging variation.

Radiomics analysis has emerged as a promising approach to aid in cancer diagnosis and treatment. However, radiomics research currently lacks standardization, and radiomics features can be highly dependent on acquisition and pre-processing techniques used. In this study, we aim to investigate the effect of various image normalization techniques on robustness of radiomics features extracted from breast cancer patient MRI scans.

MRI scans from the publicly available MAMA-MIA dataset and an internal breast MRI test set depicting triple negative breast cancer (TNBC) were used. We compared the effect of commonly used image normalization techniques on radiomics feature robustnessusing Concordance-Correlation-Coefficient (CCC) between multiple combinations of normalization approaches. We also trained machine learning-based prediction models of pathologic complete response (pCR) on radiomics after different normalization techniques were used and compared their areas under the receiver operating characteristic curve (ROC-AUC).

For predicting complete pathological response from pre-treatment breast cancer MRI radiomics, the highest overall ROC-AUC was achieved by using a combination of three different normalization techniques indicating their potentially powerful role when working with heterogeneous imaging data. The effect of normalization was more pronounced with smaller training data and normalization may be less important with increasing abundance of training data. Additionally, we observed considerable differences between MRI data sets and their feature robustness towards normalization.

Overall, we were able to demonstrate the importance of selecting and standardizing normalization methods for accurate and reliable radiomics analysis in breast MRI scans especially with small training data sets.

This study evaluates the effectiveness of machine learning (ML) models that incorporate clinical and 2-deoxy-2-[18F]fluoro-D-glucose ([18F]-FDG)-positron emission tomography (PET)-radiomic features for predicting outcomes in gallbladder cancer patients.

The study analyzed 52 gallbladder cancer patients who underwent pre-treatment [18F]-FDG-PET/CT scans between January 2011 and December 2021. Twenty-seven patients were assigned to the training cohort between January 2011 and January 2018, and the data randomly split into training (70%) and validation (30%) sets. The independent test cohort consisted of 25 patients between February 2018 and December 2021. Eight clinical features (T stage, N stage, M stage, Union for International Cancer Control [UICC] stage, histology, tumor size, carcinoembryonic antigen level, and carbohydrate antigen 19-9 level) and 49 radiomic features were used to forecast progression-free survival (PFS). Three feature selection methods were applied including the univariate statistical feature selection test method, least absolute shrinkage and selection operator Cox regression method and recursive feature elimination method, and two ML algorithms (Cox proportional hazard and random survival forest [RSF]) were employed. Predictive performance was assessed using the concordance index (C-index).

Two clinical variables (UICC stage, N stage) and three radiomic features (total lesion glycolysis, grey-level size-zone matrix_grey level non-uniformity and grey-level run-length matrix_run-length non-uniformity) were identified by the statistical feature selection method as significant for PFS prediction. The RSF model incorporating these features demonstrated strong predictive performance, with C-indices above 0.80 in both training and testing sets (training 0.81, testing 0.89). This model almost closely matched the actual and predicted progression timelines with a low mean absolute error of 1.435, a median absolute error of 0.082, and a root mean square error of 2.359.

This study highlights the potential of using ML approaches with clinical and pre-treatment [18F]-FDG-PET radiomic data for predicting the prognosis of gallbladder cancer.

Cognitively unimpaired (CU) amyloid beta (Aβ)+ individuals with elevated plasma glial fibrillary acidic protein (GFAP) have an increased risk of Alzheimer's disease (AD)-related progression. We tested the utility of plasma GFAP for population enrichment CU populations in clinical trials.

We estimated longitudinal progression, effect size, and costs of hypothetical clinical trials designed to test an estimated 25% drug effect on reducing tau positron emission tomography (PET) accumulation in the medial temporal lobe (MTL) and temporal neocortical region (NEO-T).

CU GFAP+/Aβ+ individuals present an increased annual rate of change and effect size in tau PETMTL and tau PETNEO-T compared to the other groups. An enrichment strategy selecting CU GFAP+/Aβ+ individuals would require a smaller sample size (≈ 57% reduction) and fewer Aβ PET scans (≈ 74% reduction) than trials enriched with Aβ PET alone, reducing total clinical trial costs by up to 64%.

Our results suggest that clinical trials focusing on preclinical AD recruiting Aβ+ individuals with elevated GFAP levels would improve cost effectiveness.

Cognitively unimpaired (CU) glial fibrillary acidic protein (GFAP)+/amyloid beta (Aβ)+ shows increased changes in tau positron emission tomography (PET) . CU GFAP+/Aβ+ enriched clinical trials require a reduced sample size compared to Aβ+ only. CU GFAP+/Aβ+ enrichment reduces Aβ PET scans required and costs. CU GFAP+/Aβ+ enrichment allows the selection of individuals at early stages of the Alzheimer's disease continuum.

Medical imaging data frequently encounter image-generation heterogeneity and class imbalance properties, challenging strong generalized predictive performances with data-driven machine-learning methods. The purpose of this study was to investigate the impact of harmonization and oversampling methods on multi-center imbalanced datasets, with specific application to PET-based radiomics modeling for histologic subtype prediction in non-small cell lung cancer (NSCLC).

The retrospective study included 245 patients with adenocarcinoma (ADC) and 78 patients with squamous cell carcinoma (SCC) from 4 centers. Utilizing 1502 radiomics features per patient, we trained, validated, and tested 4 machine-learning classifiers, to investigate the effect of no harmonization (NoH) or 4 feature harmonization methods, paired with no oversampling (NoO) or 5 oversampling methods on subtype prediction. Model performance was evaluated using the average area under the ROC curve (AUROC) and G-mean via 5 times 5-fold cross-validations. Statistical comparisons of the combined models against baseline (NoH + NoO) were performed for each fold of cross-validation using the DeLong test.

The number of cross-combinations with both AUROC and G-mean outperforming baseline in validation and testing was 15, 4, 2, and 7 (out of 29) for random forest (RF), linear discriminant analysis (LDA), logistic regression (LR), and support vector machine (SVM), respectively. ComBat harmonization combined with oversampling (SMOTE) via RF yielded better performance than baseline (AUROC and G-mean of validation: 0.725 vs. 0.608 and 0.625 vs. 0.398; testing: 0.637 vs. 0.567 and 0.506 vs. 0.287), though statistical significances were not observed.

Applying harmonization and oversampling methods in multi-center imbalanced datasets can improve NSCLC-subtype prediction, but the effect varies widely across classifiers. We have created open-source comparisons of harmonization and oversampling on different classifiers for comprehensive evaluations in different studies.

To investigate the influence of germline succinate dehydrogenase (SDHx) pathogenic variants on 6-[18F]-fluoro-3,4-dihydroxyphenylalanine (18F-DOPA) Positron Emission Tomography (PET) radiomic signature of head and neck paragangliomas (HNPGLs).

Forty-seven patients (20 SDH pathogenic variants carriers) harboring 55 HNPGLs were retrospectively included. HNPGLs were delineated using Nestle adaptive threshold. 128 radiomic features were extracted and harmonized to correct for batch effects. Principal Component Analysis (PCA) was performed to remove redundancy and avoid collinearity. The most representative feature of each component was tested with multivariate stepwise logistic binary regression analysis (LBRA) to identify variables predictive of genetic status.

18F-DOPA Positron Emission Tomography/Computed Tomography (PET/CT) detected 28/29 carotid body HNPGLs, 23/23 jugulotympanic HNPGLs, and 4/4 vagal HNPGLs. SUVmax was significantly higher in SDH-related HNPGLs (p = 0.003). PCA allowed identification of 4 Components. The most representative variables of Component 1 and 2 (including intensity and intensity-related textural features, and not intensity-related textural features, respectively) were Intensity-based (IB)-SUVmedian and Grey Level Run Length Matrix-Long Run Low Gray Level Emphasis (GLRLM-LRLGLE). SDHx HNPGLs exhibited higher activity scores and more homogeneous texture. At patient level, SDHx cases showed significantly higher IB-SUVmedian values (p < 0.001), and lower GLRLM-LRLGLE than sporadic patients (p = 0.005). IB-SUVmedian was found to be an independent predictor of genetic status at lesion (71.0%) and patient level (77.8%).

The present study pioneers the application of 18F-DOPA PET radiomics for HNPGLs, suggesting the influence of germline SDH pathogenic variants on 18F-FDOPA uptake intensity and textural heterogeneity. Integrating radiomics with genetic data provides new insights into the correlation between PET features and underlying molecular dysregulation.

This study compared the prognostic value of total metabolic tumor volume (TMTV) in lymphoma measured with the recently proposed SUV4.0 segmentation threshold versus the 41% SUVmax across LYSA trials and its impact on intensity and dissemination PET features.

A total of 1960 baseline PET/CT scans of Diffuse Large B cell lymphoma (DLBCL), follicular lymphoma (FL) and Hodgkin lymphoma (HL) patients were collected. After a semi-automatic preselection of region of interest, two different segmentation threshold were applied: 41% SUVmax (TMTV41%) and SUV > 4.0 (TMTV4.0).

The correlation between TMTV4.0 and TMTV41% was ρ = 0.90 for DLBCL, ρ = 0.65 for FL and ρ = 0.60 for HL. For SUVmax, SUVpeak, Dmax and Dbulk features, a strong correlation was observed with ρ > 0.95 whatever the lymphoma subtypes. The predictability of TMTV was high and comparable for the two methods with superimposable confidence intervals for the three subtypes. At the 90th percentile TMTV value, the predicted 7-year PFS was 51.13% with TMTV4.0 vs. 49.7% with TMTV41% for DLBCL patients, 45.5% vs. 39.8% for FL patients, and 82.6% vs. 80.5% for HL patients. A minority of patients showed a predicted PFS deviation > 10% between the two methods: 2.33% in DLBCL, 6.51% in FL and 1% in HL.

TMTV measured with the SUV4.0 threshold provides a comparable PFS prediction than the 41%SUVmax method supporting its routine adoption particularly in the diffuse large B cell lymphoma subtype.

Background: Patients with low tumour burden follicular lymphoma (FL) are managed with an initial watchful waiting (WW) approach. The way to better predict the time-to-treatment (TTT) is still under investigation for its possible clinical impact. This study explored whether radiomic features extracted from baseline 18F-FDG PET/CT could predict TTT in FL patients on WW. Methods: Thirty-eight patients on initial WW (grade 1-3a) were retrospectively included from 2010 to 2019. Eighty-one PET/CT morphological and first-level intensity radiomic features were extracted from the total metabolic tumour burden (TMTV), the lesion having the highest SUVmax and a reference volume-of-interest placed on the healthy liver. Models using linear regression (LR) and support vector machine (SVM) were constructed to assess the feasibility of using radiomic features to predict TTT. A leave-one-out cross-validation approach was used to assess the performance. Results: For LR models, we found a root-mean-squared error of 29.4, 28.6, 26.4 and 26.8 and an R2 of 0.03, 0.08, 0.21 and 0.20, respectively, incrementing the features from one to four. Accordingly, the best model included three features: the liver minimum SUV value, the liver SUV skewness and the sum of squared SUV values in the TMTV. For SVM models, accuracies of 0.79, 0.63, 0.76 and 0.68 and areas under the curve of 0.80, 0.72, 0.77 and 0.63 were found, respectively, incrementing the features from one to four. The best performing model used one feature, namely the median value of the lesion containing the SUVmax value. Conclusions: The baseline PET/CT radiomic approach has the potential to predict TTT in FL patients on WW. Integrating radiomics with clinical parameters could further aid in patient stratification.

Coronary artery disease (CAD) has one of the highest mortality rates in humans worldwide. Single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) provides clinicians with myocardial metabolic information non-invasively. However, there are some limitations to interpreting SPECT images performed by physicians or automatic quantitative approaches. Radiomics analyzes images objectively by extracting quantitative features and can potentially reveal biological characteristics that the human eye cannot detect. However, the reproducibility and repeatability of some radiomic features can be highly susceptible to segmentation and imaging conditions.

We aimed to assess the reproducibility of radiomic features extracted from uncorrected MPI-SPECT images reconstructed with 15 different settings before and after ComBat harmonization, along with evaluating the effectiveness of ComBat in realigning feature distributions.

A total of 200 patients (50% normal and 50% abnormal) including rest and stress (without attenuation and scatter corrections) MPI-SPECT images were included. Images were reconstructed using 15 combinations of filter cut-off frequencies, filter orders, filter types, reconstruction algorithms, number of iterations and subsets resulting in 6000 images. Image segmentation was performed on the left ventricle in the first reconstruction for each patient and applied to 14 others. A total of 93 radiomic features were extracted from the segmented area, and ComBat was used to harmonize them. The intraclass correlation coefficient (ICC) and overall concordance correlation coefficient (OCCC) tests were performed before and after ComBat to examine the impact of each parameter on feature robustness and to assess harmonization efficiency. The ANOVA and the Kruskal-Wallis tests were performed to evaluate the effectiveness of ComBat in correcting feature distributions. In addition, the Student's t-test, Wilcoxon rank-sum, and signed-rank tests were implemented to assess the significance level of the impacts made by each parameter of different batches and patient groups (normal vs. abnormal) on radiomic features.

Before applying ComBat, the majority of features (ICC: 82, OCCC: 61) achieved high reproducibility (ICC/OCCC ≥ 0.900) under every batch except Reconstruction. The largest and smallest number of poor features (ICC/OCCC < 0.500) were obtained by IterationSubset and Order batches, respectively. The most reliable features were from the first-order (FO) and gray-level co-occurrence matrix (GLCM) families. Following harmonization, the minimum number of robust features increased (ICC: 84, OCCC: 78). Applying ComBat showed that Order and Reconstruction were the least and the most responsive batches, respectively. The most robust families, in a descending order, were found to be FO, neighborhood gray-tone difference matrix (NGTDM), GLCM, gray-level run length matrix (GLRLM), gray-level size zone matrix (GLSZM), and gray-level dependence matrix (GLDM) under Cut-off, Filter, and Order batches. The Wilcoxon rank-sum test showed that the number of robust features significantly differed under most batches in the Normal and Abnormal groups.

The majority of radiomic features show high levels of robustness across different OSEM reconstruction parameters in uncorrected MPI-SPECT. ComBat is effective in realigning feature distributions and enhancing radiomic features reproducibility.

This study aimed to assess the reproducibility of MRI-derived radiomic features across multiple gray-level discretization levels for classifying Parkinson's disease (PD) subtypes, and to evaluate the impact of ComBat harmonization on feature stability and machine learning performance.

T1-weighted MRI scans from 140 PD patients (70 tremor-dominant, 70 postural instability gait difficulty) and 70 healthy controls were obtained from the Parkinson's progression markers initiative (PPMI) database. Radiomic features were extracted from 16 brain regions using 6 discretization levels (8, 16, 32, 64, 128, and 256 bins). ComBat harmonization was applied using a combined batch variable incorporating both scanner models and discretization levels. Intraclass correlation coefficients (ICC) and Kruskal-Wallis tests assessed feature reproducibility before and after harmonization. Support vector machine classifiers were used for PD subtype classification.

ComBat harmonization significantly improved feature reproducibility across all feature groups. The percentage of features showing excellent robustness (ICC ≥ 0.90) increased substantially after harmonization. The proportion of features significantly affected by discretization levels was reduced following harmonization. Classification accuracy improved dramatically, from a range of 0.42-0.49 before harmonization to 0.86-0.96 after harmonization across most discretization levels. AUC values similarly increased from 0.60-0.67 to 0.93-0.99 after harmonization.

ComBat harmonization significantly enhanced the reproducibility of radiomic features across discretization levels and improved PD subtype classification performance. This study highlights the importance of harmonization in radiomics research for PD and suggests potential clinical applications in personalized treatment planning.

To develop and identify machine learning (ML) models using pretreatment 2-deoxy-2-[18F]fluoro-D-glucose ([18F]-FDG)-positron emission tomography (PET)-based radiomic features to differentiate benign from malignant parotid gland diseases (PGDs).

This retrospective study included 62 patients with 63 PGDs who underwent pretreatment [18F]-FDG-PET/computed tomography (CT). The lesions were assigned to the training (n = 44) and testing (n = 19) cohorts. In total, 49 [18F]-FDG-PET-based radiomic features were utilized to differentiate benign from malignant PGDs using five different conventional ML algorithmic models (random forest, neural network, k-nearest neighbors, logistic regression, and support vector machine) and the deep learning (DL)-based ensemble ML model. In the training cohort, each conventional ML model was constructed using the five most important features selected by the recursive feature elimination method with the tenfold cross-validation and synthetic minority oversampling technique. The DL-based ensemble ML model was constructed using the five most important features of the bagging and multilayer stacking methods. The area under the receiver operating characteristic curves (AUCs) and accuracies were used to compare predictive performances.

In total, 24 benign and 39 malignant PGDs were identified. Metabolic tumor volume and four GLSZM features (GLSZM_ZSE, GLSZM_SZE, GLSZM_GLNU, and GLSZM_ZSNU) were the five most important radiomic features. All five features except GLSZM_SZE were significantly higher in malignant PGDs than in benign ones (each p < 0.05). The DL-based ensemble ML model had the best performing classifier in the training and testing cohorts (AUC = 1.000, accuracy = 1.000 vs AUC = 0.976, accuracy = 0.947).

The DL-based ensemble ML model using [18F]-FDG-PET-based radiomic features can be useful for differentiating benign from malignant PGDs. The DL-based ensemble ML model using [18F]-FDG-PET-based radiomic features can overcome the previously reported limitation of [18F]-FDG-PET/CT scan for differentiating benign from malignant PGDs. The DL-based ensemble ML approach using [18F]-FDG-PET-based radiomic features can provide useful information for managing PGD.

Dose escalation may only be suitable for some patients with esophageal cancer (EC) undergoing concurrent chemoradiotherapy (CCRT). This study aimed to identify specific subgroups of patients for whom dose escalation was most beneficial.

Between January 2008 and December 2022, 187 patients with EC underwent CCRT; 94 patients received high-dose (HD) radiotherapy (RT; >64.8 Gy), and 93 patients received low-dose (LD) RT. We developed a model on the basis of clinical and radiomic features to compare the predicted survival probabilities of patients with EC receiving HD- and LD-RT. Patients suitable for HD-RT could be identified. We validated our findings of a suitable HD subgroup by evaluating the actual overall survival (OS) across different subgroups in the testing set.

Our model comprised HD and LD submodels, each predicting patient survival under their respective RT doses. The HD and LD submodels achieved concordance indexes of 0.78 and 0.75 in their respective testing sets. The average areas under the receiver operating characteristic curve over years 1-3 were 0.890 and 0.807 in the HD and LD testing sets, respectively. By comparing patients' predicted survival under HD- and LD-RT in the model, we classified the patients in the testing set into the HD-suitable (HDS) subgroup and the HD-unsuitable subgroup. In the subgroup analysis, HD-RT led to a better OS benefit for the identified HDS subgroup compared with LD-RT (P = .014). Among the HD-treated patients, the HDS subgroup showed better OS than did patients identified as unsuitable (P < .001).

We identified a suitable subgroup of patients with EC who might benefit from HD-RT. This model could aid clinicians in prescribing RT doses.

Previously, we developed an "activity painting" tool for PET image simulation; however, it could simulate heterogeneous patterns only in the air. We aimed to improve this phantom technique to simulate arbitrary lesions in a radioactive background to perform relevant multi-center radiomic analysis. We conducted measurements moving a 22Na point source in a 20-liter background volume filled with 5 kBq/mL activity with an adequately controlled robotic system to prevent the surge of the water. Three different lesion patterns were "activity-painted" in five PET/CT cameras, resulting in 8 different reconstructions. We calculated 46 radiomic indeces (RI) for each lesion and imaging setting, applying absolute and relative discretization. Reproducibility and reliability were determined by the inter-setting coefficient of variation (CV) and the intraclass correlation coefficient (ICC). Hypothesis tests were used to compare RI between lesions. By simulating precisely the same lesions, we confirmed that the reconstructed voxel size and the spatial resolution of different PET cameras were critical for higher order RI. Considering conventional RIs, the SUVpeak and SUVmean proved the most reliable (CV<10%). CVs above 25% are more common for higher order RIs, but we also found that low CVs do not necessarily imply robust parameters but often rather insensitive RIs. Based on the hypothesis test, most RIs could clearly distinguish between the various lesions using absolute resampling. ICC analysis also revealed that most RIs were more reproducible with absolute discretization. The activity painting method in a real radioactive environment proved suitable for precisely detecting the radiomic differences derived from the different camera settings and texture characteristics. We also found that inter-setting CV is not an appropriate metric for analyzing RI parameters' reliability and robustness. Although multicentric cohorts are increasingly common in radiomics analysis, realistic texture phantoms can provide indispensable information on the sensitivity of an RI and how an individual RI parameter measures the texture.

Radiomics is an emerging field of medical imaging that aims at improving the accuracy of diagnosis, prognosis, treatment planning and monitoring non-invasively through the automated or semi-automated quantitative analysis of high-dimensional image features. Specifically in the field of nuclear medicine, radiomics utilizes imaging methods such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) to evaluate biomarkers related to metabolism, blood flow, cellular activity and some biological pathways. Lung cancer ranks among the leading causes of cancer-related deaths globally, and radiomics analysis has shown great potential in guiding individualized therapy, assessing treatment response, and predicting clinical outcomes. In this review, we summarize the current state-of-the-art radiomics progress in lung cancer, highlighting the potential benefits and existing limitations of this approach. The radiomics workflow was introduced first including image acquisition, segmentation, feature extraction, and model building. Then the published literatures were described about radiomics-based prediction models for lung cancer diagnosis, differentiation, prognosis and efficacy evaluation. Finally, we discuss current challenges and provide insights into future directions and potential opportunities for integrating radiomics into routine clinical practice.

Osimertinib, a third-generation EGFR-TKI, is known for its high efficacy against brain metastases (BM) in non-small cell lung cancer (NSCLC) due to its ability to penetrate the blood-brain barrier. This study aims to evaluate the use of brain MRI radiomics in predicting the intracranial efficacy to osimertinib in NSCLC patients with BM.

This study analyzed 115 brain metastases from NSCLC patients with the EGFR-T790M mutation treated with second-line osimertinib. The primary endpoint was intracranial response, and the secondary endpoint was intracranial progression-free survival (iPFS). We performed tumor delineation, image preprocessing, and radiomics feature extraction. Using a 5-fold cross-validation strategy, we built radiomic models with eight feature selectors and eight machine learning classifiers. The models' performance was evaluated by the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis.

The dataset of 115 brain metastases was divided into training and validation sets in a 7:3 ratio. The radiomic model utilizing the mRMR feature selector and stepwise logistic regression classifier showed the highest predictive accuracy, with AUCs of 0.879 for the training cohort and 0.786 for the validation cohort. This model outperformed a clinical-MRI morphological model, which included age, ring enhancement, and peritumoral edema (AUC: 0.794 for the training cohort and 0.697 for the validation cohort). The radiomic model also showed strong performance in calibration and decision curve analyses. Using a radiomic-score threshold of 199, patients were classified into two groups with significantly different median iPFS (3.0 months vs. 15.4 months, p < 0.001).

This study demonstrates that MRI radiomics can effectively predict the intracranial efficacy of osimertinib in NSCLC patients with brain metastases. This approach holds promise for assisting clinicians in personalizing treatment strategies.

Hip involvement is an important reason of disability in patients with ankylosing spondylitis (AS). Unveiling the potential phenotype of hip involvement in AS remains an unmet need to understand its biological mechanisms and improve clinical decision-making. Radiomics, a promising quantitative image analysis method that had been successfully used to describe the phenotype of a wide variety of diseases, while it was less reported in AS. The objective of this study was to investigate the feasibility of radiomics-based approach to profile hip involvement in AS.

A total of 167 patients with AS was included. Radiomic features were extracted from pelvis MRI after image preprocessing and feature engineering. Then, we performed unsupervised machine learning method to derive radiomics-based phenotypes. The validation and interpretation of derived phenotypes were conducted from the perspectives of clinical backgrounds and MRI characteristics. The association between derived phenotypes and radiographic outcomes was evaluated by multivariable analysis.

1321 robust radiomic features were extracted and four biologically distinct phenotypes were derived. According to patient clinical backgrounds, phenotype I (38, 22.8%) and II (34, 20.4%) were labelled as high-risk while phenotype III (24, 14.4%) and IV (71, 42.5%) were at low risk for hip involvement. Consistently, the high-risk phenotypes were associated with higher prevalence of MRI-detected lesion than the low-risk. Moreover, phenotype I had significant acute inflammation signs than phenotype II, while phenotype IV was enthesitis-predominant. Importantly, the derived phenotypes were highly predictive of radiographic outcomes of patients, as the high-risk phenotypes were 3 times more likely to have radiological hip lesion than the low-risk [27 (58.7%) vs 16 (28.6%); adjusted odds ratio (OR) 2.95 (95% CI 1.10, 7.92)].

We confirmed for the first time, the clinical actionability of profiling hip involvement in AS by radiomics method. Four distinct phenotypes of hip involvement in AS were identified and importantly, the high-risk phenotypes could predict structural damage of hip involvement in AS.

To investigate the prognostic value of 18F-FDG PET-based intensity, volumetric features, and deep learning (DL) across different generations of PET scanners in patients with epidermal growth factor receptor (EGFR)-mutated lung adenocarcinoma receiving tyrosine kinase inhibitor (TKI) treatment.

We retrospectively analyzed the pre-treatment 18F-FDG PET of 217 patients with advanced-stage lung adenocarcinoma and actionable EGFR mutations who received TKI as first-line treatment. Patients were separated into analog (n = 166) and digital (n = 51) PET cohorts. 18F-FDG PET-derived intensity, volumetric features, ResNet-50 DL of the primary tumor, and clinical variables were used to predict progression-free survival (PFS). Independent prognosticators were used to develop prediction model. Model was developed and validated in the analog and digital PET cohorts, respectively.

In the analog PET cohort, female sex, stage IVB status, exon 19 deletion, SUVmax, metabolic tumor volume, and positive DL prediction independently predicted PFS. The model devised from these six prognosticators significantly predicted PFS in the analog (HR = 1.319, p < 0.001) and digital PET cohorts (HR = 1.284, p = 0.001). Our model provided incremental prognostic value to staging status (c-indices = 0.738 vs. 0.558 and 0.662 vs. 0.598 in the analog and digital PET cohorts, respectively). Our model also demonstrated a significant prognostic value for overall survival (HR = 1.198, p < 0.001, c-index = 0.708 and HR = 1.256, p = 0.021, c-index = 0.664 in the analog and digital PET cohorts, respectively).

Combining 18F-FDG PET-based intensity, volumetric features, and DL with clinical variables may improve the survival stratification in patients with advanced EGFR-mutated lung adenocarcinoma receiving TKI treatment. Implementing the prediction model across different generations of PET scanners may be feasible and facilitate tailored therapeutic strategies for these patients.

This study investigated the impact of ComBat harmonization on the reproducibility of radiomic features extracted from magnetic resonance images (MRI) acquired on different scanners, using various data acquisition parameters and multiple image pre-processing techniques using a dedicated MRI phantom. Four scanners were used to acquire an MRI of a nonanatomic phantom as part of the TCIA RIDER database. In fast spin-echo inversion recovery (IR) sequences, several inversion durations were employed, including 50, 100, 250, 500, 750, 1000, 1500, 2000, 2500, and 3000 ms. In addition, a 3D fast spoiled gradient recalled echo (FSPGR) sequence was used to investigate several flip angles (FA): 2, 5, 10, 15, 20, 25, and 30 degrees. Nineteen phantom compartments were manually segmented. Different approaches were used to pre-process each image: Bin discretization, Wavelet filter, Laplacian of Gaussian, logarithm, square, square root, and gradient. Overall, 92 first-, second-, and higher-order statistical radiomic features were extracted. ComBat harmonization was also applied to the extracted radiomic features. Finally, the Intraclass Correlation Coefficient (ICC) and Kruskal-Wallis's (KW) tests were implemented to assess the robustness of radiomic features. The number of non-significant features in the KW test ranged between 0-5 and 29-74 for various scanners, 31-91 and 37-92 for three times tests, 0-33 to 34-90 for FAs, and 3-68 to 65-89 for IRs before and after ComBat harmonization, with different image pre-processing techniques, respectively. The number of features with ICC over 90% ranged between 0-8 and 6-60 for various scanners, 11-75 and 17-80 for three times tests, 3-83 to 9-84 for FAs, and 3-49 to 3-63 for IRs before and after ComBat harmonization, with different image pre-processing techniques, respectively. The use of various scanners, IRs, and FAs has a great impact on radiomic features. However, the majority of scanner-robust features is also robust to IR and FA. Among the effective parameters in MR images, several tests in one scanner have a negligible impact on radiomic features. Different scanners and acquisition parameters using various image pre-processing might affect radiomic features to a large extent. ComBat harmonization might significantly impact the reproducibility of MRI radiomic features.

This study aimed to establish and evaluate the value of integrated models involving 18F-FDG PET/CT-based radiomics and clinicopathological information in the prediction of pathological complete response (pCR) to neoadjuvant therapy (NAT) for non-small cell lung cancer (NSCLC).

A total of 106 eligible NSCLC patients were included in the study. After volume of interest (VOI) segmentation, 2,016 PET-based and 2,016 CT-based radiomic features were extracted. To select an optimal machine learning model, a total of 25 models were constructed based on five sets of machine learning classifiers combined with five sets of predictive feature resources, including PET-based alone radiomics, CT-based alone radiomics, PET/CT-based radiomics, clinicopathological features, and PET/CT-based radiomics integrated with clinicopathological features. Area under the curves (AUCs) of receiver operator characteristic (ROC) curves were used as the main outcome to assess the model performance.

The hybrid PET/CT-derived radiomic model outperformed PET-alone and CT-alone radiomic models in the prediction of pCR to NAT. Moreover, addition of clinicopathological information further enhanced the predictive performance of PET/CT-derived radiomic model. Ultimately, the support vector machine (SVM)-based PET/CT radiomics combined clinicopathological information presented an optimal predictive efficacy with an AUC of 0.925 (95% CI 0.869-0.981) in the training cohort and an AUC of 0.863 (95% CI 0.740-0.985) in the test cohort. The developed nomogram involving radiomics and pathological type was suggested as a convenient tool to enable clinical application.

The 18F-FDG PET/CT-based SVM radiomics integrated with clinicopathological information was an optimal model to non-invasively predict pCR to NAC for NSCLC.

Neuroscience studies have improved significantly through multi-site datasets. Brain dynamic functional network connectivity (dFNC) often uses clustering to represent various patterns of functional connectivity that repeat with time. Data collected from different sources combined may be affected by site effects. This can be mitigated by harmonizing the dataset before doing additional analysis. Here, we explore the impact of ComBat harmonization on dFNC data collected from two different mild traumatic brain injury (mTBI) studies. Our research showed that site effects significantly influence the formation of dFNC states. We developed a new method to measure the distribution of samples among the clusters based on site effects in clustering called an 'Inclusivity' model. With this inclusivity model, we evaluate how site effects influence the clustering of dFNC states before and after data harmonization.

To investigate the [68Ga]DOTATOC PET radiomic profile of head and neck paragangliomas (HNPGLs) and identify radiomic characteristics useful as predictors of succinate dehydrogenase genes (SDHx) pathogenic variants.

Sporadic and SDHx HNPGL patients, who underwent [68Ga]DOTATOC PET/CT, were retrospectively included. HNPGLs were analyzed using LIFEx software, and extracted features were harmonized to correct for batch effects and confronted testing for multiple comparison. Stepwise discriminant analysis was conducted to remove redundancy and identify best discriminating features. ROC analysis was used to define optimal cut-offs. Multivariate decision-tree analysis was performed using CHAID method.

34 patients harboring 60 HNPGLs (51 SDHx in 25 patients) were included. Three sporadic and nine SDHx HNPGLs were metastatic. At stepwise discriminant analysis, both GLSZM-Zone Size Non-Uniformity (ZSNU, reflecting tumor heterogeneity) and IB-TLSRE (total lesion somatostatin receptor expression) were independent predictors of genetic status, with 96.4% of lesions and 91.6% of patients correctly classified after cross validation (p < 0.001). Among non-metastatic patients, GLSZM-ZSNU and IB-TLSRE were significantly higher in sporadic than SDHx HNPGLs (p < 0.001). No differences were revealed in metastatic patients. Decision-tree analysis highlights multifocality and IB-TLSRE as useful variables, correctly identifying 6/9 sporadic and 24/25 SDHx patients. Model failed to classify one SDHA and three sporadic patients (2 metastatic).

Radiomics features GLSZM-ZSNU and IB-TLSRE appear to reflect HNPGLs SDHx status and tumor behavior (metastatic vs. non-metastatic). If validated, especially IB-TLSRE might represent a simple and time-efficient radiomic index for SDHx variants early screening and prediction of tumor behavior in HNPGL cases.

To improve reproducibility and predictive performance of PET radiomic features in multicentric studies by cycle-consistent generative adversarial network (GAN) harmonization approaches.

GAN-harmonization was developed to harmonize whole-body PET scans to perform image style and texture translation between different centers and scanners. GAN-harmonization was evaluated by application to two retrospectively collected open datasets and different tasks. First, GAN-harmonization was performed on a dual-center lung cancer cohort (127 female, 138 male) where the reproducibility of radiomic features in healthy liver tissue was evaluated. Second, GAN-harmonization was applied to a head and neck cancer cohort (43 female, 154 male) acquired from three centers. Here, the clinical impact of GAN-harmonization was analyzed by predicting the development of distant metastases using a logistic regression model incorporating first-order statistics and texture features from baseline 18F-FDG PET before and after harmonization.

Image quality remained high (structural similarity: left kidney ≥ 0.800, right kidney ≥ 0.806, liver ≥ 0.780, lung ≥ 0.838, spleen ≥ 0.793, whole-body ≥ 0.832) after image harmonization across all utilized datasets. Using GAN-harmonization, inter-site reproducibility of radiomic features in healthy liver tissue increased at least by ≥ 5 ± 14% (first-order), ≥ 16 ± 7% (GLCM), ≥ 19 ± 5% (GLRLM), ≥ 16 ± 8% (GLSZM), ≥ 17 ± 6% (GLDM), and ≥ 23 ± 14% (NGTDM). In the head and neck cancer cohort, the outcome prediction improved from AUC 0.68 (95% CI 0.66-0.71) to AUC 0.73 (0.71-0.75) by application of GAN-harmonization.

GANs are capable of performing image harmonization and increase reproducibility and predictive performance of radiomic features derived from different centers and scanners.

To investigate the usefulness of machine learning (ML) models using pretreatment 18F-FDG-PET-based radiomic features for predicting adverse clinical events (ACEs) in patients with cardiac sarcoidosis (CS).

This retrospective study included 47 patients with CS who underwent 18F-FDG-PET/CT scan before treatment. The lesions were assigned to the training (n = 38) and testing (n = 9) cohorts. In total, 49 18F-FDG-PET-based radiomic features and the visibility of right ventricle 18F-FDG uptake were used to predict ACEs using seven different ML algorithms (namely, decision tree, random forest [RF], neural network, k-nearest neighbors, Naïve Bayes, logistic regression, and support vector machine [SVM]) with tenfold cross-validation and the synthetic minority over-sampling technique. The ML models were constructed using the top four features ranked by the decrease in Gini impurity. The AUCs and accuracies were used to compare predictive performances.

Patients who developed ACEs presented with a significantly higher surface area and gray level run length matrix run length non-uniformity (GLRLM_RLNU), and lower neighborhood gray-tone difference matrix_coarseness and sphericity than those without ACEs (each, p < 0.05). In the training cohort, all seven ML algorithms had a good classification performance with AUC values of > 0.80 (range: 0.841-0.944). In the testing cohort, the RF algorithm had the highest AUC and accuracy (88.9% [8/9]) with a similar classification performance between training and testing cohorts (AUC: 0.945 vs 0.889). GLRLM_RLNU was the most important feature of the modeling process of this RF algorithm.

ML analyses using 18F-FDG-PET-based radiomic features may be useful for predicting ACEs in patients with CS.

While precision medicine applications of radiomics analysis are promising, differences in image acquisition can cause "batch effects" that reduce reproducibility and affect downstream predictive analyses. Harmonization methods such as ComBat have been developed to correct these effects, but evaluation methods for quantifying batch effects are inconsistent. In this study, we propose the use of the multivariate statistical test PERMANOVA and the Robust Effect Size Index (RESI) to better quantify and characterize batch effects in radiomics data. We evaluate these methods in both simulated and real radiomics features extracted from full-field digital mammography (FFDM) data. PERMANOVA demonstrated higher power than standard univariate statistical testing, and RESI was able to interpretably quantify the effect size of site at extremely large sample sizes. These methods show promise as more powerful and interpretable methods for the detection and quantification of batch effects in radiomics studies.

Defining the exact histological features of salivary gland malignancies before treatment remains an unsolved problem that compromises the ability to tailor further therapeutic steps individually. Radiomics, a new methodology to extract quantitative information from medical images, could contribute to characterizing the individual cancer phenotype already before treatment in a fast and non-invasive way. Consequently, the standardization and implementation of radiomic analysis in the clinical routine work to predict histology of salivary gland cancer (SGC) could also provide improvements in clinical decision-making. In this study, we aimed to investigate the potential of radiomic features as imaging biomarker to distinguish between high grade and low-grade salivary gland malignancies. We have also investigated the effect of image and feature level harmonization on the performance of radiomic models. For this study, our dual center cohort consisted of 126 patients, with histologically proven SGC, who underwent curative-intent treatment in two tertiary oncology centers. We extracted and analyzed the radiomics features of 120 pre-therapeutic MRI images with gadolinium (T1 sequences), and correlated those with the definitive post-operative histology. In our study the best radiomic model achieved average AUC of 0.66 and balanced accuracy of 0.63. According to the results, there is significant difference between the performance of models based on MRI intensity normalized images + harmonized features and other models (p value < 0.05) which indicates that in case of dealing with heterogeneous dataset, applying the harmonization methods is beneficial. Among radiomic features minimum intensity from first order, and gray level-variance from texture category were frequently selected during multivariate analysis which indicate the potential of these features as being used as imaging biomarker. The present bicentric study presents for the first time the feasibility of implementing MR-based, handcrafted radiomics, based on T1 contrast-enhanced sequences and the ComBat harmonization method in an effort to predict the formal grading of salivary gland carcinoma with satisfactory performance.

Objectives: Radiomics and machine learning are innovative approaches to improve the clinical management of NSCLC. However, there is less information about the additive value of FDG PET-based radiomics compared with clinical and imaging variables. Methods: This retrospective study included 320 NSCLC patients who underwent PET/CT with FDG at initial staging. VOIs were placed on primary tumors only. We included a total of 94 variables, including 87 textural features extracted from PET studies, SUVmax, MTV, TLG, TNM stage, histology, age, and gender. We used the least absolute shrinkage and selection operator (LASSO) regression to select variables with the highest predictive value. Although several radiomics variables are available, the added value of these predictors compared with clinical and imaging variables is still under evaluation. Three hundred and twenty NSCLC patients were included in this retrospective study and underwent 18F-FDG PET/CT at initial staging. In this study, we evaluated 94 variables, including 87 textural features, SUVmax, MTV, TLG, TNM stage, histology, age, and gender. Image-based predictors were extracted from a volume of interest (VOI) positioned on the primary tumor. The least absolute shrinkage and selection operator (LASSO) Cox regression was used to reduce the number of variables and select only those with the highest predictive value. The predictive model implemented with the variables selected using the LASSO analysis was compared with a reference model using only a tumor stage and SUVmax. Results: NGTDM coarseness, SUVmax, and TNM stage survived the LASSO analysis and were used for the radiomic model. The AUCs obtained from the reference and radiomic models were 80.82 (95%CI, 69.01-92.63) and 81.02 (95%CI, 69.07-92.97), respectively (p = 0.98). The median OS in the reference model was 17.0 months in high-risk patients (95%CI, 11-21) and 113 months in low-risk patients (HR 7.47, p < 0.001). In the radiomic model, the median OS was 16.5 months (95%CI, 11-20) and 113 months in high- and low-risk groups, respectively (HR 9.64, p < 0.001). Conclusions: Our results indicate that a radiomic model composed using the tumor stage, SUVmax, and a selected radiomic feature (NGTDM_Coarseness) predicts survival in NSCLC patients similarly to a reference model composed only by the tumor stage and SUVmax. Replication of these preliminary results is necessary.

The aim of the study was to externally validate two AI models for the classification of prostate mpMRI sequences and segmentation of the prostate gland on T2WI.

MpMRI data from 719 patients were retrospectively collected from two hospitals, utilizing nine MR scanners from four different vendors, over the period from February 2018 to May 2022. Med3D deep learning pretrained architecture was used to perform image classification,UNet-3D was used to segment the prostate gland. The images were classified into one of nine image types by the mode. The segmentation model was validated using T2WI images. The accuracy of the segmentation was evaluated by measuring the DSC, VS,AHD.Finally,efficacy of the models was compared for different MR field strengths and sequences.

20,551 image groups were obtained from 719 MR studies. The classification model accuracy is 99%, with a kappa of 0.932. The precision, recall, and F1 values for the nine image types had statistically significant differences, respectively (all P < 0.001). The accuracy for scanners 1.436 T, 1.5 T, and 3.0 T was 87%, 86%, and 98%, respectively (P < 0.001). For segmentation model, the median DSC was 0.942 to 0.955, the median VS was 0.974 to 0.982, and the median AHD was 5.55 to 6.49 mm,respectively.These values also had statistically significant differences for the three different magnetic field strengths (all P < 0.001).

The AI models for mpMRI image classification and prostate segmentation demonstrated good performance during external validation, which could enhance efficiency in prostate volume measurement and cancer detection with mpMRI.

These models can greatly improve the work efficiency in cancer detection, measurement of prostate volume and guided biopsies.

Magnetic resonance imaging (MRI) scans are highly sensitive to acquisition and reconstruction parameters which affect feature stability and model generalizability in radiomic research. This work aims to investigate the effect of image pre-processing and harmonization methods on the stability of brain MRI radiomic features and the prediction performance of radiomic models in patients with brain metastases (BMs).

Two T1 contrast enhanced brain MRI data-sets were used in this study. The first contained 25 BMs patients with scans at two different time points and was used for features stability analysis. The effect of gray level discretization (GLD), intensity normalization (Z-score, Nyul, WhiteStripe, and in house-developed method named N-Peaks), and ComBat harmonization on features stability was investigated and features with intraclass correlation coefficient >0.8 were considered as stable. The second data-set containing 64 BMs patients was used for a classification task to investigate the informativeness of stable features and the effects of harmonization methods on radiomic model performance.

Applying fixed bin number (FBN) GLD, resulted in higher number of stable features compare to fixed bin size (FBS) discretization (10 ± 5.5 % higher). `Harmonization in feature domain improved the stability for non-normalized and normalized images with Z-score and WhiteStripe methods. For the classification task, keeping the stable features resulted in good performance only for normalized images with N-Peaks along with FBS discretization.

To develop a robust MRI based radiomic model we recommend using an intensity normalization method based on a reference tissue (e.g N-Peaks) and then using FBS discretization.

To recognize patients at high risk of refractory disease, the identification of novel prognostic parameters improving stratification of newly diagnosed Hodgkin Lymphoma (HL) is still needed. This study investigates the potential value of metabolic and texture features, extracted from baseline 18F-FDG Positron Emission Tomography/Computed Tomography (PET) and Contrast-Enhanced Computed Tomography scan (CECT), together with clinical data, in predicting first-line therapy refractoriness (R) of classical HL (cHL) with mediastinal bulk involvement. We reviewed 69 cHL patients who underwent staging PET and CECT. Lesion segmentation and texture parameter extraction were performed using the freeware software LIFEx 6.3. The prognostic significance of clinical and imaging features was evaluated in relation to the development of refractory disease. Receiver operating characteristic curve, Cox proportional hazard regression and Kaplan-Meier analyses were performed to examine the potential independent predictors and to evaluate their prognostic value. Among clinical characteristics, only stage according to the German Hodgkin Group (GHSG) classification system significantly differed between R and not-R. Among CECT variables, only parameters derived from second order matrices (gray-level co-occurrence matrix (GLCM) and gray-level run length matrix (GLRLM) demonstrated significant prognostic power. Among PET variables, SUVmean, several variables derived from first (histograms, shape), and second order analyses (GLCM, GLRLM, NGLDM) exhibited significant predictive power. Such variables obtained accuracies greater than 70% at receiver operating characteristic analysis and their PFS curves resulted statistically significant in predicting refractoriness. At multivariate analysis, only HISTO_EntropyPET extracted from PET (HISTO_EntropyPET ) and GHSG stage resulted as significant independent predictors. Their combination identified 4 patient groups with significantly different PFS curves, with worst prognosis in patients with higher HISTO_EntropyPET values, regardless of the stage. Imaging radiomics may provide a reference for prognostic evaluation of patients with mediastinal bulky cHL. The best prognostic value in the prediction of R versus not-R disease was reached by combining HISTO_EntropyPET with GHSG stage.

The domain shift, or acquisition shift in medical imaging, is responsible for potentially harmful differences between development and deployment conditions of medical image analysis techniques. There is a growing need in the community for advanced methods that could mitigate this issue better than conventional approaches. In this paper, we consider configurations in which we can expose a learning-based pixel level adaptor to a large variability of unlabeled images during its training, i.e. sufficient to span the acquisition shift expected during the training or testing of a downstream task model. We leverage the ability of convolutional architectures to efficiently learn domain-agnostic features and train a many-to-one unsupervised mapping between a source collection of heterogeneous images from multiple unknown domains subjected to the acquisition shift and a homogeneous subset of this source set of lower cardinality, potentially constituted of a single image. To this end, we propose a new cycle-free image-to-image architecture based on a combination of three loss functions : a contrastive PatchNCE loss, an adversarial loss and an edge preserving loss allowing for rich domain adaptation to the target image even under strong domain imbalance and low data regimes. Experiments support the interest of the proposed contrastive image adaptation approach for the regularization of downstream deep supervised segmentation and cross-modality synthesis models.

To identify which dynamic contrast-enhanced (DCE-)MRI features best predict histological response to neoadjuvant chemotherapy in patients with an osteosarcoma.

Patients with osteosarcoma who underwent DCE-MRI before and after neoadjuvant chemotherapy prior to resection were retrospectively included at two different centers. Data from the center with the larger cohort (training cohort) was used to identify which method for region-of-interest selection (whole slab or focal area method) and which change in DCE-MRI features (time to enhancement, wash-in rate, maximum relative enhancement and area under the curve) gave the most accurate prediction of histological response. Models were created using logistic regression and cross-validated. The most accurate model was then externally validated using data from the other center (test cohort).

Fifty-five (27 poor response) and 30 (19 poor response) patients were included in training and test cohorts, respectively. Intraclass correlation coefficient of relative DCE-MRI features ranged 0.81-0.97 with the whole slab and 0.57-0.85 with the focal area segmentation method. Poor histological response was best predicted with the whole slab segmentation method using a single feature threshold, relative wash-in rate <2.3. Mean accuracy was 0.85 (95%CI: 0.75-0.95), and area under the receiver operating characteristic curve (AUC-index) was 0.93 (95%CI: 0.86-1.00). In external validation, accuracy and AUC-index were 0.80 and 0.80.

In this study, a relative wash-in rate of <2.3 determined with the whole slab segmentation method predicted histological response to neoadjuvant chemotherapy in osteosarcoma. Consistent performance was observed in an external test cohort.