Radiomic Feature Extraction Research Articles

Automated deep learning approach for identifying etiologies of hypertrophic phenocopies by echocardiography

Abstract Background Left ventricular hypertrophy (LVH) is an indicator of poor prognosis across various cardiovascular diseases and is developed by various etiologies. Transthoracic echocardiography (TTE) is the most available cardiac imaging tool to detect LVH. However, it is still challenging to differentiate three non-ischemic cardiomyopathies which frequently presenting LVH: Fabry cardiomyopathy (FC), hypertrophic cardiomyopathy (HCM), and cardiac amyloidosis (CA) from morphology or cardiac function evaluated by TTE. Purpose We aimed to develop stepwise deep learning models to differentiate FC, HCM, and CA using radiomics extracted from TTE. Methods We conducted a retrospective cohort study between 2011 and 2022 in Taiwan. Patients with left ventricular (LV) mass index≥95g/m² in women or ≥115g/m² in men, or the maximal LV wall thickness≥13mm on echocardiography were enrolled. The TTE automatic processing models included four steps: image quality assessment, view classification, auto-segmentation of interventricular septum (IVS), and dynamic radiomics analysis. One echocardiographic study was scanned per patient, and each study contains multiple cardiac cycles. One sample was extracted from each cycle and all samples were evaluated by image quality assessment and view classification model that based on VGG16. Samples with apical four (A4C) and apical five chamber (A5C) views were retained for following steps. Manual contouring of the septal region served as the ground truth for auto-segmentation based on U-Net. The last step was the extraction of dynamic radiomic features from IVS and the dynamic radiomic features were analyzed using an LSTM model to differentiate three hypertrophic cardiomyopathies. Results There were 103 patients with FC, 96 patients with HCM, and 49 patients with CA in this study. Patients with CA were older and had lower LV ejection fraction (EF) than those with FC or HCM (age: 65.8±9.7 vs. 59.3±11.1 vs. 58.6±14.2 years, P=0.002; LVEF: 56.6±10.8 vs. 61.9±8.8 vs. 64.5±8.3%, P<0.001). In addition, patient with FC had larger LV mass index than those with HCM or CA. (Fig. 2A) The details of the performances of four stepwise models were provided in Figure 2 (B-F). The image quality assessment and view classification models achieved accuracies of 91.7% and 95.2%, respectively. The auto-segmentation had the median Dice score as 0.88 for contouring IVS. The results indicated that deep learning algorithms had great ability to recognize different views with proper image quality, and could contour the IVS adequately. The accuracy of dynamic radiomics-based LSTM model to differentiate three hypertrophic phenocopies was 74.5%. Conclusion We developed an automatic process with four stepwise models for analyzing the radiomics of IVS on TTE to differentiate FC, HCM and CA with the modest performance by LSTM model. The developed automatic models may provide potential implications to identify various etiologies LVH in the future.Figure 1-Workflow of this studyFigure 2-Performances of stepwise models

ESR Essentials: radiomics-practice recommendations by the European Society of Medical Imaging Informatics.

Radiomics is a method to extract detailed information from diagnostic images that cannot be perceived by the naked eye. Although radiomics research carries great potential to improve clinical decision-making, its inherent methodological complexities make it difficult to comprehend every step of the analysis, often causing reproducibility and generalizability issues that hinder clinical adoption. Critical steps in the radiomics analysis and model development pipeline-such as image, application of image filters, and selection of feature extraction parameters-can greatly affect the values of radiomic features. Moreover, common errors in data partitioning, model comparison, fine-tuning, assessment, and calibration can reduce reproducibility and impede clinical translation. Clinical adoption of radiomics also requires a deep understanding of model explainability and the development of intuitive interpretations of radiomic features. To address these challenges, it is essential for radiomics model developers and clinicians to be well-versed in current best practices. Proper knowledge and application of these practices is crucial for accurate radiomics feature extraction, robust model development, and thorough assessment, ultimately increasing reproducibility, generalizability, and the likelihood of successful clinical translation. In this article, we have provided researchers with our recommendations along with practical examples to facilitate good research practices in radiomics. KEY POINTS: Radiomics' inherent methodological complexity should be understood to ensure rigorous radiomic model development to improve clinical decision-making. Adherence to radiomics-specific checklists and quality assessment tools ensures methodological rigor. Use of standardized radiomics tools and best practices enhances clinical translation of radiomics models.

Synergizing traditional CT imaging with radiomics: a novel model for preoperative diagnosis of gastric neuroendocrine and mixed adenoneuroendocrine carcinoma

ObjectiveTo develop diagnostic models for differentiating gastric neuroendocrine carcinoma (g-NEC) and gastric mixed adeno-neuroendocrine carcinoma (g-MANEC) from gastric adenocarcinoma (g-ADC) based on traditional contrast enhanced CT imaging features and radiomics features.MethodsWe retrospectively analyzed 90 g-(MA)NEC (g-MANEC and g-NEC) patients matched 1:1 by T-stage with 90 g-ADC patients. Traditional CT features were analyzed using univariable and multivariable logistic regression. Tumor segmentation and radiomics features extraction were performed with Slicer and PyRadiomics. Feature selection was conducted through univariable analysis, correlation analysis, LASSO, and multivariable stepwise logistic. The combined model incorporated clinical and radiomics predictors. Diagnostic performance was assessed with ROC curves and DeLong’s test. The models’ diagnostic efficacy was further validated in subgroup of g-NEC vs. g-ADC and g-MANEC vs. g-ADC cases.ResultsTumor necrosis and lymph node metastasis were independent predictors for differentiating g-(MA)NEC from g-ADC (P < 0.05). The clinical model’s AUC was 0.700 (training) and 0.667(validation). Five radiomics features were retained, with the radiomics model showing AUC of 0.809 (training) and 0.802 (validation). The combined model’s AUCs were 0.853 (training) and 0.812 (validation), significantly outperforming the clinical model (P < 0.05). Subgroup analysis revealed that the combined model exhibited acceptable performance in differentiating g-NEC from g-ADC and g-MANEC from g-ADC, with AUC of 0.887 and 0.823 in the training cohort and 0.852 and 0.762 in the validation cohort.ConclusionA combined model based on traditional CT imaging and radiomic features provides a non-invasive and effective preoperative diagnostic method for differentiating g-(MA)NEC from g-ADC.

P10.22.B PREDICTING INTRAOPERATIVE 5-ALA-INDUCED TUMOR FLUORESCENCE VIA MRI AND DEEP LEARNING IN GLIOMAS WITH RADIOGRAPHIC LOWER-GRADE CHARACTERISTICS

Abstract BACKGROUND Lower-grade gliomas typically exhibit 5-aminolevulinic acid (5-ALA)-induced fluorescence in only 20-30% of cases, a rate that can be increased by doubling the administered dose of 5-ALA. Fluorescence can depict anaplastic foci, which can be precisely resected to avoid undergrading. We aimed to analyze whether a deep learning model can predict intraoperative fluorescence based on preoperative magnetic resonance imaging (MRI). Material and Methods: The MRI images of gliomas lacking high-grade characteristics (necrosis, extended contrast-enhancement, a.o.) consisted of T1, T1-post gadolinium, and FLAIR sequences. The preprocessed MRIs were fed into an encoder-decoder convolutional neural network (U-Net), pre-trained for tumor segmentation using those three MRI sequences. We used the outputs of the bottleneck layer of the U-Net in the Variational Autoencoder (VAE) as features for classification. We identified and utilized the most effective features in a Random Forest classifier using the principal component analysis (PCA) and the partial least square discriminant analysis (PLS-DA) algorithms. We evaluated the performance of the classifier using a 10-fold cross-validation procedure. RESULTS We evaluated a cohort of 163 glioma patients categorized as fluorescent (n=83) or non-fluorescent (n=80). Our proposed approach’s performance was evaluated using metrics such as mean balanced accuracy, mean sensitivity, and mean specificity. The optimal results were obtained by employing top-performing features selected by PCA, resulting in a mean balanced accuracy of 80% and mean sensitivity and specificity of 84% and 76%, respectively. CONCLUSION Our findings highlight the potential of a U-Net model, coupled with a random forest classifier, for intraoperative fluorescence prediction. We achieved good accuracy using advanced techniques such as deep learning-based tumor segmentation and Variational Autoencoder for radiomics feature extraction. While the model can still be improved, it has the potential for evaluating when to administer 5-ALA to gliomas lacking typical high-grade radiographic features.

MATHEMATICAL MODELLING OF SURVIVAL IN LOW GRADE GLIOMAS AT MALIGNANT TRANSFORMATION WITH XGBOOST

Abstract AIMS To develop non-linear machine learning models using the XGBoost algorithm to predict a continuous (overall survival (OS) and a binary survival outcome (OS > 5 years) using clinical, molecular and genetic, and radiomic data. METHOD Patients with LGGs treated at a single institution (2005-2020) with histology and MRIs at the time of malignant transformation (MT) were retrospectively included in this study. MRIs underwent in-house tumour segmentation pipeline with radiomic feature extraction of whole-tumour, enhancing, non-enhancing and oedema components, and masked disconnectome map components. Patients were split into training and testing sets for the development of the survival models, which were assessed with mean absolute error (MAE) and root mean square error (RMSE) for the prediction of OS; and receiver operating characteristics analysis for the prediction of OS > 5 years. RESULTS Of 553 patients, 415 patients were included in the training set and 138 patients in the testing set. The XGB Regressor model was able to predict overall survival (OS) from the time of malignant transformation (tMRI) with an MAE of 953 days (RMSE: 1163 days). The XGB Classifier model was able to predict the probability of OS > 5 years from tMRI with an accuracy of 64% (sensitivity: 58%, specificity: 70%). Age, IDH1 mutation, 1p/19q co- deletion, regularity of tumour shape, and disconnectome-related perilesional components were most predictive of survival outcome. CONCLUSION This study has investigated the predictive capabilities of clinical, molecular and genetic, and radiomic data to develop survival analysis models, using XGBoost, to predict OS and OS > 5 years in patients with LGG at tMRI. We corroborate previous findings that age, 1p/19q co-deletion and IDH1 mutation are positive prognosticators for survival. However, further investigation into the radiomics of the disconnectome, especially of the perilesional oedema compartment, presents an intriguing and novel avenue for survival analysis of patients with LGG.

MODELLING GBM RECURRENCE: THE INflUENCE OF PERI-LESIONAL OEDEMA AND THE DISCONNECTOME

Abstract AIMS Residual tumour burden after surgery in GBM patients is a prognostic imaging biomarker, often involving the area of T2 signal elevation surrounding the tumour core, histopathologically representing infiltrative oedema with tumour cells. We present the first stage of a model to improve the prediction GBM recurrence by specifically investigating the peri-lesional oedema compartment. METHOD UK-wide sample of histologically proven IDHwt GBMs with MRIs at the time of diagnosis underwent state-of- the-art in-house tumour segmentation pipeline and radiomic feature extraction of whole-tumour, enhancing, non-enhancing and oedema components, and masked disconnectome map components. Of these, early GBM recurrence, within 6 months after completion of radiotherapy, was curated and distribution mapped across the brain. RESULTS 1464 patients were identified: male = 901, median age 63 years, average overall survival 597 days. The radiomic features: Compactness 1 and 2, Surface Area, Spherical Disproportion, Sphericity, and Maximum 2D Diameter column of the disconnectome-related perilesional oedema compartment were most predictive of survival outcome (p value <0.01). Of these, 47 patients with average overall survival 357 days and average time from end of radiotherapy to progression of 131 days were identified. CONCLUSION Our early results reveal the relationship between perilesional oedema MRI features and survival, with disconnectome-related perilesional oedema features most predictive. This suggests that extended brain network aberrations caused by gliomas may also influence tumour recurrence. Further work will focus on the development of more robust predictive models of tumour recurrence that have potential for significant impact on surgical and radiotherapy planning.

Radiomics parameters of epicardial adipose tissue predict mortality in acute pulmonary embolism

BackgroundAccurate prediction of short-term mortality in acute pulmonary embolism (APE) is very important. The aim of the present study was to analyze the prognostic role of radiomics values of epicardial adipose tissue (EAT) in APE.MethodsOverall, 508 patients were included into the study, 209 female (42.1%), mean age, 64.7 ± 14.8 years. 4.6%and 12.4% died (7- and 30-day mortality, respectively). For external validation, a cohort of 186 patients was further analysed. 20.2% and 27.7% died (7- and 30-day mortality, respectively). CTPA was performed at admission for every patient before any previous treatment on multi-slice CT scanners. A trained radiologist, blinded to patient outcomes, semiautomatically segmented the EAT on a dedicated workstation using ImageJ software. Extraction of radiomic features was applied using the pyradiomics library. After correction for correlation among features and feature cleansing by random forest and feature ranking, we implemented feature signatures using 247 features of each patient. In total, 26 feature combinations with different feature class combinations were identified. Patients were randomly assigned to a training and a validation cohort with a ratio of 7:3. We characterized two models (30-day and 7-day mortality). The models incorporate a combination of 13 features of seven different image feature classes.FindingsWe fitted the characterized models to a validation cohort (n = 169) in order to test accuracy of our models. We observed an AUC of 0.776 (CI 0.671–0.881) and an AUC of 0.724 (CI 0.628–0.820) for the prediction of 30-day mortality and 7-day mortality, respectively. The overall percentage of correct prediction in this regard was 88% and 79% in the validation cohorts. Lastly, the AUC in an independent external validation cohort was 0.721 (CI 0.633–0.808) and 0.750 (CI 0.657–0.842), respectively.InterpretationRadiomics parameters of EAT are strongly associated with mortality in patients with APE.Clinical trial numberNot applicable.

Rapid patient-specific organ dose estimation in computed tomography scans via integration of radiomics features and neural networks.

Computed tomography (CT) offers detailed cross-sectional images of internal anatomy for disease detection but carries a risk of solid cancer or blood malignancies due to exposure to X-ray radiation. This study aimed to develop a new method to quickly predict patient-specific organ doses from CT examinations by training neural networks (NNs) based on radiomics features. CT Digital Imaging and Communications in Medicine (DICOM) image data were exported to DeepViewer, a clinical autosegmentation software, to segment the regions of interest (ROIs) for patient organs. Radiomics feature extraction was performed based on the selected CT data and ROIs. Reference organ doses were computed using Monte Carlo (MC) simulations. Patient-specific organ doses were predicted by training a NN model based on radiomics features and reference doses. For the dose prediction performance, the relative root mean squared error (RRMSE), mean absolute percentage error (MAPE), and coefficient of determination (R2) were evaluated on the test sets. The robustness of the NN model was evaluated via the random rearrangement of patient samples in the training and test sets. The maximal difference between the reference and predicted doses was less than 1 mGy for all investigated organs. The range of MAPE was 1.68% to 5.2% for head organs, 11.42% to 15.2% for chest organs, and 5.0% to 8.0% for abdominal organs; the maximal R2 values were 0.93, 0.86, and 0.89 for the head, chest, and abdominal organs, respectively. The radiomics feature-based NN model can achieve accurate prediction of patient-specific organ doses at a high speed of less than 1 second using a single central processing unit, which supports its use as a user-friendly online clinical application.

Machine learning-based discrimination of benign and malignant breast lesions on US: The contribution of shear-wave elastography

PurposeTo build and validate a combined radiomics and machine learning (ML) approach using B-mode US and SWE images to differentiate benign from malignant solid breast lesions (BLs) and compare its performance with that of an expert radiologist. MethodsPatients with at least one BI-RADS 2–6 BL who performed breast US integrated with SWE were retrospectively included. B-mode US and SWE images were manually segmented to extract radiomics features. A multi-step feature selection process was performed and a predictive model built using the Logistic Regression algorithm. The diagnostic accuracy was evaluated with the AUC and Matthews Correlation Coefficient (MCC) metrics. The performance of the ML classifier was compared to that of an expert radiologist. Results427 Bls were included and divided into a training (286 BLs, of which 127 benign and 159 malignant) and a test set (141 BLs, of which 59 benign and 82 malignant). Of 1098 features extracted from B-mode US and SWE images, 13 were finally selected. The ML classifier showed an AUC of 0.768 and 0.746, and an MCC of 0.403 and 0.423 in the training and test sets, respectively. The performance was higher than that of the expert radiologist assessing only B-mode US images, but significantly lower when SWE images were also provided. ConclusionA ML approach based on B-mode US and SWE images may represent a potential tool in the characterization of BLs. SWE still gives its most relevant contribution in the clinical setting rather than included in a radiomics pipeline.

Risk stratification and overall survival prediction in extensive stage small cell lung cancer after chemotherapy with immunotherapy based on CT radiomics

The prognosis of extensive-stage small cell lung cancer is usually poor. In this study, a combined model based on pre-treatment CT radiomics and clinical features was constructed to predict the OS of extensive-stage small cell lung cancer after chemotherapy with immunotherapy.Clinical data of 111 patients with extensive stage small-cell lung cancer who received first-line immunotherapy combined with chemotherapy in our hospital from December 2019 to December 2021 were retrospectively collected. Finally, 93 patients were selected for inclusion in the study, and CT images were obtained through PACS system before treatment. All patients were randomly divided into a training set (n = 66) and a validation set (n = 27). Images were imported into ITK-SNAP to outline areas of interest, and Python software was used to extract radiomics features. A total of 1781 radiomics features were extracted from each patient’s images. The feature dimensions were reduced by MRMR and LASSO methods, and the radiomics features with the greatest predictive value were screened. The weight coefficient of radiomics features was calculated, and the linear combination of the feature parameters and the weight coefficient was used to calculate Radscore. Univariate cox regression analysis was used to screen out the factors significantly associated with prognosis from the radiomics and clinical features, and multivariate cox regression analysis was performed to establish the prognosis prediction model of extensive stage small cell lung cancer. The degree of metastases was selected as a significant clinical prognostic factor by univariate cox regression analysis. Seven radiomics features with significance were selected by LASSO-COX regression analysis, and the Radscore was calculated according to the coefficient of the radiomics features. An alignment diagram survival prediction model was constructed by combining Radscore with the number of metastatic lesions. The study population was stratified into those who survived less than 11 months, and those with a greater than 11 month survival. The C-index was 0.722 (se = 0.044) and 0.68(se = 0.074) in the training and the validation sets, respectively. The Log_rank test results of the combination model were as follows: training set: p < 0.0001, validation set: p = 0.00042. In this study, a combined model based on radiomics and clinical features could predict OS in patients with extensive stage small cell lung cancer after chemotherapy with immunotherapy, which could help guide clinical treatment strategies.

Computed tomography-based multi-organ radiomics nomogram model for predicting the risk of esophagogastric variceal bleeding in cirrhosis.

Radiomics has been used in the diagnosis of cirrhosis and prediction of its associated complications. However, most current studies predict the risk of esophageal variceal bleeding (EVB) based on image features at a single level, which results in incomplete data. Few studies have explored the use of global multi-organ radiomics for non-invasive prediction of EVB secondary to cirrhosis. To develop a model based on clinical and multi-organ radiomic features to predict the risk of first-instance secondary EVB in patients with cirrhosis. In this study, 208 patients with cirrhosis were retrospectively evaluated and randomly split into training (n = 145) and validation (n = 63) cohorts. Three areas were chosen as regions of interest for extraction of multi-organ radiomic features: The whole liver, whole spleen, and lower esophagus-gastric fundus region. In the training cohort, radiomic score (Rad-score) was created by screening radiomic features using the inter-observer and intra-observer correlation coefficients and the least absolute shrinkage and selection operator method. Independent clinical risk factors were selected using multivariate logistic regression analyses. The radiomic features and clinical risk variables were combined to create a new radiomics-clinical model (RC model). The established models were validated using the validation cohort. The RC model yielded the best predictive performance and accurately predicted the EVB risk of patients with cirrhosis. Ascites, portal vein thrombosis, and plasma prothrombin time were identified as independent clinical risk factors. The area under the receiver operating characteristic curve (AUC) values for the RC model, Rad-score (liver + spleen + esophagus), Rad-score (liver), Rad-score (spleen), Rad-score (esophagus), and clinical model in the training cohort were 0.951, 0.930, 0.801, 0.831, 0.864, and 0.727, respectively. The corresponding AUC values in the validation cohort were 0.930, 0.886, 0.763, 0.792, 0.857, and 0.692. In patients with cirrhosis, combined multi-organ radiomics and clinical model can be used to non-invasively predict the probability of the first secondary EVB.

Radiomics prediction models of left atrial appendage hypercoagulability based on machine learning algorithms: an exploration about cardiac computed tomography angiography imaging.

Transesophageal echocardiography (TEE) is the standard method for diagnosing left atrial appendage (LAA) hypercoagulability in patients with atrial fibrillation (AF), which means LAA thrombus/sludge, dense spontaneous echo contrast and slow LAA blood flow velocity (< 0.25m/s). Based on machine learning algorithms, cardiac computed tomography angiography (CCTA) radiomics features were adopted to construct prediction models and explore a suitable approach for diagnosing LAA hypercoagulability and adjusting anticoagulation. This study included 652 patients with non-valvular AF. The univariate analysis were used to select meaningful clinical characteristics to predict LAA hypercoagulability. Then 3D Slicer software was adopted to extract radiomics features from CCTA imaging. The radiomics score was calculated using the least absolute shrinkage and selection operator logistic regression analysis to predict LAA hypercoagulability. We then combined clinical characteristics and radiomics scores to construct a nomogram model. Finally, we got prediction models based on machine learning algorithms and logistic regression separately. The area under the receiver operating characteristic curve of radiomics score was 0.8449 in the training set and 0.7998 in the validation set. The nomogram model had a concordance index of 0.838. The final machine-learning based prediction models had good performances (best f1 score = 0.85). Radiomics features of long maximum diameter and high uniformity of Hounsfield unit in left atrial were significant predictors of the hypercoagulable state in LAA, with better predictive efficacy than clinical characteristics. Our combined models based on machine learning were reliable for hypercoagulable state screening and anticoagulation adjustment.

Prediction of intraoperative press-fit stability of the acetabular cup in total hip arthroplasty using radiomics-based machine learning models

BackgroundPreoperative prediction of the acetabular cup press-fit stability in total hip arthroplasty is necessary for clinical decision-making. This study aims to establish and validate machine learning models to investigate the feasibility of predicting the intraoperative press-fit stability of the acetabular cup in total hip arthroplasty (THA). Methods226 patients who underwent primary THA from 2018 to 2022 in our hospital were retrospectively enrolled. Patients were divided into press-fit stable or unstable groups according to the intraoperative pull-out test of the implanted cup. Then, they were randomly assigned to the training or test cohort in an 8:2 ratio. We used 3Dslicer software to segment the region of interest (ROI) of the patient’s bilateral hip X-ray to extract radiomics features. The least absolute shrinkage and selection operator (LASSO) regression was used in our feature selection. Finally, four machine learning models were employed in this study, including support vector machine (SVM), random forest (RF), logistic regression (LR), and XGBoost (XGB). Decision curve analysis (DCA), and receiver operating characteristic (ROC) curves of the models were plotted. The area under the curve (AUC), diagnostic accuracy, sensitivity, and specificity were calculated as well. The AUCs of the four models were compared using the DeLong test. ResultsTwenty-seven valuable radiomics features were determined by dimensionality reduction and selection. Regarding to the DeLong test, the AUC of the XGB model was significantly different from those of the other three models. (p < 0.05). Among all models, the XGB model exhibited the best performance with an AUC of 0.823 (95 % CI: 0.711–0.919) in the test cohort and showed optimal clinical efficacy according to the DCA. ConclusionMachine learning models based on X-ray radiomics can accurately predict the intraoperative press-fit stability of implanted cups preoperatively, providing surgeons with valuable information to lower the complication risk in THA.

Magnetic resonance imaging-based radiomics for predicting infiltration levels of CD68+ tumor-associated macrophages in glioblastomas.

Tumor-associated macrophages (TAMs) are important biomarkers of tumor invasion and prognosis in patients with glioblastoma. We combined the imaging and radiomics features of preoperative MRI to predict CD68+ macrophage infiltration. Clinical, MRI image, and pathology data of 188 patients with glioblastoma were analyzed. Overall, 143 patients were included in the training (n = 101) and validation (n = 42) sets, whereas 45patients were included in an independent test set. The optimal cut-off value (14.8%) was based on the minimum p-value formed by the Kaplan-Meier survival analysis and log-rank tests which divided patients into groups with high CD68+ TAMs(≥ 14.8%) and low CD68+ TAMs (< 14.8%). Regions of interest and radiomics features extraction were based on contrast-enhanced T1-weighted images (CE-T1WI) and T2WI. Multi-parameter stepwise regression was used to create the clinical, radiomics, and combined models, each evaluated using the receiver operating characteristic curve. Decision curve analysis was used to assess the clinical applicability of the nomogram. Aclinical model based on the minimum apparent diffusion coefficient (ADCmin) revealed an area under the curve (AUC) of 0.768, 0.764, and 0.624 for the training set, validation set, and test set, respectively. The 2D radiomics model, based on two features, revealed an AUC of 0.783, 0.724, and 0.789 for the training, validation, and test sets, respectively. The 3D radiomics model, based on three features, revealed AUCs of 0.823, 0.811, and 0.787 for the training, validation, and test sets, respectively. The combined model, with ADCmin and radiomics features, showed the best performance, with AUCs of 0.865, 0.822, and 0.776 for the training, validation, and test sets, respectively. The calibration curve of the combined model nomogram showed good agreement between the estimated and actual probabilities. The combined model constructed using ADCmin, aquantitative imaging parameter, combined with five key radiomics features can be used to evaluate the extent of CD68+ macrophages before surgery.

Role of Artificial intelligence model in prediction of low back pain using T2 weighted MRI of Lumbar spine

Background Low back pain (LBP), the primary cause of disability, is the most common musculoskeletal disorder globally and the primary cause of disability. Magnetic resonance imaging (MRI) studies are inconclusive and less sensitive for identifying and classifying patients with LBP. Hence, this study aimed to investigate the role of artificial intelligence (AI) models in the prediction of LBP using T2 weighted MRI image of the lumbar spine. Methods This was a prospective case-control study. A total of 200 MRI patients (100 cases and controls each) referred for lumbar spine and whole spine screening were included. The scans were performed using 3.0 Tesla MRI (United Imaging Healthcare). T2 weighted images of the lumbar spine were segmented to extract radiomic features. Machine learning (ML) models, such as random forest, decision tree, logistic regression, K-nearest neighbors, adaboost, and deep learning methods (DL), such as ResNet and GoogleNet, were used, and performance measures were calculated. Results Our study showed that Random forest and AdaBoost are the most reliable ML models for predicting LBP. Random forest showed high performance with area under curve (AUC) values from 0.83 to 0.88 across all lumbar vertebrae and L2-L3, L3-L4, and L4-L5 intervertebral discs (IVDs), with AUCs of 0.88 the highest at L5-S1 IVD (0.92). Adaboost demonstrated high performance at the L2-L5 vertebrae with AUC values of 0.82 to 0.90, with the highest AUC (0.97) at the L5-S1 IVD. Among the DL models, GoogleNet outperformed the other models at 30 epochs with an accuracy of 0.85, followed by ResNet 18 (30 epochs) with an accuracy of 0.84. Conclusion The study demonstrated that ML and DL models can effectively predict LBP from MRI T2 weighted image of the lumbar spine. ML and DL models could also enhance the diagnostic accuracy of LBP, potentially leading to better patient management and outcomes.

Deep Learning Features Can Improve Radiomics-Based Prostate Cancer Aggressiveness Prediction

PURPOSE Emerging evidence suggests that the use of artificial intelligence can assist in the timely detection and optimization of therapeutic approach in patients with prostate cancer. The conventional perspective on radiomics encompassing segmentation and the extraction of radiomic features considers it as an independent and sequential process. However, it is not necessary to adhere to this viewpoint. In this study, we show that besides generating masks from which radiomic features can be extracted, prostate segmentation and reconstruction models provide valuable information in their feature space, which can improve the quality of radiomic signatures models for disease aggressiveness classification. MATERIALS AND METHODS We perform 2,244 experiments with deep learning features extracted from 13 different models trained using different anatomic zones and characterize how modeling decisions, such as deep feature aggregation and dimensionality reduction, affect performance. RESULTS While models using deep features from full gland and radiomic features consistently lead to improved disease aggressiveness prediction performance, others are detrimental. Our results suggest that the use of deep features can be beneficial, but an appropriate and comprehensive assessment is necessary to ensure that their inclusion does not harm predictive performance. CONCLUSION The study findings reveal that incorporating deep features derived from autoencoder models trained to reconstruct the full prostate gland (both zonal models show worse performance than radiomics only models), combined with radiomic features, often lead to a statistically significant increase in model performance for disease aggressiveness classification. Additionally, the results also demonstrate that the choice of feature selection is key to achieving good performance, with principal component analysis (PCA) and PCA + relief being the best approaches and that there is no clear difference between the three proposed latent representation extraction techniques.

A Comparative Study on the Diagnostic Performance of Multiple Radiomics Models in Differentiating PGL and MIA in Pulmonary Ground-Glass Nodules

Rationale and ObjectivesHistological subtypes of lung cancers are critical for clinical treatment decision. The aim of this study is to compare the diagnostic performance of multiple radiomics models in differentiating PGL and MIA in pulmonary GGN, in order to identify the most optimal diagnostic model. Materials and methodsPatients presenting with GGNs on lung CT, confirmed as PGL or MIA through surgical pathology between October 2015 and June 2023, were included. The GGNs were randomly divided into training and validation sets at a 7:3 ratio. Clinical imaging characteristics were analyzed by univariate and multivariate logistic regression to identify independent risk factors for predicting MIA, leading to the development of a clinical model. ITK-SNAP and Pyradiomics were employed for segmentation and radiomics feature extraction. Subsequently, radiomics and combined models were established. The diagnostic performance of the three models was compared using ROC curves and quantitatively assessed by AUC, accuracy, specificity, and sensitivity. ResultsA total of 116 cases of GGNs with pathologically confirmed PGLs and MIAs were included. The clinical model identified three independent predictors. The radiomics model identified seven distinct radiomic features. A combined model was constructed by integrating clinical imaging features with radiomic features. In the training set, the combined model demonstrated a higher AUC than the radiomics model, with AUCs of 0.87 and 0.85 respectively. In the validation set, the radiomics model outperformed the combined model with an AUC of 0.83 versus 0.82. Notably, the radiomics model achieved the highest accuracy and specificity, while the combined model demonstrated the highest sensitivity. However, both models performed significantly better than the clinical model. ConclusionThe independent radiomics model can serve as a rapid, non-invasive diagnostic tool for differentiating between the PGL and MIA.

Inteligencia Artificial, Imagen Médica y Medicina de Precisión: Avances y Perspectivas

The role of artificial intelligence (AI) applied to medical imaging in developing and strengthening personalized medicine is described as a continuous process of improvement, a set of opportunities, and a professional challenge of enormous significance. This paper outlines the main processes in which AI is involved regarding imaging, including data preparation, image harmonization, automatic segmentation of organs and lesions, labeling, extraction of radiomic features, and the development of predictive clinical models. It will also address aspects related to the integration of these solutions into clinical practice to enhance accuracy and efficiency in the care process, diagnosis, and treatment, making it more personalized, efficient, and precise. Projects like PRIMAGE and CHAIMELEON highlight the transformative potential of AI and the fundamental role of interdisciplinary collaboration in realizing this potential, based on ongoing multiprofesional collaboration to address the ethical, regulatory, technical, and clinical challenges that accompany these advancements.

Diagnostic performance of MRI-based radiomics models using machine learning approaches for the triple classification of parotid tumors

Rationale and objectivesPreoperative differentiation of malignant tumors (MT), pleomorphic adenomas (PA), and other benign tumors of the parotid gland is critical to clinical strategy, this study aimed to develop and validate a T2-weighted image (T2WI) based radiomics model through machine learning approaches for the triple classification of parotid gland tumors. Materials and methodsWe retrospectively enrolled 147 patients from January 2010 to July 2022. T2WIs were used to extract radiomics features. Max-Relevance and Min-Redundancy (mRMR) and Extreme Gradient Boosting (XGBoost) algorithms were used to select features. Using a 5-fold cross-validation strategy, radiomics models were constructed using a Support Vector Machine (SVM), Logistic Regression (LR), and k-Nearest Neighbor (KNN) for the triple classification of parotid tumors. The three models were evaluated and compared using the receiver operator characteristic (ROC) curve, sensitivity, specificity, and accuracy. ResultsA total of 1057 radiomics features were extracted, and 8 features were selected to developed the radiomics model, including First-order Median, First-order Skewness, First-order Minimum, Original_shape_Flatness, Glcm Inverse Variance, Glcm Inverse Variance, Glszm Low Gray Level Zone Emphasis, and Glszm Small Area Low Gray Level Emphasis. The mean area under the curves (AUCs) for the radiomics models in training and validation sets through LR, SVM and KNN were 0.85 and 0.80, 0.85 and 0.80 and 0.83 and 0.79, respectively. ConclusionThe T2WI-based radiomics models through LR, SVM and KNN demonstrated good performance in the triple classification of parotid tumors.

Machine learning model for non-alcoholic steatohepatitis diagnosis based on ultrasound radiomics

BackgroundNon-Alcoholic Steatohepatitis (NASH) is a crucial stage in the progression of Non-Alcoholic Fatty Liver Disease(NAFLD). The purpose of this study is to explore the clinical value of ultrasound features and radiological analysis in predicting the diagnosis of Non-Alcoholic Steatohepatitis.MethodAn SD rat model of hepatic steatosis was established through a high-fat diet and subcutaneous injection of CCl4. Liver ultrasound images and elastography were acquired, along with serum data and histopathological results of rat livers.The Pyradiomics software was used to extract radiomic features from 2D ultrasound images of rat livers. The rats were then randomly divided into a training set and a validation set, and feature selection was performed through dimensionality reduction. Various machine learning (ML) algorithms were employed to build clinical diagnostic models, radiomic models, and combined diagnostic models. The efficiency of each diagnostic model for diagnosing NASH was evaluated using Receiver Operating Characteristic (ROC) curves, Clinical Decision Curve Analysis (DCA), and calibration curves.ResultsIn the machine learning radiomic model for predicting the diagnosis of NASH, the Area Under the Curve (AUC) of ROC curve for the clinical radiomic model in the training set and validation set were 0.989 and 0.885, respectively. The Decision Curve Analysis revealed that the clinical radiomic model had the highest net benefit within the probability threshold range of > 65%. The calibration curve in the validation set demonstrated that the clinical combined radiomic model is the optimal method for diagnosing Non-Alcoholic Steatohepatitis.ConclusionThe combined diagnostic model constructed using machine learning algorithms based on ultrasound image radiomics has a high clinical predictive performance in diagnosing Non-Alcoholic Steatohepatitis.