3D Regions Of Interest Research Articles

Multiparametric Characterization of Focal Cortical Dysplasia Using 3D MR Fingerprinting.

To develop a multiparametric machine-learning (ML) framework using high-resolution 3 dimensional (3D) magnetic resonance (MR) fingerprinting (MRF) data for quantitative characterization of focal cortical dysplasia (FCD). We included 119 subjects, 33 patients with focal epilepsy and histopathologically confirmed FCD, 60 age- and gender-matched healthy controls (HCs), and 26 disease controls (DCs). Subjects underwent whole-brain 3 Tesla MRF acquisition, the reconstruction of which generated T1 and T2 relaxometry maps. A 3D region of interest was manually created for each lesion, and z-score normalization using HC data was performed. We conducted 2D classification with ensemble models using MRF T1 and T2 mean and standard deviation from gray matter and white matter for FCD versus controls. Subtype classification additionally incorporated entropy and uniformity of MRF metrics, as well as morphometric features from the morphometric analysis program (MAP). We translated 2D results to individual probabilities using the percentage of slices above an adaptive threshold. These probabilities and clinical variables were input into a support vector machine for individual-level classification. Fivefold cross-validation was performed and performance metrics were reported using receiver-operating-characteristic-curve analyses. FCD versus HC classification yielded mean sensitivity, specificity, and accuracy of 0.945, 0.980, and 0.962, respectively; FCD versus DC classification achieved 0.918, 0.965, and 0.939. In comparison, visual review of the clinical magnetic resonance imaging (MRI) detected 48% (16/33) of the lesions by official radiology report. In the subgroup where both clinical MRI and MAP were negative, the MRF-ML models correctly distinguished FCD patients from HCs and DCs in 98.3% of cross-validation trials. Type II versus non-type-II classification exhibited mean sensitivity, specificity, and accuracy of 0.835, 0.823, and 0.83, respectively; type IIa versus IIb classification showed 0.85, 0.9, and 0.87. In comparison, the transmantle sign was present in 58% (7/12) of the IIb cases. The MRF-ML framework presented in this study demonstrated strong efficacy in noninvasively classifying FCD from normal cortex and distinguishing FCD subtypes. ANN NEUROL 2024;96:944-957.

CT radiomics to predict pathologic complete response after neoadjuvant immunotherapy plus chemoradiotherapy in locally advanced esophageal squamous cell carcinoma.

To develop and validate a CT-based radiomics model to predict pathologic complete response (pCR) after neoadjuvant immunotherapy plus chemoradiotherapy (NICRT) in locally advanced esophageal squamous cell carcinoma (ESCC). A total of 105 patients with locally advanced ESCC receiving NICRT from February 2019 to December 2023 were enrolled. Patients were randomly divided into the training cohort and the test cohort at a 3:1 ratio. Enhanced CT scans were obtained before NICRT treatment. The 2D and 3D regions of interest were segmented, and features were extracted, followed by feature selection. Six algorithms were applied to construct the radiomics and clinical models. These models were evaluated by area under curve (AUC), accuracy, sensitivity, and specificity, and their respective optimal algorithms were further compared. Forty-eight patients (45.75%) achieved pCR after NICRT. The AUC values of three algorithms in 2D radiomics models were higher than those in the 3D radiomics model and clinical model. Among these, the 2D radiomics model based on eXtreme Gradient Boosting (XGBoost) exhibited the best performance, with an AUC of 0.89 (95% CI, 0.81-0.97), accuracy of 0.85, sensitivity of 0.86, and specificity of 0.84 in the training cohort, and an AUC of 0.80 (95% CI, 0.64-0.97), accuracy of 0.77, sensitivity of 0.84, and specificity of 0.69 in the test cohort. Calibration curves also showed good agreement between predicted and actual response, and the decision curve analysis further confirmed its clinical applicability. The 2D radiomics model can effectively predict pCR to NICRT in locally advanced ESCC. Question Can CT-based radiomics predict pathologic complete response (pCR) after neoadjuvant immunotherapy plus chemoradiotherapy (NICRT) in locally advanced esophageal squamous cell carcinoma (ESCC)? Findings The model based on eXtreme Gradient Boosting (XGBoost) performed best, with an AUC of 0.89 in the training and 0.80 in the test cohort. Clinical relevance This CT-based radiomics model exhibits promising performance for predicting pCR to NICRT in locally advanced ESCC, which may be valuable in personalized treatment plan optimization.

Prediction of lymphovascular invasion in esophageal squamous cell carcinoma by computed tomography-based radiomics analysis: 2D or 3D ?

BackgroundTo compare the performance between one-slice two-dimensional (2D) and whole-volume three-dimensional (3D) computed tomography (CT)-based radiomics models in the prediction of lymphovascular invasion (LVI) status in esophageal squamous cell carcinoma (ESCC).MethodsTwo hundred twenty-four patients with ESCC (158 LVI-absent and 66 LVI-present) were enrolled in this retrospective study. The enrolled patients were randomly split into the training and testing sets with a 7:3 ratio. The 2D and 3D radiomics features were derived from the primary tumors’ 2D and 3D regions of interest (ROIs) using 1.0 mm thickness contrast-enhanced CT (CECT) images. The 2D and 3D radiomics features were screened using inter-/intra-class correlation coefficient (ICC) analysis, Wilcoxon rank-sum test, Spearman correlation test, and the least absolute shrinkage and selection operator, and the radiomics models were built by multivariate logistic stepwise regression. The performance of 2D and 3D radiomics models was assessed by the area under the receiver operating characteristic (ROC) curve. The actual clinical utility of the 2D and 3D radiomics models was evaluated by decision curve analysis (DCA).ResultsThere were 753 radiomics features from 2D ROIs and 1130 radiomics features from 3D ROIs, and finally, 7 features were retained to construct 2D and 3D radiomics models, respectively. ROC analysis revealed that in both the training and testing sets, the 3D radiomics model exhibited higher AUC values than the 2D radiomics model (0.930 versus 0.852 and 0.897 versus 0.851, respectively). The 3D radiomics model showed higher accuracy than the 2D radiomics model in the training and testing sets (0.899 versus 0.728 and 0.788 versus 0.758, respectively). In addition, the 3D radiomics model has higher specificity and positive predictive value, while the 2D radiomics model has higher sensitivity and negative predictive value. The DCA indicated that the 3D radiomics model provided higher actual clinical utility regarding overall net benefit than the 2D radiomics model.ConclusionsBoth 2D and 3D radiomics features can be employed as potential biomarkers to predict the LVI in ESCC. The performance of the 3D radiomics model is better than that of the 2D radiomics model for the prediction of the LVI in ESCC.

Utility of diffusion-weighted imaging in differentiating benign and malignant breast lesions.

Breast cancer presents a significant global health burden. An accurate differentiation between benign and malignant lesions is imperative for timely intervention. While dynamic contrast enhanced MRI (DCE-MRI) is highly sensitive, its specificity is limited. This has led to the exploration of diffusion-weighted imaging (DWI) in distinguishing between benign and malignant breast lesions. The study aimed to explore the diagnostic utility of DWI in distinguishing between benign and malignant breast lesions. Assessment of 38 breast lesions using DWI with a b value of 800 s/mm2, performed with 3 Tesla MRI. The diagnostic performance of two different region of Interest (ROI) placement approaches was compared to obtain a feasible cut-off value of apparent diffusion coefficient (ADC) to differentiate between malignant and benign lesions. The histopathological reports were used as the gold standard. ADC values of malignant lesions were significantly lower than those of benign lesions (0.84 × 10-3 mm2/s vs. 1.54 × 10-3 mm2/s). The average ADC measured using a small-sized 2D ROI including the darkest part in the ADC map, performed better than the large 2D ROI covering the entire lesion. Using a cut-off value of 0.98 × 10-3 mm2/s, ADC obtained high sensitivity (90%) and specificity (88.9%) in distinguishing between benign and malignant breast lesions. Utilising quantitative analysis of DWI with ADC value measurement, reliably distinguished between benign and malignant breast lesions in this cohort, especially when employing a higher b value of 800 s/mm2.

Utilizing computed tomography-based renal parenchymal volumes to calculate split renal function in patients with ureteral stricture disease.

Evaluation of split renal function (SRF) is critical for guiding surgical treatment decisions for patients with ureteral stricture disease (USD). We aimed to determine whether computed tomography (CT)-based renal parenchymal volumes may be used to predict SRF in patients with USD. We retrospectively reviewed all patients undergoing surgical management for USD at a single institution from October 2021 to January 2024. Patients who had preoperative nuclear medicine scan (NMS) and CT scan with intravenous contrast that were obtained within six weeks of each other were included. Interval between NMS and CT could be longer if the affected renal unit was drained with ureteral stent and/or percutaneous nephrostomy. Volume measurements were obtained using the 3D Region of Interest (ROI) Tool on Visage®7 Enterprise Imaging Platform (Visage Inc., San Diego, USA) by two investigators that were blinded to NMS derived SRF. Intraclass correlation coefficient (ICC) was used to assess consistency between investigators. Predictive accuracy was assessed using Pearson correlation coefficient (r) and linear regression. 40 of 160 patients met inclusion criteria. There was excellent reliability in calculating renal parenchymal volume between raters (ICC = 0.990). There was a strong linear correlation between estimated CT SRF and NMS SRF (r = 0.912, p < 0.00001). A linear regression model found RObservedSRF = -0.013 + 1.015(REstimatedSRF), with r2 = 0.832. CT-derived parenchymal volume analysis may be used to estimate SRF in patients with USD. This may obviate the need to obtain preoperative renal scans for SRF measurement in selected patients when assessing surgical management options.

Radiomics predicts the prognosis of patients with clear cell renal cell carcinoma by reflecting the tumor heterogeneity and microenvironment

PurposeWe aimed to develop and externally validate a CT-based deep learning radiomics model for predicting overall survival (OS) in clear cell renal cell carcinoma (ccRCC) patients, and investigate the association of radiomics with tumor heterogeneity and microenvironment.MethodsThe clinicopathological data and contrast-enhanced CT images of 512 ccRCC patients from three institutions were collected. A total of 3566 deep learning radiomics features were extracted from 3D regions of interest. We generated the deep learning radiomics score (DLRS), and validated this score using an external cohort from TCIA. Patients were divided into high and low-score groups by the DLRS. Sequencing data from the corresponding TCGA cohort were used to reveal the differences of tumor heterogeneity and microenvironment between different radiomics score groups. What’s more, univariate and multivariate Cox regression were used to identify independent risk factors of poor OS after operation. A combined model was developed by incorporating the DLRS and clinicopathological features. The SHapley Additive exPlanation method was used for interpretation of predictive results.ResultsAt multivariate Cox regression analysis, the DLRS was identified as an independent risk factor of poor OS. The genomic landscape of different radiomics score groups was investigated. The heterogeneity of tumor cell and tumor microenvironment significantly varied between both groups. In the test cohort, the combined model had a great predictive performance, with AUCs (95%CI) for 1, 3 and 5-year OS of 0.879(0.868–0.931), 0.854(0.819–0.899) and 0.831(0.813–0.868), respectively. There was a significant difference in survival time between different groups stratified by the combined model. This model showed great discrimination and calibration, outperforming the existing prognostic models (all p values < 0.05).ConclusionThe combined model allowed for the prognostic prediction of ccRCC patients by incorporating the DLRS and significant clinicopathologic features. The radiomics features could reflect the tumor heterogeneity and microenvironment.

Radiomic feature robustness in CT scans affected by fiducial marker induced streak artifacts for patients with hepatocellular carcinoma.

Stereotactic body radiation therapy for hepatocellular carcinoma necessitates the implantation of gold fiducial markers in the liver, resulting in artifacts on computed tomography (CT) scans, which affect radiomic feature values. This report aims to assess the effect of these artifacts on radiomic features and how removing CT slices affects radiomic features extracted from 3D regions of interest (ROI). First, the range variation in 38 tumor contours unaffected by artifacts was assessed after sequentially and randomly removing 25%, 50%, 75% of slices. Subsequently, the agreement of feature values before and after removing ROI slices containing artifacts from 186 patients' CT scans was assessed with Lin's concordance correlation coefficient (CCC) and a Wilcoxon signed-rank test. In artifact-free tumor volumes, at least 71% of features remain robust with up to 50% of slices removed, while 56% remains robust with up to 75% of slices removed. When comparing contours before and after removing slices containing artifacts, around a third of features in the tumor contour and surrounding area remain robust (CCC>0.9), compared to 44% in the healthy liver. Concerning the tumor, 13% (Gray Level Size Zone Matrix) to 61% (first order) of the features remain robust (CCC>0.9). Over 90% of features differ significantly as assessed by Wilcoxon signed-rank test, however. This study demonstrates that removing slices containing artifacts is a feasible solution for the CT fiducial problem in this patient population and provides insight into which features are affected.

Abstract 2309: 3D spatial quantification of lymphocyte infiltration and collagen features in the tumor microenvironment using a novel assay: 3D I/O Pro™

Abstract Background: The composition of the tumor microenvironment (TME) is a major determinant of response to therapy in many solid tumors. To-date, characterization of the TME has been based on limited analysis of thin tissue sections. Here we demonstrate the utility of an end-to-end 3D spatial biology workflow, the 3D I/O Pro™, based on whole tissue imaging, to identify and quantify tumor cells, lymphocytes, and collagen features in human FFPE tissue samples. Methods: Five human colorectal cancer FFPE blocks were deparaffinized, stained with nuclear (TO-PRO-3) and general protein (eosin) dyes, and cleared using a modified iDISCO protocol. Entire samples were imaged at 2 µm/pixel resolution with a hybrid open-top light-sheet microscope, the 3Di™. Regions of interest (ROIs) with a volume of 0.5 mm3 were re-imaged at 0.33 µm/pixel. Cell nuclei and collagen were segmented and 3D spatial relationships between tumor cells, lymphocytes, and collagen fibers were quantified. Analyses were performed within select ROIs using 3Dai™ tools including U-Net, CytoMAP, CT-FIRE, CurveAlign, and custom Python scripts. Results: The ratio of stromal to tumor parenchymal lymphocytes (lymphocyte infiltration ratio) varied from 1.4 up to 9.1 in 3D volumes and 1 to 25 in 2D virtual sections taken throughout all 5 CRC samples. The ratio of the collagen fibers within 100 µm of the tumor border-oriented perpendicular to the tumor-stromal boundary compared to those oriented parallel to the tumor-stromal boundary was 0.37 in a proof-of-concept evaluation of a 2D virtual section with a lymphocyte infiltration ratio of 12. Qualitatively, areas with perpendicular collagen had more lymphocyte infiltration into the tumor parenchyma than areas with parallel collagen orientation. Conclusions: We demonstrated that the 3D I/O Pro™ pipeline can quantify lymphocyte density in tumor parenchyma and stroma and analyze collagen features, including orientation, within 3D ROIs in the TME. This workflow allows us to characterize tumors based on many complex spatial relationships and could have broad applicability in research and development of novel cancer therapies that target tumor fibrosis or other features of the TME. In the future, we plan to correlate features from the 3D TME with response to immunotherapy and use these features to refine histologic definitions of immune spatial phenotypes. Citation Format: Caleb Stoltzfus, Jasmine Wilson, Nathan Grant, Alexandra Alvarsson, Anna DeWitt, David Simmons, Brandy Olin, Bonnie Phillips, Laura A. Dillon, G. Travis Clifton, Nicholas P. Reder. 3D spatial quantification of lymphocyte infiltration and collagen features in the tumor microenvironment using a novel assay: 3D I/O Pro™ [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2309.

COverlap: a Fiji toolset for the 3D co-localization of two fluorescent nuclear markers in confocal images

With the increasing complexity and throughput of microscopy experiments, it has become essential for biologists to navigate computational means of analysis to produce automated and reproducible workflows. Bioimage analysis workflows being largely underreported in method sections of articles, it is however quite difficult to find practical examples of documented scripts to support beginner programmers in biology. Here, we introduce COverlap, a Fiji toolset composed of four macros, for the 3D segmentation and co-localization of fluorescent nuclear markers in confocal images. The toolset accepts batches of multichannel z-stack images, segments objects in two channels of interest, and outputs object counts and labels, as well as co-localization results based on the physical overlap of objects. The first macro is a preparatory step that produces maximum intensity projections of images for visualization purposes. The second macro assists users in selecting batch-suitable segmentation parameters by testing them on small portions of the images. The third macro performs automated segmentation and co-localization analysis, and saves the parameters used, the results table, the 3D regions of interest (ROIs) of co-localizing objects, and two types of verification images with segmentation and co-localization masks for each image of the batch. The fourth macro allows users to review the verification images displaying segmentation masks and the location of co-localization events, and to perform corrections such as ROI adjustment, z-stack reslicing, and volume estimation correction in an automatically documented manner. To illustrate how COverlap operates, we present an experiment in which we identified rare endothelial proliferation events in adult rat brain slices on more than 350 large tiled z-stacks. We conclude by discussing the reproducibility and generalizability of the toolset, its limitations for different datasets, and its potential use as a template that is adaptable to other types of analyses.

Region of interest (ROI) extraction and crack detection for UAV-based bridge inspection using point cloud segmentation and 3D-to-2D projection

For digital-image-based bridge inspection tasks, images captured by camera-carrying unmanned aircraft vehicles (UAVs) usually contain both the region of interest (ROI) and the background. However, accurately detecting cracks in concrete surface images containing background information is challenging. To improve UAV-based bridge inspection, an image extraction and crack detection methodology is presented in this paper. First, a deep-learning-based semantic segmentation network RandLA-BridgeNet for large-scale bridge point clouds, which can facilitate 3D ROI extraction, is trained and tested. Second, an image ROI extraction method based on 3D-to-2D projection is presented to generate images containing only the ROI. Finally, a data-driven deep learning convolutional neural network (CNN) called the grid-based classification and box-based detection fusion model (GCBD) is utilized to identify cracks in the processed images. An experiment is conducted on highway bridge images to validate the presented methodology. The overall semantic segmentation and image ROI extraction accuracies are 97.0% and 98.9%, respectively. After ROI extraction, 47.9% of the grid cells, which represent background misrecognition, are filtered, greatly improving the crack identification accuracy.

Abstract A149: Characterization of HER3 targeted radioligand therapy using molecular imaging

Abstract Background: Radioligand therapy (RLT) is an emerging treatment modality that has shown potential to improve survival in cancer patients that often have limited or non-effective therapeutic options. We previously demonstrated that anti-HER3 antibody radioconjugate (HER3 ARC) has antitumor efficacy against HER3 expressing tumor xenograft models. One important step in the early development of RLT is to understand the targeting properties and the distribution of the RLT in vivo. Non-invasive molecular imaging such as single photon emission computed tomography (SPECT) and positron emission tomography (PET) offer unique opportunities to understand drug behavior in vivo, permitting real time drug targeting and distribution analysis. In this study, we assessed the targeting properties and distribution of our HER3 ARC in HER3 expressing preclinical xenograft tumor models using both SPECT and PET. Methods: HER3 monoclonal antibody was conjugated with p-SCN-Bn-deferoxamine (DFO) or p-SCN-Bn-DOTA (DOTA) then labeled with Zirconium-89 (89Zr) or Indium-111 (111In), respectively. Flow cytometry and gamma counting was used to characterize HER3 ARC biological activity in HER3-expressing tumor cell lines. Balb/c mice subcutaneously inoculated with HER3-expressing xenograft tumors (lung, ovarian and colorectal) were intravenously administered with 89Zr-DFO-HER3 or 111In-DOTA-HER3. PET or SPECT scans were acquired 72 and 120-144hrs post injection of the radiotracer. Separate cohorts of mice were simultaneously administered with 89Zr-DFO-HER3 or 111In-DOTA-HER3 and 20-fold excess of native HER3 antibody as control to block radiotracer binding. Tissue uptake was quantified by drawing 3D or 2D regions of interest (ROI) for PET or SPECT images, respectively. Results: HER3 ARC radiotracers were successfully prepared and demonstrated similar binding to HER3 expressing cells compared to native HER3 antibody. In tumor-bearing mice, 89Zr-DFO-HER3 and 111In-DOTA-HER3 specifically accumulated in HER3 expressing tumors and cleared throughout the reticuloendothelial system. High tumor uptake was observed 120hrs post radiotracer injection using both imaging modalities. In mice pre-treated with 20-fold excess of native antibody, tumor uptake was significantly lower, showing tumor binding specificity of ARC. PET and SPECT images showed similar tracer distribution profile in the HER3 tumors. As expected, radiotracer clearance organ signals were also visualized. Conclusions: In this study we show that PET or SPECT imaging using 89Zr or 111In, respectively, offers the potential to assess the distribution of HER3 targeted radioligand therapy during early stages of drug development. HER3 ARC showed highly specific HER3 tumor uptake that remained at least 120hrs post injection. Our findings warrant further investigation to support advancement of HER3 ARC as a therapeutic option. Citation Format: Sumit Mukherjee, Debbie Lewis, Jason Li, Le-Cun Xu, Amanda Chin, Mary Chen, Monideepa Roy, Patrik Brodin, William van der Touw, Helen Kotanides, Denis Beckford-Vera. Characterization of HER3 targeted radioligand therapy using molecular imaging [abstract]. In: Proceedings of the AACR-NCI-EORTC Virtual International Conference on Molecular Targets and Cancer Therapeutics; 2023 Oct 11-15; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2023;22(12 Suppl):Abstract nr A149.

Computed tomography and guidelines-based human-machine fusion model for predicting resectability of the pancreatic cancer.

The study aims to develop a hybrid machine learning model for predicting resectability of the pancreatic cancer, which is based on computed tomography (CT) and National Comprehensive Cancer Network (NCCN) guidelines. We retrospectively studied 349 patients. One hundred seventy-one cases from Center 1 and 92 cases from Center 2 were used as the primary training cohort, and 66 cases from Center 3 and 20 cases from Center 4 were used as the independent test dataset. Semi-automatic module of ITK-SNAP software was used to assist CT image segmentation to obtain three-dimensional (3D) imaging region of interest (ROI). There were 788 handcrafted features extracted for 3D ROI using PyRadiomics. The optimal feature subset consists of three features screened by three feature selection methods as the input of the SVM to construct the conventional radiomics-based predictive model (cRad). 3D ROI was used to unify the resolution by 3D spline interpolation method for constructing the 3D tumor imaging tensor. Using 3D tumor image tensor as input, 3D kernelled support tensor machine-based predictive model (KSTM), and 3D ResNet-based deep learning predictive model (ResNet) were constructed. Multi-classifier fusion ML model is constructed by fusing cRad, KSTM, and ResNet using multi-classifier fusion strategy. Two experts with more than 10years of clinical experience were invited to reevaluate each patient based on their CECT following the NCCN guidelines to obtain resectable, unresectable, and borderline resectable diagnoses. The three results were converted into probability values of 0.25, 0.75, and 0.50, respectively, according to the traditional empirical method. Then it is used as an independent classifier and integrated with multi-classifier fusion machine learning (ML) model to obtain the human-machine fusion ML model (HMfML). Multi-classifier fusion ML model's area under receiver operating characteristic curve (AUC; 0.8610), predictive accuracy (ACC: 80.23%), sensitivity (SEN: 78.95%), and specificity (SPE: 80.60%) is better than cRad, KSTM, and ResNet-based single-classifier models and their two-classifier fusion models. This means that three different models have mined complementary CECT feature expression from different perspectives and can be integrated through CFS-ER, so that the fusion model has better performance. HMfML's AUC (0.8845), ACC (82.56%), SEN (84.21%), SPE (82.09%). This means that ML models might learn extra information from CECT that experts cannot distinguish, thus complementing expert experience and improving the performance of hybrid ML models. HMfML can predict PC resectability with high accuracy.

Numerical investigation reveals challenges in measuring the contrast recovery coefficients in PET

Objective. Contrast recovery coefficient (CRC) is essential for image quality (IQ) assessment in positron emission tomography (PET), typically measured according to the National Electrical Manufacturers Association (NEMA) NU 2 standard. This study quantifies systematic uncertainties of the CRC measurement by a numerical investigation of the effects from scanner-independent parameters like voxel size, region-of-interest (ROI) misplacement, and sphere position on the underlying image grid. Approach. CRC measurements with 2D and 3D ROIs were performed on computer-generated images of a NEMA IQ-like phantom, using voxel sizes of 1–4 mm for sphere diameters of 5–40 mm—first in absence of noise and blurring, then with simulated spatial resolution and image noise with varying noise levels. The systematic uncertainties of the CRC measurement were quantified from above variations of scanner-independent parameters. Subsampled experimental images of a NEMA IQ phantom were additionally used to investigate the impact of ROI misplacement at different noise levels. Main results. In absence of noise and blurring, systematic uncertainties were up to 28.8% and 31.0% with 2D and 3D ROIs, respectively, for the 10 mm sphere, with the highest impact from ROI misplacement. In all cases, smaller spheres showed higher uncertainties with larger voxels. Contrary to prior assumptions, the use of 3D ROIs did not exhibit less susceptibility for parameter changes. Experimental and computer-generated images both demonstrated considerable variations on individual CRC measurements when background coefficient-of-variation exceeded 20%, despite negligible effects on mean CRC. Significance. This study underscores the effect of scanner-independent parameters on reliability, reproducibility, and comparability of CRC measurements. Our findings highlight the trade-off between the benefits of smaller voxel sizes and noise-induced CRC fluctuations, which is not considered in the current version of the NEMA IQ standards. The results furthermore warrant adjustments to the standard to accommodate the advances in sensitivity and spatial resolution of current-generation PET scanners.

Diagnostic performance of CT scan–based radiomics for prediction of lymph node metastasis in gastric cancer: a systematic review and meta-analysis

ObjectiveThe purpose of this study was to evaluate the diagnostic performance of computed tomography (CT) scan–based radiomics in prediction of lymph node metastasis (LNM) in gastric cancer (GC) patients.MethodsPubMed, Embase, Web of Science, and Cochrane Library databases were searched for original studies published until 10 November 2022, and the studies satisfying the inclusion criteria were included. Characteristics of included studies and radiomics approach and data for constructing 2 × 2 tables were extracted. The radiomics quality score (RQS) and Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) were utilized for the quality assessment of included studies. Overall sensitivity, specificity, diagnostic odds ratio (DOR), and area under the curve (AUC) were calculated to assess diagnostic accuracy. The subgroup analysis and Spearman’s correlation coefficient was done for exploration of heterogeneity sources.ResultsFifteen studies with 7,010 GC patients were included. We conducted analyses on both radiomics signature and combined (based on signature and clinical features) models. The pooled sensitivity, specificity, DOR, and AUC of radiomics models compared to combined models were 0.75 (95% CI, 0.67–0.82) versus 0.81 (95% CI, 0.75–0.86), 0.80 (95% CI, 0.73–0.86) versus 0.85 (95% CI, 0.79–0.89), 13 (95% CI, 7–23) versus 23 (95% CI, 13–42), and 0.85 (95% CI, 0.81–0.86) versus 0.90 (95% CI, 0.87–0.92), respectively. The meta-analysis indicated a significant heterogeneity among studies. The subgroup analysis revealed that arterial phase CT scan, tumoral and nodal regions of interest (ROIs), automatic segmentation, and two-dimensional (2D) ROI could improve diagnostic accuracy compared to venous phase CT scan, tumoral-only ROI, manual segmentation, and 3D ROI, respectively. Overall, the quality of studies was quite acceptable based on both QUADAS-2 and RQS tools.ConclusionCT scan–based radiomics approach has a promising potential for the prediction of LNM in GC patients preoperatively as a non-invasive diagnostic tool. Methodological heterogeneity is the main limitation of the included studies.Systematic review registrationhttps://www.crd.york.ac.uk/Prospero/display_record.php?RecordID=287676, identifier CRD42022287676.

Computed tomography radiomics identification of T1–2 and T3–4 stages of esophageal squamous cell carcinoma: two-dimensional or three-dimensional?

BackgroundTo evaluate two-dimensional (2D) and three-dimensional (3D) computed tomography (CT) radiomics analysis for the T stage of esophageal squamous cell carcinoma (ESCC).Methods398 patients with pathologically confirmed ESCC were divided into training and testing sets. All patients underwent chest CT scans preoperatively. For each tumor, based on CT images, a 2D region of interest (ROI) was outlined on the largest cross-sectional area, and a 3D ROI was outlined layer by layer on each section of the tumor. The radiomics platform was used for feature extraction. For feature selection, stepwise logistic regression was used. The receiver operating characteristic (ROC) curve was used to assess the diagnostic performance of the 2D radiomics model versus the 3D radiomics model. The differences were compared using the DeLong test. The value of the clinical utility of the two radiomics models was evaluated.Results1595 radiomics features were extracted. After screening, two radiomics models were constructed. In the training set, the difference between the area under the curve (AUC) of the 2D radiomics model (AUC = 0.831) and the 3D radiomics model (AUC = 0.830) was not statistically significant (p = 0.973). In the testing set, the difference between the AUC of the 2D radiomics model (AUC = 0.807) and the 3D radiomics model (AUC = 0.797) was also not statistically significant (p = 0.748). A 2D model was equally useful as a 3D model in clinical situations.ConclusionThe performance of 2D radiomics model is comparable to that of 3D radiomics model in distinguishing between the T1-2 and T3-4 stages of ESCC. In addition, 2D radiomics model may be a more feasible option due to the shorter time required for segmenting the ROI.

MRI-based machine learning models predict the malignant biological behavior of meningioma

BackgroundThe WHO grade and Ki-67 index are independent indices used to evaluate the malignant biological behavior of meningioma. This study aims to develop MRI-based machine learning models to predict the malignant biological behavior of meningioma from the perspective of the WHO grade, Ki-67 index, and their combination.MethodsThis multicenter, retrospective study included 313 meningioma patients, of which 70 were classified as high-grade (WHO II/III) and 243 as low-grade (WHO I). The Ki-67 expression was classified into low-expression (n = 216) and high-expression (n = 97) groups with a threshold of 5%. Among them, there were 128 patients with malignant biological behavior whose WHO grade or Ki-67 index increased either or both. Data from Center A and B are were utilized for model development, while data from Center C and D were used for external validation. Radiomic features were extracted from the maximum cross-sectional area (2D) region of Interest (ROI) and the whole tumor volume (3D) ROI using different paraments from the T1, T2-weighted, and T1 contrast-enhanced sequences (T1CE), followed by five independent feature selections and eight classifiers. 240 prediction models were constructed to predict the WHO grade, Ki-67 index and their combination respectively. Models were evaluated by cross-validation in training set (n = 224). Suitable models were chosen by comparing the cross-validation (CV) area under the curves (AUC) and their relative standard deviations (RSD). Clinical and radiological features were collected and analyzed; meaningful features were combined with radiomic features to establish the clinical-radiological-radiomic (CRR) models. The receiver operating characteristic (ROC) analysis was used to evaluate those models in validation set. Radiomic models and CRR models were compared by Delong test.Results1218 and 1781 radiomic features were extracted from 2D ROI and 3D ROI of each sequence. The selected grade, Ki-67 index and their combination radiomic models were T1CE-2D-LASSO-LR, T1CE-3D-LASSO-NB, and T1CE-2D-LASSO-LR, with cross-validated AUCs on the training set were 0.857, 0.798, and 0.888, the RSDs were 0.06, 0.09, and 0.05, the validation set AUCs were 0.829, 0.752, and 0.904, respectively. Heterogeneous enhancement was found to be associated with high grade and Ki-67 status, while surrounding invasion was associated with the high grade status, peritumoral edema and cerebrospinal fluid space surrounding tumor were correlated with the high Ki-67 status. The Delong test showed that these significant radiological features did not significantly improve the predictive performance. The AUCs for CRR models predicting grade, Ki-67 index, and their combination in the validation set were 0.821, 0.753, and 0.906, respectively.ConclusionsThis study demonstrated that MRI-based machine learning models could effectively predict the grade, Ki-67 index of meningioma. Models considering these two indices might be valuable for improving the predictive sensitivity and comprehensiveness of prediction of malignant biological behavior of meningioma.

Multi-dimensional radiomics analysis to predict visceral pleural invasion in lung adenocarcinoma of ≤3 cm maximum diameter

To explore the value of radiomics analysis in preoperatively predicting visceral pleural invasion (VPI) of lung adenocarcinoma (LAC) with ≤3 cm maximum diameter and to compare the performance of two-dimensional (2D) and three-dimensional (3D) computed tomography (CT) radiomics models. A total of 391 LAC patients were enrolled retrospectively, of whom 142 were VPI (+) and 249 were VPI (-). Radiomics features were extracted from 2D and 3D regions of interest (ROIs) of tumours in CT images. 2D and 3D radiomics models were developed combining the optimal radiomics features by using the logistic regression machine-learning method and radiomics scores (rad-scores) were calculated. Nomograms were constructed by integrating independent risk factors and rad-scores. The performance of each model was evaluated by using the receiver operator characteristic (ROC) curve, decision curve analysis (DCA), clinical impact curve (CIC), and calculating the area under the curve (AUC). There was no difference in the VPI prediction between 2D and 3D radiomics models (training group: 2D AUC=0.835, 3D AUC=0.836, p=0.896; validation group: 2D AUC=0.803, 3D AUC=0.794, p=0.567). The 2D and 3D nomograms performed similarly regarding discrimination (training group: 2D AUC=0.867, 3D AUC=0.862, p=0.409, validation group: 2D AUC=0.835, 3D AUC=0.827, p=0.558), and outperformed their corresponding radiomics models and the clinical model. DCA and CIC revealed that the 2D nomogram had slightly better clinical utility. The 2D radiomics model has a similar discrimination capability compared with the 3D radiomics model. The 2D nomogram performs slightly better for individual VPI prediction in LAC.

Ensemble method employing quantitative biomarkers and deep learning approach from structural magnetic resonance imaging to diagnose Alzheimer’s disease

Abstract Introduction: Alzheimer’s disease (AD) is a neurodegenerative condition that impairs activities of daily living and sharply declines gross cognitive ability. Over 152 million individuals worldwide will live with the dreaded consequence of a longer lifespan by the year 2050, making it a pressing public health issue. Magnetic resonance imaging (MRI) provides excellent soft tissue contrast and helps image the brain in vivo, non-invasively. Aims and Objectives: To summarize AD’s anatomical, physiological, and pathophysiological changes and derivation of quantifiable biomarkers from MRI to develop artificial intelligence (AI) based computer-aided detection (CAD) system to classify subjects among AD, mild cognitive impairment (MCI), and cognitively normal (CN). Materials and Methods: This retrospective study uses clinical and standardized, pre-processed, quality-controlled, and quality-checked—structural MRI imaging (diagnosed/labeled) data of 1069 subjects, age, gender, and class matched, taken from Alzheimer’s disease neuroimaging initiative. A pipeline is developed to get quantified biomarkers from the assessment of (1) cortical thickness, (2) volumetric segmentation for whole brain volumes, and (3) region of interest (ROI) areas most affected in AD. A gradient boosting method is used to predict class labels. The second approach implements a convolution neural network (CNN) model comprising 3D ROI. Results: Implemented CAD system using an ensemble gradient boosting approach has demonstrated good receiver operating characteristics characteristic and yielded balanced accuracy (BA) of 82.31%, 78.52%, and 72.73%, and the CNN approach has given better results 88.44%, 82.96%, and 74.34% for classification task AD versus CN, AD versus MCI, and MCI versus CN, respectively. Conclusion: This study has used a substantially large dataset of 1069 subjects. The deep learning-based efficient and optimal CNN model has used significantly large ROI-based 3-Dimentional volume, resulting in impressive performance improvements over comparable methods. The CNN model had given higher accuracy (6.13% for AD vs. CN, 4.44% for AD vs. MCI and 1.61% for MCI vs. CN) over gradient boosting, as the model uses significantly large ROI-based 3D brain volume and an inherent capability of it in learning most discriminative features automatically. However, quantitative biomarkers derived from brain morphometry, which accesses structural changes, yield reasonable estimates over pathophysiological alterations across the brain and augment a clinician with insightful and a holistic view, resulting in higher confidence over predicated class label by CNN and is a step closer to explainable AI. Accuracy for MCI versus CN drops as these classes share similar features and characteristics and can be improved by integrating biomarkers from other MRI modalities.

Same-day repeatability and Between-Sequence reproducibility of Mean ADC in PI-RADS lesions

PurposeThis study aimed to assess repeatability after repositioning (inter-scan), intra-rater, inter-rater and inter-sequence variability of mean apparent diffusion coefficient (ADC) measurements in MRI-detected prostate lesions. MethodForty-three patients with suspicion for prostate cancer were included and received a clinical prostate bi-/multiparametric MRI examination with repeat scans of the T2-weighted and two DWI-weighted sequences (ssEPI and rsEPI). Two raters (R1 and R2) performed single-slice, 2D regions of interest (2D-ROIs) and 3D-segmentation-ROIs (3D-ROIs). Mean bias, corresponding limits of agreement (LoA), mean absolute difference, within-subject coefficient of variation (CoV) and repeatability/reproducibility coefficient (RC/RDC) were calculated. Bradley & Blackwood test was used for variance comparison. Linear mixed models (LMM) were used to account for multiple lesions per patient. ResultsInter-scan repeatability, intra-rater and inter-sequence reproducibility analysis of ADC showed no significant bias. 3D-ROIs demonstrated significantly less variability than 2D-ROIs (p < 0.01). Inter-rater comparison demonstrated small significant systematic bias of 57 × 10-6 mm2/s for 3D-ROIs (p < 0.001). Intra-rater RC, with the lowest variation, was 145 and 189 × 10-6 mm2/s for 3D- and 2D-ROIs, respectively. For 3D-ROIs of ssEPI, RCs and RDCs were 190–198 × 10-6 mm2/s for inter-scan, inter-rater and inter-sequence variation. No significant differences were found for inter-scan, inter-rater and inter-sequence variability. ConclusionsIn a single-scanner setting, single-slice ADC measurements showed considerable variation, which may be lowered using 3D-ROIs. For 3D-ROIs, we propose a cut-off of ∼ 200 × 10-6 mm2/s for differences introduced by repositioning, rater or sequence effects. The results suggest that follow-up measurements should be possible by different raters or sequences.

Novel use of alternate (Alt) response (Rp) criteria (Cr) for early prediction of outcomes in pancreatic (P) neuroendocrine tumors (NETs): Utilizing banked imaging data from the ECOG-ACRIN E2211 study.

4133 Background: The evaluation of treatment (Tx) Rp in NETs using CT/MRI scans can be difficult. Previous studies, including the E2211, have shown improved progression-free survival (PFS) but no significant difference in Rp as measured by RECIST 1.1 (R1.1) Cr. Thus, making it difficult to determine Tx effectiveness with short-term imaging, further complicated by differences in imaging protocols and inter-reader variability. Incorporating tumor density, using smaller threshold changes in tumor size, and novel quantitative features may be more sensitive and precise than R1.1 (e.g. CHOI in GIST) but they have not been studied for NETs. Methods: Banked CT images from the E2211 trial were repurposed to study novel Cr. Three radiologists provided a R1.1 assessment. Density and radiomic features were calculated from a 2D region of interest in the portal venous phase (as per CHOI Cr). Patients were classified as responders (PR), stable disease (SD), or progressive disease (PD) based on R1.1 and CHOI and compared with PFS to determine the best predictor. Radiomic features were analyzed using pyradiomics. Wilcoxon tests were used to compare scan quality by center type, agreement was measured using Cohen’s Kappa, and predictive value was quantified using the c-statistic and AUC for time-varying outcomes. Results: 67 patients had their scans repurposed for the study. Inter-reader agreement for overall imaging studies was fair, with only 5 of 9 PD cases noted by reviewer 1 being agreed upon by reviewer 2 (proportional agreement for PD = 0.56) and a Kappa of 0.6 (95% CI 0.41 – 0.79). The CHOI Cr showed improved prediction of PFS compared to R1.1 (12-month AUC 0.75 for CHOI vs 0.69 for R1.1, c-statistic 0.66 vs 0.64, p=0.22), and was able to predict improved PFS for PR versus SD at the first interval scan. This was not seen with R1.1 (table). Radiomics and texture analysis as a potential adjunct to CHOI is being studied and results will be presented. Conclusions: This study using repurposed banked imaging from an NCTN trial, found that there was significant variability in the interpretation of images, and that the R1.1 Cr had a poor correlation with Rp in PNETs. The CHOI Cr emerged as a better alternative as it may have a better correlation with PFS and the ability to predict outcomes at the point of first disease assessment. Importantly, CHOI Cr can be easily applied in clinical practice to inform treatment decisions. Further research is needed to confirm these findings and define the therapeutic importance. [Table: see text]