Machine Learning-based Radiomics Research Articles

Diagnostic accuracy of a machine learning-based radiomics approach of MR in predicting IDH mutations in glioma patients: a systematic review and meta-analysis.

To assess the diagnostic accuracy of machine learning (ML)-based radiomics for predicting isocitrate dehydrogenase (IDH) mutations in patients with glioma. A systematic search of PubMed, Web of Science, Embase, and the Cochrane Library from inception to 1 September 2023, was conducted to collect all articles investigating the diagnostic performance of ML for the prediction of IDH mutations in gliomas. Two reviewers independently screened all papers for eligibility. Methodological quality and risk of bias were assessed using the METhodological RadiomICs Score and Quality Assessment of Diagnostic Accuracy Studies-2, respectively. The pooled sensitivity, specificity, and 95% confidence intervals were calculated, and the area under the receiver operating characteristic curve (AUC) was obtained. In total, 14 original articles assessing 1740 patients with gliomas were included. The AUC of ML for predicting IDH mutation was 0.90 (0.87-0.92). The pooled sensitivity, specificity, and diagnostic odds ratio were 0.83 (0.71-0.90), 0.84 (0.74-0.90), and 25 (12,50) respectively. In subgroup analyses, modeling methods, glioma grade, and the combination of magnetic resonance imaging and clinical features affected the diagnostic performance in predicting IDH mutations in gliomas. ML-based radiomics demonstrated excellent diagnostic performance in predicting IDH mutations in gliomas. Factors influencing the diagnosis included the modeling methods employed, glioma grade, and whether the model incorporated clinical features. https://www.crd.york.ac.uk/PROSPERO/#myprospero, PROSPERO registry (CRD 42023395444).

Radiomics and machine learning analysis of liver magnetic resonance imaging for prediction and early detection of tumor response in colorectal liver metastases.

The aim of this study was to demonstrate the effectiveness of a machine learning-based radiomics model for distinguishing tumor response and overall survival in patients with unresectable colorectal liver metastases (CRLM) treated with targeted biological therapy. We prospectively recruited 17 patients with unresectable liver metastases of colorectal cancer, who had been given targeted biological therapy as the first line of treatment. All patients underwent liver magnetic resonance imaging (MRI) three times up until 8 weeks after chemotherapy. We evaluated the diagnostic performance of machine learning-based radiomics model in tumor response of liver MRI compared with the guidelines for the Response Evaluation Criteria in Solid Tumors. Overall survival was evaluated using the Kaplan-Meier analysis and compared to the Cox proportional hazard ratios following univariate and multivariate analyses. Performance measurement of the trained model through metrics showed the accuracy of the machine learning model to be 76.5%, and the area under the receiver operating characteristic curve was 0.857 (95% confidence interval [CI], 0.605-0.976; P < 0.001). For the patients classified as non-progressing or progressing by the radiomics model, the median overall survival was 17.5 months (95% CI, 12.8-22.2), and 14.8 months (95% CI, 14.2-15.4), respectively (P = 0.431, log-rank test). Machine learning-based radiomics models could have the potential to predict tumor response in patients with unresectable CRLM treated with biologic therapy.

Is it possible to detect cribriform adverse pathology in prostate cancer with magnetic resonance imaging machine learning-based radiomics?

Rationale and objectives: Cribriform patterns are accepted as aggressive variants of prostate cancer. These adverse pathologies are closely associated with early biochemical recurrence, metastasis, castration resistance, and poor disease-related survival. A few publications exist to diagnose these two adverse pathologies with multiparametric magnetic resonance imaging (mpMRI). Most of these publications are retrospective and are not studies that have made a difference in diagnosing adverse pathology. It is also known that fusion biopsies taken from lesions detected in mpMRI are insufficient to detect these adverse pathologies. Our study aims to diagnose this adverse pathology using machine learning-based radiomics data from MR images. Materials and methods: A total of 88 patients who had pathology results indicating the presence of cribriform pattern and prostate adenocarcinoma underwent preoperative MRI examinations and radical prostatectomy. Manual slice-by-slice 3D volumetric segmentation was performed on all axial images. Data processing and machine learning analysis were conducted using Python 3.9.12 (Jupyter Notebook, Pycaret Library). Results: Two radiologists, SE and MAG, with 7 and 8 years of post-graduate experience, respectively, evaluated the images using the 3D-Slicer software without knowledge of the histopathological findings. One hundred seventeen radiomic tissue features were extracted from T1 weighted (T1W) and apparent diffusion coefficient (ADC) sequences for each patient. The interobserver agreement for these features was analyzed using the intraclass correlation coefficient (ICC). Features with excellent interobserver agreement (ICC &gt; 0.90) were further analyzed for collinearity between predictors using Pearson’s correlation. Variables showing a very high correlation (r ≥ ±0.80) were disregarded. The selected features for T1W and ADC images were First-order maximum, First-order skewness, First-order 10th percentile for ADC, and Gray level size zone matrix, Large area low gray level emphasis for T1W.As a result of the classification of PyCaret, the three best models were found. A single model was obtained by blending these three models. AUC, accuracy, recall, precision, and F1 scores were 0.79, 0.77, 0.85, 0.82, and 0.83, respectively. Conclusion: ML-based MRI radiomics of prostate cancer can predict the cribriform pattern. This prognostic factor cannot be determined through qualitative radiological evaluation and may be overlooked in preoperative histopathological specimens.

The current landscape of machine learning-based radiomics in arteriovenous malformations: a systematic review and radiomics quality score assessment.

Arteriovenous malformations (AVMs) are rare vascular anomalies involving a disorganization of arteries and veins with no intervening capillaries. In the past 10 years, radiomics and machine learning (ML) models became increasingly popular for analyzing diagnostic medical images. The goal of this review was to provide a comprehensive summary of current radiomic models being employed for the diagnostic, therapeutic, prognostic, and predictive outcomes in AVM management. A systematic literature review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines, in which the PubMed and Embase databases were searched using the following terms: (cerebral OR brain OR intracranial OR central nervous system OR spine OR spinal) AND (AVM OR arteriovenous malformation OR arteriovenous malformations) AND (radiomics OR radiogenomics OR machine learning OR artificial intelligence OR deep learning OR computer-aided detection OR computer-aided prediction OR computer-aided treatment decision). A radiomics quality score (RQS) was calculated for all included studies. Thirteen studies were included, which were all retrospective in nature. Three studies (23%) dealt with AVM diagnosis and grading, 1 study (8%) gauged treatment response, 8 (62%) predicted outcomes, and the last one (8%) addressed prognosis. No radiomics model had undergone external validation. The mean RQS was 15.92 (range: 10-18). We demonstrated that radiomics is currently being studied in different facets of AVM management. While not ready for clinical use, radiomics is a rapidly emerging field expected to play a significant future role in medical imaging. More prospective studies are warranted to determine the role of radiomics in the diagnosis, prediction of comorbidities, and treatment selection in AVM management.

Radiomic Features at Contrast-Enhanced CT Predict Virus-Driven Liver Fibrosis: A Multi-Institutional Study.

Liver fibrosis is a major cause of morbidity and mortality among in patients with chronic hepatitis. Radiomics, particularly of the spleen, may improve diagnostic accuracy and treatment strategies. External validations are necessary to ensure reliability and generalizability. In this retrospective study, we developed 3 radiomics models using contrast-enhanced computed tomography scans from 167 patients with liver fibrosis (training group) between January 2020 and December 2021. Radiomic features were extracted from arterial venous, portal venous, and equilibrium phase images. Recursive feature selection random forest and the least absolute shrinkage and selection operator logistic regression were used for feature selection and dimensionality reduction. Performance was assessed by area under the curve, C-index, calibration plots, and decision curve analysis. External validation was performed on 114 patients from 2 institutions. Twenty-five radiomic features were significantly associated with fibrosis stage, with 80% of the top 10 features originating from portal venous phase spleen images. The radiomics models showed good performance in the validation cohort (C-indices 0.723-0.808) and excellent calibration. Decision curve analysis indicated clinical benefits, with machine learning-based radiomics models (Random Forest score and support vector machine based radiomics score) providing more significant advantages. Radiomic features offer significant benefits over existing serum indices for staging virus-driven liver fibrosis, underscoring the value of radiomics in enhancing diagnostic accuracy. Specifically, radiomics analysis of the spleen presents additional noninvasive options for assessing fibrosis, highlighting its potential in improving patient management and outcomes.

Machine learning-based radiomics for predicting outcomes in cervical cancer patients undergoing concurrent chemoradiotherapy

PurposesTo investigate the value of machine learning-based radiomics for predicting disease-free survival (DFS) and overall survival (OS) undergoing concurrent chemoradiotherapy (CCRT) for patients with locally advanced cervical cancer (LACC). Materials and methodsIn this multicentre study, 700 patients with IB2-IVA cervical cancer who underwent CCRT with ongoing follow-up were retrospectively analyzed. Three-dimensional radiomics features of primary lesions and its surrounding 5 mm region in T2WI sequences were collected. Six machine learning methods were used to construct the optimal radiomics model for accurate prediction of DFS and OS after CCRT in LACC patients. Eventually, TCGA and GEO databases were used to explore the mechanisms of radiomics in predicting the progression and survival of cervical cancer. This study adhered CLEAR for reporting and its quality was assessed using RQS and METRICS. ResultsIn the prediction of DFS, the RSF model combined tumor and peritumor radiomics demonstrated the best predictive efficacy, with the AUC for predicting 1-year, 3-year, and 5-year DFS in the training, validation, and test sets of 0.986, 0.989, 0.990, and 0.884, 0.838, 0.823, and 0.829, 0.809, 0.841, respectively. In the prediction of OS, the GBM model best performer, with AUC of 0.999, 0.995, 0.978, and 0.981, 0.975, 0.837, and 0.904, 0.860, 0.905. Differential genes in TCGA and GEO suggest that the prediction of radiomics model may be associated with KDELR2 and HK2. ConclusionMachine learning-based radiomics models help to predict DFS and OS after CCRT in LACC patients, and the combination of tumor and peritumor information has higher predictive efficacy, which can provide a reliable basis for therapeutic decision-making in cervical cancer patients.

Preoperative prediction for early recurrence of hepatocellular carcinoma using machine learning-based radiomics.

To develop a contrast-enhanced computed tomography (CECT) based radiomics model using machine learning method and assess its ability of preoperative prediction for the early recurrence of hepatocellular carcinoma (HCC). A total of 297 patients confirmed with HCC were assigned to the training dataset and test dataset based on the 8:2 ratio, and the follow-up period of the patients was from May 2012 to July 2017. The lesion sites were manually segmented using ITK-SNAP, and the pyradiomics platform was applied to extract radiomic features. We established the machine learning model to predict the early recurrence of HCC. The accuracy, AUC, standard deviation, specificity, and sensitivity were applied to evaluate the model performance. 1,688 features were extracted from the arterial phase and venous phase images, respectively. When arterial phase and venous phase images were employed correlated with clinical factors to train a prediction model, it achieved the best performance (AUC with 95% CI 0.8300(0.7560-0.9040), sensitivity 89.45%, specificity 79.07%, accuracy 82.67%, p value 0.0064). The CECT-based radiomics may be helpful to non-invasively reveal the potential connection between CECT images and early recurrence of HCC. The combination of radiomics and clinical factors could boost model performance.

Machine learning-based radiomics score improves prognostic prediction accuracy of stage II/III gastric cancer: A multi-cohort study.

Although accurately evaluating the overall survival (OS) of gastric cancer patients remains difficult, radiomics is considered an important option for studying prognosis. To develop a robust and unbiased biomarker for predicting OS using machine learning and computed tomography (CT) image radiomics. This study included 181 stage II/III gastric cancer patients, 141 from Lichuan People's Hospital, and 40 from the Cancer Imaging Archive (TCIA). Primary tumors in the preoperative unenhanced CT images were outlined as regions of interest (ROI), and approximately 1700 radiomics features were extracted from each ROI. The skeletal muscle index (SMI) and skeletal muscle density (SMD) were measured using CT images from the lower margin of the third lumbar vertebra. Using the least absolute shrinkage and selection operator regression with 5-fold cross-validation, 36 radiomics features were identified as important predictors, and the OS-associated CT image radiomics score (OACRS) was calculated for each patient using these important predictors. Patients with a high OACRS had a poorer prognosis than those with a low OACRS score (P < 0.05) and those in the TCIA cohort. Univariate and multivariate analyses revealed that OACRS was a risk factor [RR = 3.023 (1.896-4.365), P < 0.001] independent of SMI, SMD, and pathological features. Moreover, OACRS outperformed SMI and SMD and could improve OS prediction (P < 0.05). A novel biomarker based on machine learning and radiomics was developed that exhibited exceptional OS discrimination potential.

Machine learning-based radiomics analysis in predicting RAS mutational status using magnetic resonance imaging.

To assess the efficacy of radiomics features, obtained by magnetic resonance imaging (MRI) with hepatospecific contrast agent, in pre-surgical setting, to predict RAS mutational status in liver metastases. Patients with MRI in pre-surgical setting were enrolled in a retrospective study. Manual segmentation was made by means 3D Slicer image computing, and 851 radiomics features were extracted as median values using the PyRadiomics Python package. The features were extracted considering the agreement with the Imaging Biomarker Standardization Initiative (IBSI). Balancing was performed through synthesis of samples for the underrepresented classes using the self-adaptive synthetic oversampling (SASYNO) approach. Inter- and intraclass correlation coefficients (ICC) were calculated to assess the between-observer and within-observer reproducibility of all radiomics characteristics. For continuous variables, nonparametric Wilcoxon-Mann-Whitney test was utilized. Benjamini and Hochberg's false discovery rate (FDR) adjustment for multiple testing was used. Receiver operating characteristics (ROC) analysis with the calculation of area under the ROC curve (AUC), sensitivity (SENS), specificity (SPEC), positive predictive value (PPV), negative predictive value (NPV) and accuracy (ACC) were assessed for each parameter. Linear and non-logistic regression model (LRM and NLRM) and different machine learning-based classifiers including decision tree (DT), k-nearest neighbor (KNN) and support vector machine (SVM) were considered. Moreover, features selection were performed before and after a normalized procedure using two different methods (3-sigma and z-score). McNemar test was used to assess differences statistically significant between dichotomic tables. All statistical procedures were done using MATLAB R2021b Statistics and Machine Toolbox (MathWorks, Natick, MA, USA). Seven normalized radiomics features, extracted from arterial phase, 11 normalized radiomics features, from portal phase, 12 normalized radiomics features from hepatobiliary phase and 12 normalized features from T2-W SPACE sequence were robust predictors of RAS mutational status. The multivariate analysis increased significantly the accuracy in RAS prediction when a LRM was used, combining 12 robust normalized features extracted by VIBE hepatobiliary phase reaching an accuracy of 99%, a sensitivity 97%, a specificity of 100%, a PPV of 100% and a NPV of 98%. No statistically significant increase was obtained, considering the tested classifiers DT, KNN and SVM, both without normalization and with normalization methods. Normalized approach in MRI radiomics analysis allows to predict RAS mutational status.

Preoperative Prediction of Perineural Invasion and Prognosis in Gastric Cancer Based on Machine Learning through a Radiomics–Clinicopathological Nomogram

Purpose: The aim of this study was to construct and validate a nomogram for preoperatively predicting perineural invasion (PNI) in gastric cancer based on machine learning, and to investigate the impact of PNI on the overall survival (OS) of gastric cancer patients. Methods: Data were collected from 162 gastric patients and analyzed retrospectively, and radiomics features were extracted from contrast-enhanced computed tomography (CECT) scans. A group of 42 patients from the Cancer Imaging Archive (TCIA) were selected as the validation set. Univariable and multivariable analyses were used to analyze the risk factors for PNI. The t-test, Max-Relevance and Min-Redundancy (mRMR) and the least absolute shrinkage and selection operator (LASSO) were used to select radiomics features. Radscores were calculated and logistic regression was applied to construct predictive models. A nomogram was developed by combining clinicopathological risk factors and the radscore. The area under the curve (AUC) values of receiver operating characteristic (ROC) curves, calibration curves and clinical decision curves were employed to evaluate the performance of the models. Kaplan–Meier analysis was used to study the impact of PNI on OS. Results: The univariable and multivariable analyses showed that the T stage, N stage and radscore were independent risk factors for PNI (p &lt; 0.05). A nomogram based on the T stage, N stage and radscore was developed. The AUC of the combined model yielded 0.851 in the training set, 0.842 in the testing set and 0.813 in the validation set. The Kaplan–Meier analysis showed a statistically significant difference in OS between the PNI group and the non-PNI group (p &lt; 0.05). Conclusions: A machine learning-based radiomics–clinicopathological model could effectively predict PNI in gastric cancer preoperatively through a non-invasive approach, and gastric cancer patients with PNI had relatively poor prognoses.

Development of a machine learning-based radiomics signature for estimating breast cancer TME phenotypes and predicting anti-PD-1/PD-L1 immunotherapy response

BackgroundsSince breast cancer patients respond diversely to immunotherapy, there is an urgent need to explore novel biomarkers to precisely predict clinical responses and enhance therapeutic efficacy. The purpose of our present research was to construct and independently validate a biomarker of tumor microenvironment (TME) phenotypes via a machine learning-based radiomics way. The interrelationship between the biomarker, TME phenotypes and recipients’ clinical response was also revealed.MethodsIn this retrospective multi-cohort investigation, five separate cohorts of breast cancer patients were recruited to measure breast cancer TME phenotypes via a radiomics signature, which was constructed and validated by integrating RNA-seq data with DCE-MRI images for predicting immunotherapy response. Initially, we constructed TME phenotypes using RNA-seq of 1089 breast cancer patients in the TCGA database. Then, parallel DCE-MRI images and RNA-seq of 94 breast cancer patients obtained from TCIA were applied to develop a radiomics-based TME phenotypes signature using random forest in machine learning. The repeatability of the radiomics signature was then validated in an internal validation set. Two additional independent external validation sets were analyzed to reassess this signature. The Immune phenotype cohort (n = 158) was divided based on CD8 cell infiltration into immune-inflamed and immune-desert phenotypes; these data were utilized to examine the relationship between the immune phenotypes and this signature. Finally, we utilized an Immunotherapy-treated cohort with 77 cases who received anti-PD-1/PD-L1 treatment to evaluate the predictive efficiency of this signature in terms of clinical outcomes.ResultsThe TME phenotypes of breast cancer were separated into two heterogeneous clusters: Cluster A, an "immune-inflamed" cluster, containing substantial innate and adaptive immune cell infiltration, and Cluster B, an "immune-desert" cluster, with modest TME cell infiltration. We constructed a radiomics signature for the TME phenotypes ([AUC] = 0.855; 95% CI 0.777–0.932; p < 0.05) and verified it in an internal validation set (0.844; 0.606–1; p < 0.05). In the known immune phenotypes cohort, the signature can identify either immune-inflamed or immune-desert tumor (0.814; 0.717–0.911; p < 0.05). In the Immunotherapy-treated cohort, patients with objective response had higher baseline radiomics scores than those with stable or progressing disease (p < 0.05); moreover, the radiomics signature achieved an AUC of 0.784 (0.643–0.926; p < 0.05) for predicting immunotherapy response.ConclusionsOur imaging biomarker, a practicable radiomics signature, is beneficial for predicting the TME phenotypes and clinical response in anti-PD-1/PD-L1-treated breast cancer patients. It is particularly effective in identifying the "immune-desert" phenotype and may aid in its transformation into an "immune-inflamed" phenotype.

Machine learning-based radiomics to distinguish pulmonary nodules between lung adenocarcinoma and tuberculosis.

Radiomics is increasingly utilized to distinguish pulmonary nodules between lung adenocarcinoma (LUAD) and tuberculosis (TB). However, it remains unclear whether different segmentation criteria, such as the inclusion or exclusion of the cavity region within nodules, affect the results. A total of 525 patients from two medical centers were retrospectively enrolled. The radiomics features were extracted according to two regions of interest (ROI) segmentation criteria. Multiple logistic regression models were trained to predict the pathology: (1) The clinical model relied on clinical-radiological semantic features; (2) The radiomics models (radiomics+ and radiomics-) utilized radiomics features from different ROIs (including or excluding cavities); (3) the composite models (composite+ and composite-) incorporated both above. In the testing set, the radiomics+/- models and the composite+/- models still possessed efficient prediction performance (AUC ≥ 0.94), while the AUC of the clinical model was 0.881. In the validation set, the AUC of the clinical model was only 0.717, while that of the radiomics+/- models and the composite+/- models ranged from 0.801 to 0.825. The prediction performance of all the radiomics+/- and composite+/- models were significantly superior to that of the clinical model (p < 0.05). Whether the ROI segmentation included or excluded the cavity had no significant effect on these models (radiomics+ vs. radiomics-, composite+ model vs. composite-) (p > 0.05). The present study established a machine learning-based radiomics strategy for differentiating LUAD from TB lesions. The ROI segmentation including or excluding the cavity region may exert no significant effect on the predictive ability.

Machine learning-based radiomics strategy for prediction of acquired EGFR T790M mutation following treatment with EGFR-TKI in NSCLC

The epidermal growth factor receptor (EGFR) Thr790 Met (T790M) mutation is responsible for approximately half of the acquired resistance to EGFR-tyrosine kinase inhibitor (TKI) in non-small-cell lung cancer (NSCLC) patients. Identifying patients at diagnosis who are likely to develop this mutation after first- or second-generation EGFR-TKI treatment is crucial for better treatment outcomes. This study aims to develop and validate a radiomics-based machine learning (ML) approach to predict the T790M mutation in NSCLC patients at diagnosis. We collected retrospective data from 210 positive EGFR mutation NSCLC patients, extracting 1316 radiomics features from CT images. Using the LASSO algorithm, we selected 10 radiomics features and 2 clinical features most relevant to the mutations. We built models with 7 ML approaches and assessed their performance through the receiver operating characteristic (ROC) curve. The radiomics model and combined model, which integrated radiomics features and relevant clinical factors, achieved an area under the curve (AUC) of 0.80 (95% confidence interval [CI] 0.79–0.81) and 0.86 (0.87–0.88), respectively, in predicting the T790M mutation. Our study presents a convenient and noninvasive radiomics-based ML model for predicting this mutation at the time of diagnosis, aiding in targeted treatment planning for NSCLC patients with EGFR mutations.

Tailored Intraoperative MRI Strategies in High-Grade Glioma Surgery: A Machine Learning-Based Radiomics Model Highlights Selective Benefits.

In high-grade glioma (HGG) surgery, intraoperative MRI (iMRI) has traditionally been the gold standard for maximizing tumor resection and improving patient outcomes. However, recent Level 1 evidence juxtaposes the efficacy of iMRI and 5-aminolevulinic acid (5-ALA), questioning the continued justification of iMRI because of its associated costs and extended surgical duration. Nonetheless, drawing from our clinical observations, we postulated that a subset of intricate HGGs may continue to benefit from the adjunctive application of iMRI. In a prospective study of 73 patients with HGG, 5-ALA was the primary technique for tumor delineation, complemented by iMRI to detect residual contrast-enhanced regions. Suboptimal 5-ALA efficacy was defined when (1) iMRI detected contrast-enhanced remnants despite 5-ALA's indication of a gross total resection or (2) surgeons observed residual fluorescence, contrary to iMRI findings. Radiomic features from preoperative MRIs were extracted using a U2-Net deep learning algorithm. Binary logistic regression was then used to predict compromised 5-ALA performance. Resections guided solely by 5-ALA achieved an average removal of 93.14% of contrast-enhancing tumors. This efficacy increased to 97% with iMRI integration, albeit not statistically significant. Notably, for tumors with suboptimal 5-ALA performance, iMRI's inclusion significantly improved resection outcomes (P-value: .00013). The developed deep learning-based model accurately pinpointed these scenarios, and when enriched with radiomic parameters, showcased high predictive accuracy, as indicated by a Nagelkerke R2 of 0.565 and a receiver operating characteristic of 0.901. Our machine learning-driven radiomics approach predicts scenarios where 5-ALA alone may be suboptimal in HGG surgery compared with its combined use with iMRI. Although 5-ALA typically yields favorable results, our analyses reveal that HGGs characterized by significant volume, complex morphology, and left-sided location compromise the effectiveness of resections relying exclusively on 5-ALA. For these intricate cases, we advocate for the continued relevance of iMRI.

Using machine learning-based radiomics to differentiate between glioma and solitary brain metastasis from lung cancer and its subtypes

ObjectiveTo establish a machine learning-based radiomics model to differentiate between glioma and solitary brain metastasis from lung cancer and its subtypes, thereby achieving accurate preoperative classification.Materials and methodsA retrospective analysis was conducted on MRI T1WI-enhanced images of 105 patients with glioma and 172 patients with solitary brain metastasis from lung cancer, which were confirmed pathologically. The patients were divided into the training group and validation group in an 8:2 ratio for image segmentation, extraction, and filtering; multiple layer perceptron (MLP), support vector machine (SVM), random forest (RF), and logistic regression (LR) were used for modeling; fivefold cross-validation was used to train the model; the validation group was used to evaluate and assess the predictive performance of the model, ROC curve was used to calculate the accuracy, sensitivity, and specificity of the model, and the area under curve (AUC) was used to assess the predictive performance of the model.ResultsThe accuracy and AUC of the MLP differentiation model for high-grade glioma and solitary brain metastasis in the validation group was 0.992, 1.000, respectively, while the sensitivity and specificity were 1.000, 0.968, respectively. The accuracy and AUC for the MLP and SVM differentiation model for high-grade glioma and small cell lung cancer brain metastasis in the validation group was 0.966, 1.000, respectively, while the sensitivity and specificity were 1.000, 0.929, respectively. The accuracy and AUC for the MLP differentiation model for high-grade glioma and non-small cell lung cancer brain metastasis in the validation group was 0.982, 0.999, respectively, while the sensitivity and specificity were 0.958, 1.000, respectively.ConclusionThe application of machine learning-based radiomics has a certain clinical value in differentiating glioma from solitary brain metastasis from lung cancer and its subtypes. In the HGG/SBM and HGG/NSCLC SBM validation groups, the MLP model had the best diagnostic performance, while in the HGG/SCLC SBM validation group, the MLP and SVM models had the best diagnostic performance.

Machine learning-based radiomics for amyotrophic lateral sclerosis diagnosis

Timely diagnosis and accurate phenotyping of amyotrophic lateral sclerosis (ALS) is of paramount importance for the clinical management of patients. Magnetic Resonance Imaging (MRI) plays a key role in the clinical work-up of ALS. In this study we investigated the usefulness of radiomics analysis on T1-weighted MRI to define a machine learning-based classification pipeline.We collected 53 controls and 84 patients with ALS from three different scanners. Following dataset harmonization, radiomics analysis was conducted using different features selection and machine learning algorithms to identify the best combination in distinguishing ALS patients from controls and “Classic” from “non-Classical” ALS motor phenotypes.The combined Least Absolute Shrinkage and Selection Operator with Support Vector Machine (SVM) algorithm classified ALS patients with an accuracy of 81.1%. The Maximum Relevance Minimum Redundancy with SVM pipeline was able to distinguish “Classic” from “non-Classical” motor phenotypes with 92.9% accuracy.Radiomics is a promising approach to characterize brain abnormalities in patients with ALS. Radiomics could help to improve diagnosis and may prove useful to assess disease severity and longitudinally monitor ALS patients along the disease course.

A machine learning-based radiomics model for prediction of tumor mutation burden in gastric cancer.

Purpose: To evaluate the potential of machine learning (ML)-based radiomics approach for predicting tumor mutation burden (TMB) in gastric cancer (GC). Methods: The contrast enhanced CT (CECT) images with corresponding clinical information of 256GC patients were retrospectively collected. Patients were separated into training set (n = 180) and validation set (n = 76). A total of 3,390 radiomics features were extracted from three phases images of CECT. The least absolute shrinkage and selection operator (LASSO) model was used for feature screening. Seven machine learning (ML) algorithms were employed to find the optimal classifier. The predictive ability of radiomics model (RM) was evaluated with receiver operating characteristic. The correlation between RM and TMB values was evaluated using Spearman's correlation coefficient. The explainability of RM was assessed by the Shapley Additive explanations (SHAP) method. Results: Logistic regression algorithm was chosen for model construction. The RM showed good predictive ability of TMB status with AUCs of 0.89 [95% confidence interval (CI): 0.85-0.94] and 0.86 (95% CI: 0.74-0.98) in the training and validation sets. The correlation analysis revealed a good correlation between RM and TMB levels (correlation coefficient: 0.62, p < 0.001). The RM also showed favorable and stable predictive accuracy within the cutoff value range 6-16 mut/Mb in both sets. Conclusion: The ML-based RM offered a promising image biomarker for predicting TMB status in GC patients.

Machine Learning-Based Models for Assessing Postoperative Risk Factors in Patients with Cervical Cancer

To investigate the value of machine learning-based radiomics, intravoxel incoherent motion (IVIM) diffusion-weighted imaging and its combined model in predicting the postoperative risk factors of parametrial infiltration (PI), lymph node metastasis (LNM), deep muscle invasion (DMI), lymph-vascular space invasion (LVSI), pathological type (PT), differentiation degree (DD), and Ki-67 expression level in patients with cervical cancer. The data of 180 patients with cervical cancer were retrospectively analyzed and randomized 2:1 into a training and validation group. The IVIM-DWI and radiomics parameters of primary lesions were measured in all patients. Seven machine learning methods were used to calculate the optimal radiomics score (Rad-score), which was combined with IVIM-DWI and clinical parameters to construct nomograms for predicting the risk factors of cervical cancer, with internal and external validation. The diagnostic efficacy of the nomograms based on clinical and imaging parameters wassignificantly better than MRI assessment alone. The area under the curve (AUC) of nomograms and MRI for the assessment of PI, LNM, and DMI were 0.981 vs 0.868, 0.848 vs 0.639, and 0.896 vs 0.780, respectively. Nomograms also performed well in the assessment of LVSI, PT, DD, and Ki-67 expression levels, with AUC of 0.796, 0.854, 0.806, 0.839 and 0.840, 0.856, 0.810, 0.832 in the training and validation groups. Machine learning-based nomograms can serve as a useful tool for assessing postoperative risk factors in patients with cervical cancer.

Machine learning-based radiomics to differentiate immune-mediated necrotizing myopathy from limb-girdle muscular dystrophy R2 using MRI.

This study aimed to assess the feasibility of a machine learning-based radiomics tools to discriminate between Limb-girdle muscular dystrophy R2 (LGMDR2) and immune-mediated necrotizing myopathy (IMNM) using lower-limb muscle magnetic resonance imaging (MRI) examination. After institutional review board approval, 30 patients with genetically proven LGMDR2 (12 females; age, 34.0 ± 11.3) and 45 patients with IMNM (28 females; age, 49.2 ± 16.6) who underwent lower-limb MRI examination including T1-weighted and interactive decomposition water and fat with echos asymmetric and least-squares estimation (IDEAL) sequences between July 2014 and August 2022 were included. Radiomics features of muscles were obtained, and four machine learning algorithms were conducted to select the optimal radiomics classifier for differential diagnosis. This selected algorithm was performed to construct the T1-weighted (TM), water-only (WM), or the combined model (CM) for calf-only, thigh-only, or the calf and thigh MR images, respectively. And their diagnostic performance was studied using area under the curve (AUC) and compared to the semi-quantitative model constructed by the modified Mercuri scale of calf and thigh muscles scored by two radiologists specialized in musculoskeletal imaging. The logistic regression (LR) model was the optimal radiomics model. The performance of the WM and CM for thigh-only images (AUC 0.893, 0.913) was better than those for calf-only images (AUC 0.846, 0.880) except the TM. For "calf + thigh" images, the TM, WM, and CM models always performed best (AUC 0.953, 0.907, 0.953) with excellent accuracy (92.0, 84.0, 88.0%). The AUCs of the Mercuri model of the calf, thigh, and "calf + thigh" images were 0.847, 0.900, and 0.953 with accuracy (84.0, 84.0, 88.0%). Machine learning-based radiomics models can differentiate LGMDR2 from IMNM, performing better than visual assessment. The model built by combining calf and thigh images presents excellent diagnostic efficiency.

Simultaneous 18F-FDG PET/MRI Radiomics and Machine Learning Analysis of the Primary Breast Tumor for the Preoperative Prediction of Axillary Lymph Node Status in Breast Cancer.

In this prospective study, 117 female patients (mean age = 53 years) with 127 histologically proven breast cancer lesions (lymph node (LN) positive = 85, LN negative = 42) underwent simultaneous 18F-FDG PET/MRI of the breast. Quantitative parameters were calculated from dynamic contrast-enhanced (DCE) imaging (tumor Mean Transit Time, Volume Distribution, Plasma Flow), diffusion-weighted imaging (DWI) (tumor ADCmean), and PET (tumor SUVmax, mean and minimum, SUVmean of ipsilateral breast parenchyma). Manual whole-lesion segmentation was also performed on DCE, T2-weighted, DWI, and PET images, and radiomic features were extracted. The dataset was divided into a training (70%) and a test set (30%). Multi-step feature selection was performed, and a support vector machine classifier was trained and tested for predicting axillary LN status. 13 radiomic features from DCE, DWI, T2-weighted, and PET images were selected for model building. The classifier obtained an accuracy of 79.8 (AUC = 0.798) in the training set and 78.6% (AUC = 0.839), with sensitivity and specificity of 67.9% and 100%, respectively, in the test set. A machine learning-based radiomics model comprising 18F-FDG PET/MRI radiomic features extracted from the primary breast cancer lesions allows high accuracy in non-invasive identification of axillary LN metastasis.