Inter-reader Variability Research Articles

BIOM-54. AI-DRIVEN RISK-OF-PROGRESSION (AIRIP) CLASSIFIER FOR DISTINGUISHING RECURRENT BRAIN METASTASES FROM RADIATION TREATMENT EFFECT: A MULTI-INSTITUTIONAL COMPARATIVE STUDY WITH ADVANCED MULTIMODAL IMAGING

Abstract Following radiation therapy, a significant challenge in brain metastases (BM) management is differentiating radiation-induced-treatment effect (TrE) from tumor recurrence (TuR). TrE can be indistinguishable from TuR using conventional MRI. Advanced imaging techniques (e.g., perfusion MRI, PET/MRI) are not consistently used, and the standardized Response Assessment in Neuro-Oncology for brain metastases (RANO-BM) is sensitive to inter-reader variability. The performance of an artificial intelligence (AI)-driven risk-of-progression (AiRiP) classifier, which has been shown to capture pathophysiologic differences between TrE and TuR on routine MRI, was compared to that of clinical assessments and advanced imaging methods, in a multi-institutional setting. A total of n=261 lesions with pathologically-confirmed diagnoses in n=189 patients were analyzed. 115 lesions (73 TuR, 42 TrE) from site 1, 86 lesions (38 TuR, 48 TrE) from site 2, and 60 lesions (33 TuR, 27 TrE) from site 3 were used for training and testing the AiRiP-model. Gd-T1w, T2w, FLAIR MRI were preprocessed, and lesions were segmented by experts. Texture features (n=856) were extracted from each lesion. Random-forest classifier was employed for 3-fold cross-validation. Top-performing AiRiP-features, RANO-BM criteria, perfusion MRI and PET/MRI were evaluated in a sub-group analysis. For n=51 lesions on the test-set (site 3), 14 were classified as stable disease and 37 as TuR using RANO-BM (accuracy=54.1%). AiRiP-model achieved an accuracy of 76.5% on the same test-set and accurately classified 78.6% of the stable lesions as TrE or TuR. For another subset of lesions (n=27) on the same test-set, perfusion MRI and AiRiP-model achieved an accuracy of 59.3% and 70.4%, respectively. Lastly, for a subset of lesions (n=35) on the test-set (site 2), multimodal (perfusion, PET) imaging and AiRiP-model accurately classified 60% and 74.3% of lesions, respectively. 15 lesions were considered indeterminate via multimodal imaging, 73.3% of which AiRiP-model accurately classified as TrE or TuR. Our results suggest AI-driven models on conventional MRI may reliably distinguish TuR from TrE.

Factors of interobserver variability in prostate tumor MRI delineation: impact of PI-QUAL score.

Prostate cancer ranks as the second most common cancer in men worldwide. Dose escalation to the tumor and/or the prostate improves biochemical recurrence-free survival. However, interobserver variability in lesion contouring poses a significant limitation to such therapeutic approaches. Therefore, a study of factors influencing this variability is necessary. Three independent readers delineated the index prostate lesion (DIL) using T2w and ADC sequences for each patient. Clinical data were retrospectively collected for all participants. Logistic regression analysis was employed to examine the correlation between clinical features and a mean DICE coefficient > 0.7. Features exhibiting a p value < 0.05 on univariate analysis were subjected to multivariate analysis. The study comprised 68 patients, with a median DICE coefficient of 0.69 (95% CI 0.65-0.71), wherein 42.6% (29/68) attained a mean DICE > 0.7. Univariate analysis identified the PI-QUAL score, maximum diameter of DIL, and mean DIL volume as significant (p < 0.05) predictors. In multivariate analysis, only the PI-QUAL score (p = 0.008) remained statistically associated with a DICE coefficient > 0.7. The PI-QUAL score emerges as the primary predictive factor for minimizing inter-reader variability in intraprostatic dominant lesion segmentation. These findings underscore the importance of considering PI-QUAL scores when devising focal treatment plans. Adoption of a multi-reader approach involving diverse medical specialists (radiologists, radiotherapists, urologists) is advocated, particularly for MRIs with low PI-QUAL scores. Radiotherapy is a major treatment for patients with localized prostate cancer. Dose escalation to the tumor leads to improved cancer control. Precise delineation of the dominant intraprostatic lesion (DIL) remains a limitation to focal treatments. Features influencing inter-reader variability were never evaluated. In this study, we identified that the PI-QUAL score was the sole predictor of the inter-reader delineation variability of the DIL.

Reproducibility of Cardiac Multifrequency MR Elastography in Assessing Left Ventricular Stiffness and Viscosity.

Cardiac magnetic resonance elastography (MRE) shows promise in assessing the mechanofunctional properties of the heart but faces clinical challenges, mainly synchronization with cardiac cycle, breathing, and external harmonic stimulation. To determine the reproducibility of in vivo cardiac multifrequency MRE (MMRE) for assessing diastolic left ventricular (LV) stiffness and viscosity. Prospective. This single-center study included a total of 28 participants (mean age, 56.6 ± 23.0 years; 16 male) consisting of randomly selected healthy participants (mean age, 44.6 ± 20.1 years; 9 male) and patients with aortic stenosis (mean age, 78.3 ± 3.8 years; 7 male). 3 T, 3D multifrequency MRE with a single-shot spin-echo planar imaging sequence. Each participant underwent two cardiac MMRE examinations on the same day. Full 3D wave fields were acquired in diastole at frequencies of 80, 90, and 100 Hz during a total of three breath-holds. Shear wave speed (SWS) and penetration rate (PR) were reconstructed as a surrogate for tissue stiffness and inverse viscous loss. Epicardial and endocardial ROIs were manually drawn by two independent readers to segment the LV myocardium. Shapiro-Wilk test, Bland-Altman analysis and intraclass correlation coefficient (ICC). P-value <0.05 were considered statistically significant. Bland-Altman analyses and intraclass correlation coefficients (ICC = 0.96 for myocardial stiffness and ICC = 0.93 for viscosity) indicated near-perfect test-retest repeatability among examinations on the same day. The mean SWS for scan and re-scan diastolic LV myocardium were 2.42 ± 0.24 m/s and 2.39 ± 0.23 m/s; the mean PR were 1.24 ± 0.17 m/s and 1.22 ± 0.14 m/s. Inter-reader variability showed good to excellent agreement for myocardial stiffness (ICC = 0.92) and viscosity (ICC = 0.85). Cardiac MMRE is a promising and reproducible method for noninvasive assessment of diastolic LV stiffness and viscosity. 2 TECHNICAL EFFICACY: 1.

ProstateZones – Segmentations of the prostatic zones and urethra for the PROSTATEx dataset

Manual segmentations are considered the gold standard for ground truth in machine learning applications. Such tasks are tedious and time-consuming, albeit necessary to train reliable models. In this work, we present a dataset with expert segmentations of the prostatic zones and urethra for 200 randomly selected patients from the PROSTATEx dataset. Notably, independent duplicate segmentations were performed for 40 patients, providing inter-reader variability data. This results in a total of 240 segmentations. This dataset can be used to train machine learning models or serve as an external test set for evaluating models trained on private data, thereby addressing a current gap in the field. The delineated structures and terminology adhere to the latest Prostate Imaging Reporting and Data Systems v2.1 guidelines, ensuring consistency.

Detectability of Hypoattenuating Liver Lesions with Deep Learning CT Reconstruction: A Phantom and Patient Study.

Background CT deep learning image reconstruction (DLIR) improves image quality by reducing noise compared with adaptive statistical iterative reconstruction-V (ASIR-V). However, objective assessment of low-contrast lesion detectability is lacking. Purpose To investigate low-contrast detectability of hypoattenuating liver lesions on CT scans reconstructed with DLIR compared with CT scans reconstructed with ASIR-V in a patient and a phantom study. Materials and Methods This single-center retrospective study included patients undergoing portal venous phase abdominal CT between February and May 2021 and a low-contrast-resolution phantom scanned with the same protocol. Four reconstructions (ASIR-V at 40% strength [ASIR-V 40] and DLIR at three strengths) were generated. Five radiologists qualitatively assessed the images using the five-point Likert scale for image quality, lesion diagnostic confidence, conspicuity, and small lesion (≤1 cm) visibility. Up to two key lesions per patient, confirmed at histopathologic testing or at prior or follow-up imaging studies, were included. Lesion-to-background contrast-to-noise ratio was calculated. Interreader variability was analyzed. Intergroup qualitative and quantitative metrics were compared between DLIR and ASIR-V 40 using proportional odds logistic regression models. Results Eighty-six liver lesions (mean size, 15 mm ± 9.5 [SD]) in 50 patients (median age, 62 years [IQR, 57-73 years]; 27 [54%] female patients) were included. Differences were not detected for various qualitative low-contrast detectability metrics between ASIR-V 40 and DLIR (P > .05). Quantitatively, medium-strength DLIR and high-strength DLIR yielded higher lesion-to-background contrast-to-noise ratios than ASIR-V 40 (medium-strength DLIR vs ASIR-V 40: odds ratio [OR], 1.96 [95% CI: 1.65, 2.33]; high-strength DLIR vs ASIR-V 40: OR, 5.36 [95% CI: 3.68, 7.82]; P < .001). Low-contrast lesion attenuation was reduced by 2.8-3.6 HU with DLIR. Interreader agreement was moderate to very good for the qualitative metrics. Subgroup analysis based on lesion size of larger than 1 cm and 1 cm or smaller yielded similar results (P > .05). Qualitatively, phantom study results were similar to those in patients (P > .05). Conclusion The detectability of low-contrast liver lesions was similar on CT scans reconstructed with low-, medium-, and high-strength DLIR and ASIR-V 40 in both patient and phantom studies. Lesion-to-background contrast-to-noise ratios were higher for DLIR medium- and high-strength reconstructions compared with ASIR-V 40. © RSNA, 2024 Supplemental material is available for this article.

Deep-learning model accurately classifies multi-label lung ultrasound findings, enhancing diagnostic accuracy and inter-reader agreement.

Despite the increasing use of lung ultrasound (LUS) in the evaluation of respiratory disease, operators' competence constrains its effectiveness. We developed a deep-learning (DL) model for multi-label classification using LUS and validated its performance and efficacy on inter-reader variability. We retrospectively collected LUS and labeled as normal, B-line, consolidation, and effusion from patients undergoing thoracentesis at a tertiary institution between January 2018 and January 2022. The development and internal testing involved 7580 images fromJanuary 2018and December2020, andthe model's performance was validated on a temporally separated test set (n = 985 imagescollected after January 2021) and two external test sets (n = 319 and 54 images). Two radiologists interpreted LUS with and without DL assistance and compared diagnostic performance and agreement. The model demonstrated robust performancewith AUCs: 0.93 (95% CI 0.92-0.94) for normal, 0.87 (95% CI 0.84-0.89) for B-line, 0.82 (95% CI 0.78-0.86) for consolidation, and 0.94 (95% CI 0.93-0.95) for effusion.The model improved reader accuracy for binary discrimination (normal vs. abnormal; reader 1: 87.5-95.6%, p = 0.004;reader 2: 95.0-97.5%, p = 0.19), and agreement(k = 0.73-0.83, p = 0.01). In conclusion,the DL-based model may assist interpretation, improving accuracy and overcoming operator competence limitationsin LUS.

Clinical feasibility of deep learning–accelerated single-shot turbo spin echo sequence with enhanced denoising for pancreas MRI at 3 Tesla

PurposeTo assess the feasibility of the single-shot turbo spin echo sequence using deep learning-based reconstruction (DLR) (HASTEDL) with enhanced denoising for pancreas MRI. MethodsPatients who underwent pancreas MRI from March to April 2021 were included. Four T2-weighted images (non-FS conventional HASTE vs. HASTEDL with enhanced denoising and FS HASTEDL with enhanced denoising vs. HASTEDL) were acquired. Two readers independently assessed the image quality parameters of the two non-FS image sets using a 4-point Likert scale. The signal-to-noise ratio (SNR) of the cystic lesions and pancreatic parenchyma and the contrast-to-noise ratio between the cystic lesion and pancreatic parenchyma were calculated for all four image sets. The size of the largest cystic lesion and the diameter of pancreatic duct were measured. ResultsA total of 63 patients were included, 48 (76.2 %) of whom had 136 pancreatic cystic lesion(s). The acquisition times of conventional HASTE and HASTEDL were 69 and 18 sec, respectively. All image quality parameters except artifacts for reader 2 were significantly better for HASTEDL with enhanced denoising. Those images also received scores for overall image quality that were significantly higher than those for the conventional HASTE (3.26 ± 0.54 vs. 2.47 ± 0.56, p < 0.001). The SNR of the pancreatic cystic lesion and pancreatic parenchyma was significantly higher in the HASTEDL with enhanced denoising (p < 0.001 for both). Inter-reader variability for measuring the pancreatic cyst size (ICC, 0.999 vs. 0.995; 95 % LoA, −0.13481 to 0.14743 vs. −0.24097 to 0.27404) and duct diameter (ICC, 0.994 vs. 0.969; 95 % LoA, −0.11684 to 0.36026 vs. −0.45544 to 0.44664) was lower in HASTEDL with enhanced denoising than in the conventional HASTE. ConclusionHASTEDL with enhanced denoising could be useful for reducing the acquisition time of pancreas MRI while improving the image quality for the evaluation of pancreatic cystic lesions.

Artificial intelligence-based quantification of pulmonary HRCT (AIqpHRCT) for the evaluation of interstitial lung disease in patients with inflammatory rheumatic diseases

High-resolution computed tomography (HRCT) is important for diagnosing interstitial lung disease (ILD) in inflammatory rheumatic disease (IRD) patients. However, visual ILD assessment via HRCT often has high inter-reader variability. Artificial intelligence (AI)-based techniques for quantitative image analysis promise more accurate diagnostic and prognostic information. This study evaluated the reliability of artificial intelligence-based quantification of pulmonary HRCT (AIqpHRCT) in IRD-ILD patients and verified IRD-ILD quantification using AIqpHRCT in the clinical setting. Reproducibility of AIqpHRCT was verified for each typical HRCT pattern (ground-glass opacity [GGO], non-specific interstitial pneumonia [NSIP], usual interstitial pneumonia [UIP], granuloma). Additional, 50 HRCT datasets from 50 IRD-ILD patients using AIqpHRCT were analysed and correlated with clinical data and pulmonary lung function parameters. AIqpHRCT presented 100% agreement (coefficient of variation = 0.00%, intraclass correlation coefficient = 1.000) regarding the detection of the different HRCT pattern. Furthermore, AIqpHRCT data showed an increase of ILD from 10.7 ± 28.3% (median = 1.3%) in GGO to 18.9 ± 12.4% (median = 18.0%) in UIP pattern. The extent of fibrosis negatively correlated with FVC (ρ=-0.501), TLC (ρ=-0.622), and DLCO (ρ=-0.693) (p < 0.001). GGO measured by AIqpHRCT also significant negatively correlated with DLCO (ρ=-0.699), TLC (ρ=-0.580) and FVC (ρ=-0.423). For the first time, the study demonstrates that AIpqHRCT provides a highly reliable method for quantifying lung parenchymal changes in HRCT images of IRD-ILD patients. Further, the AIqpHRCT method revealed significant correlations between the extent of ILD and lung function parameters. This highlights the potential of AIpqHRCT in enhancing the accuracy of ILD diagnosis and prognosis in clinical settings, ultimately improving patient management and outcomes.

Three-Dimensional Transthoracic Echocardiography for Semiautomated Analysis of the Tricuspid Annulus: Validation and Normal Values

The expansion of tricuspid valve (TV) interventions has underscored the need for accurate and reproducible three-dimensional (3D) transthoracic echocardiographic (TTE) tools for evaluating the tricuspid annulus and for 3D normal values of this structure. The aims of this study were to develop new semi-automated software for 3D TTE analysis of the tricuspid annulus, compare its accuracy and reproducibility against those of multiplanar reconstruction (MPR) reference, and determine normative values. Three-dimensional TTE images of 113 patients with variable degrees of tricuspid regurgitation were analyzed using the new semiautomated software and conventional MPR methodology (as the reference standard), each by three independent readers. For each measured parameter, intertechnique agreement was assessed using linear regression and Bland-Altman analyses, and interreader variability using intraclass correlation coefficients and coefficients of variation. Additionally, 3D TTE data sets of 789 subjects from the WASE (World Alliance Societies of Echocardiography) study were analyzed using this new software to determine normal values for each tricuspid annular (TA) parameter. Semiautomated measurements showed excellent agreement with MPR reference values for all TA measurements: high correlations (all r values >0.8) and minimal biases. All measurements were more reproducible than with MPR: higher intraclass correlation coefficients (0.94-0.96 vs 0.82-0.90) and lower coefficients of variation (5%-12% vs 8%-18%). Sex- and age-related differences were identified in 3D normal values of most TA parameters. Dynamic analysis showed that TA parameters vary throughout the cardiac cycle, reaching minimal values at end-systole and maximum values in late diastole. Novel software for semiautomated analysis of TA geometry and dynamics provides accurate and reproducible measurements. Normal values of TA dimensions, parsed by sex and age, are reported.

Use of reporting templates for chest radiographs in a coronavirus disease 2019 context: measuring concordance of radiologists with three international templates.

Reporting templates for chest radiographs (CXRs) for patients presenting or being clinically managed for severe acute respiratory syndrome coronavirus 2 [coronavirus disease 2019 (COVID-19)] has attracted advocacy from international radiology societies. We aim to explore the effectiveness and useability of three international templates through the concordance of, and between, radiologists reporting on the presence and severity of COVID-19 on CXRs. Seventy CXRs were obtained from a referral hospital, 50 from patients with COVID-19 (30 rated "classic" COVID-19 appearance and 20 "indeterminate") and 10 "normal" and 10 "alternative pathology" CXRs. The recruited radiologists were assigned to three test sets with the same CXRs but with different template orders. Each radiologist read their test set three times and assigned a classification to the CXR using the Royal Australian New Zealand College of Radiology (RANZCR), British Society of Thoracic Imaging (BSTI), and Modified COVID-19 Reporting and Data System (Dutch; mCO-RADS) templates. Inter-reader variability and intra-reader variability were measured using Fleiss' kappa coefficient. Twelve Australian radiologists participated. The BSTI template had the highest inter-reader agreement (0.46; "moderate" agreement), followed by RANZCR (0.45) and mCO-RADS (0.32). Concordance was driven by strong agreement in "normal" and "alternative" classifications and was lowest for "indeterminate." General consistency was observed across classifications and templates, with intra-reader variability ranging from "good" to "very good" for COVID-19 CXRs (0.61), "normal" CXRs (0.76), and "alternative" (0.68). Reporting templates may be useful in reducing variation among radiology reports, with intra-reader variability showing promise. Feasibility and implementation require a wider approach including referring and treating doctors plus the development of training packages for radiologists specific to the template being used.

Intraindividual reproducibility of myocardial radiomic features between energy-integrating detector and photon-counting detector CT angiography

BackgroundRadiomics is not yet used in clinical practice due to concerns regarding its susceptibility to technical factors. We aimed to assess the stability and interscan and interreader reproducibility of myocardial radiomic features between energy-integrating detector computed tomography (EID-CT) and photon-counting detector CT (PCD-CT) in patients undergoing coronary CT angiography (CCTA) on both systems.MethodsConsecutive patients undergoing clinically indicated CCTA on an EID-CT were prospectively enrolled for a PCD-CT CCTA within 30 days. Virtual monoenergetic images (VMI) at various keV levels and polychromatic images (T3D) were generated for PCD-CT, with image reconstruction parameters standardized between scans. Two readers performed myocardial segmentation and 110 radiomic features were compared intraindividually between EID-CT and PDC-CT series. The agreement of parameters was assessed using the intraclass correlation coefficient and paired t-test for the stability of the parameters.ResultsEighteen patients (15 males) aged 67.6 ± 9.7 years (mean ± standard deviation) were included. Besides polychromatic PCD-CT reconstructions, 60- and 70-keV VMIs showed the highest feature stability compared to EID-CT (96%, 90%, and 92%, respectively). The interscan reproducibility of features was moderate even in the most favorable comparisons (median ICC 0.50 [interquartile range 0.20–0.60] for T3D; 0.56 [0.33–0.74] for 60 keV; 0.50 [0.36–0.62] for 70 keV). Interreader reproducibility was excellent for the PCD-CT series and good for EID-CT segmentations.ConclusionMost myocardial radiomic features remain stable between EID-CT and PCD-CT. While features demonstrated moderate reproducibility between scanners, technological advances associated with PCD-CT may lead to greater reproducibility, potentially expediting future standardization efforts.Relevance statementWhile the use of PCD-CT may facilitate reduced interreader variability in radiomics analysis, the observed interscanner variations in comparison to EID-CT should be taken into account in future research, with efforts being made to minimize their impact in future radiomics studies.Key PointsMost myocardial radiomic features resulted in being stable between EID-CT and PCD-CT on certain VMIs.The reproducibility of parameters between detector technologies was limited.PCD-CT improved interreader reproducibility of myocardial radiomic features.Graphical

Comparison of best landmarks for calculating fetal jaw measurements by ultrasound and MRI in micrognathia.

Micrognathia can be diagnosed in utero with ultrasound by measuring the jaw index and/or inferior facial angle, though it can be challenging due to fetal positioning. The jaw index can be measured with magnetic resonance imaging (MRI) using the masseter muscle, but indistinct margins can lead to inaccuracy; the easily visualized posterior teeth buds may be a better landmark. We aimed to evaluate inter-reader variability, agreement with ultrasound, and association with postnatal outcomes using MRI to measure the inferior facial angle, jaw index by masseter muscle, and jaw index by posterior teeth buds. A single-institution retrospective review was performed of singleton pregnancies with prenatally diagnosed micrognathia by ultrasound or MRI from September 2013-June 2022. Ultrasound measurements were obtained by a maternal-fetal medicine specialist and MRI measurements by two radiologists to evaluate inter-reader variability. Intraclass correlation coefficients (ICC) and Bland-Altman analysis were used to assess agreement between imaging methods and logistic regressions and ROC curves to assess associations with postnatal outcomes. Forty-three fetuses (median gestational age 26weeks (IQR 22-31); 47% male (20/43)) were included. Ultrasound measurements could not be obtained for jaw index in 15/43 (35%) fetuses and inferior facial angle in 11/43 (26%); MRI measurements were obtained by at least one reader in all cases. Jaw index by teeth buds demonstrated lowest inter-reader variability (ICC = 0.82, P < 0.001) and highest agreement with ultrasound (bias -0.23, 95% CI -2.8-2.2). All MRI measurements, but not ultrasound, predicted need for mandibular distraction (inferior facial angle P = 0.02, jaw index by masseter muscle P = 0.04, jaw index by teeth buds P = 0.01). Fetal MRI measurements, particularly jaw index measured by posterior teeth buds, demonstrate low inter-reader variability and high agreement with ultrasound, and may predict need for mandibular distraction postnatally.

Interreader and Intermodality Variability in Macular Atrophy Quantification in Neovascular Age-related Macular Degeneration: Comparison of 6 Imaging Modalities

Macular atrophy is a common complication in neovascular age-related macular degeneration (AMD) and is associated with poorer visual outcomes. This study evaluated interreader and intermodality variability in measuring macular atrophy in previously treated neovascular AMD eyes without exudation using 6 imaging modalities. Prospective, cohort study. Thirty participants with previously treated neovascular AMD showing no signs of exudation at the time of enrollment and exhibiting macular atrophy. During the same clinic visit, patients were imaged using 6 different imaging modalities: color fundus photography (CFP; Clarus, Carl Zeiss Meditec), near-infrared imaging (NIR; Spectralis, Heidelberg Engineering), structural OCT (Spectralis, Heidelberg Engineering), green fundus autofluorescence (GAF; Clarus, Carl Zeiss Meditec), blue fundus autofluorescence (BAF; Spectralis, Heidelberg Engineering), and pseudocolor imaging (MultiColor; Spectralis, Heidelberg Engineering). Two readers independently measured the macular atrophy area. Interreader and intermodality agreement. The 95% coefficient of repeatability was 5.98 mm2 for CFP, 4.46 mm2 for MultiColor, 3.90 mm2 for BAF, 3.92 mm2 for GAF, 4.86 mm2 for NIR, and 3.55 mm2 for OCT. Similarly, the coefficient of variation was lowest for OCT-based grading at 0.08 and highest for NIR-based grading at 0.28. Accordingly, the intraclass correlation coefficient was 0.742 for CFP, 0.805 for MultiColor, 0.857 for BAF, 0.850 for GAF, 0.755 for NIR, and 0.917 for OCT. The 6 different imaging modalities presented measurements with different mean values, with only a limited number of comparisons not significantly different between the instruments, although measurements were correlated. The largest size of macular atrophy was measured with structural OCT-based grading (median= 4.65 mm2; interquartile range [IQR]= 4.78 mm2) and the smallest was with CFP-based grading (median= 3.86 mm2; IQR= 5.06 mm2). Inconsistencies arose from various factors. In patients with neovascular AMD, macular atrophy measurements vary significantly depending on the imaging technique used. Color fundus photography-based assessments yielded the smallest macular atrophy sizes, whereas structural OCT-based assessments produced the largest. These discrepancies stem from both the inherent limitations of each modality in assessing retinal pigment epithelial atrophy and factors related to neovascularization, such as the coexistence of fibrosis. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Advances in multiparametric magnetic resonance imaging combined with biomarkers for the diagnosis of high-grade prostate cancer.

Clinical decisions based on the test results for prostate-specific antigen often result in overdiagnosis and overtreatment. Multiparametric magnetic resonance imaging (mpMRI) can be used to identify high-grade prostate cancer (HGPCa; Gleason score ≥3 + 4); however, certain limitations remain such as inter-reader variability and false negatives. The combination of mpMRI and prostate cancer (PCa) biomarkers (prostate-specific antigen density, Proclarix, TMPRSS2:ERG gene fusion, Michigan prostate score, ExoDX prostate intelliscore, four kallikrein score, select molecular diagnosis, prostate health index, and prostate health index density) demonstrates high accuracy in the diagnosis of HGPCa, ensuring that patients avoid unnecessary prostate biopsies with a low leakage rate. This manuscript describes the characteristics and diagnostic performance of each biomarker alone and in combination with mpMRI, with the intension to provide a basis for decision-making in the diagnosis and treatment of HGPCa. Additionally, we explored the applicability of the combination protocol to the Asian population.

Influence of Gadolinium-based Contrast Media and Inter-reader Variation on the Estimation of Intravoxel Incoherent Motion (IVIM) Parameters in Breast MR Imaging.

Gadolinium-based contrast media (GBCM) may affect apparent diffusion coefficient measurements on diffusion-weighted imaging. We aimed at investigating the effect of GBCM and inter-reader variation on intravoxel incoherent motion (IVIM) parameters in breast lesions. A total of 89 patients referred to 3T breast MRI with at least one histologically verified lesion were included. IVIM data were acquired using a single-shot echo planar imaging sequence before and after GBCM administration. D (true diffusion coefficient), D* (pseudo-diffusion coefficient) and f (perfusion fraction) were calculated and measured by two readers (R1, R2). Inter-reader and intra-reader agreements were assessed by intraclass correlation coefficients (ICCs) and Bland-Altman plots. D was comparable before and after GBCM administration and between readers. D* and f decreased after GBCM administration and showed a lower agreement between readers. Intra-reader agreement before and after GBCM administration was almost perfect for D for both R1 and R2 (ICC 0.955 and 0.887). The intra-reader agreement was substantial to moderate for D* (ICC R1 0.708, R2 0.583) and moderate for f (ICC R1 0.529 and R2 0.425). Inter-reader agreement before GBCM administration was almost perfect for D (ICC 0.905), substantial for D* (ICC 0.733), and moderate for f (ICC 0.404); after contrast media administration, it was almost perfect for D (ICC 0.876) and substantial for D* (ICC 0.654) and f (ICC 0.606). Bland-Altman plots revealed no significant bias. Administration of GBCM seems to have a stronger effect on D* and f values than on D values. This should be considered when applying IVIM in clinical practice.

Diagnostic accuracy and reliability of CT-based Node-RADS for colon cancer.

The Node-RADS classification was recently published as a classification system to better characterize lymph nodes in oncological imaging. The present analysis investigated the diagnostic benefit of the Node-RADS classification of staging computed tomography (CT) images to categorize and stage lymph nodes in patients with colon cancer. All patients were surgically resected and the lymph nodes were histopathological analyzed. All investigated lymph nodes were scored in accordance to the Node-RADS classification by two experienced radiologists. Interreader variability was assessed with Cohen's kappa analysis, discrimination analysis was performed with Mann-Whitney-U test and diagnostic accuracy was assessed with receiver-operating characteristics (ROC) curve analysis. Overall, 108 patients (n = 49 females, 45.3%) with a mean age of 70.08 ± 14.34 years were included. In discrimination analysis, the total Node-RADS score showed statistically significant differences between N- and N + stage (for reader 1: mean 1.89 ± 1.09 score for N- versus 2.93 ± 1.62 score for N+, for reader 2: 1.33 ± 0.48 score for N- versus 3.65 ± 0.94 score for N+, p = 0.001, respectively). ROC curve analysis for lymph node discrimination showed an area under the curve of 0.68. A threshold value of 2 resulted in a sensitivity of 0.62 and a specificity of 0.71. Node-RADS score derived from staging CT shows only limited diagnostic accuracy to correctly predict nodal positivity in colon cancer. The interreader variability seems to be high and should question the clinical translation for this tumour entity.

Double reading performance and the impact of adjudication on progression-free survival estimations: Findings from a lung clinical trial.

e23010 Background: The FDA recommends Blinded Independent Central Review (BICR) with double read for clinical trials with imaging. However, inter-reader variability is a concern in these trials. While studies have investigated the variability of RECIST, the primary response criteria, little attention has been given to the optimal association of readers. The evaluation of therapeutic response in phase III trials relies on the Date of first Progressive Disease (DoPD), with the Discrepancy Rate (DR) as the preferred index for measuring inter-reader variability in DoPD evaluation. Another important index measures readers' bias, assessing their tendency to over or under-estimate diagnoses. In cases of discrepancies, a third reader is brought in for adjudication. However, the impact of adjudication on trials' Progression-Free Survival (PFS) is not well-documented. Our study examines the variability in a lung clinical trial using RECIST, analyzing double reading performance, reader association prediction, and the impact of adjudication on PFS estimations. Methods: We retrospectively analyzed five phase III lung clinical trials using the RECIST 1.1 criteria in BICR with double reads. The trials involved 7 readers organized into 11 teams, each reader having participated in multiple trials and performed over 50 reads, resulting in 1017 patients' reviews. Our analysis included: Calculation of DR and bias for each team, and testing the correlation between DR and bias. Computation of the signed bias for each individual reader. Evaluation of a probabilistic model to predict the DR for each team and the bias for each reader. Comparison of PFS between single and double reads after adjudication and endorsement of one of the readings. Results: A multiple comparisons test did not reveal any difference between teams’ DR (Marascuilo test; q &gt; 0.05). The average DR across all teams was 39.9% [95%CI; 37.8; 41.9]. However, we did find significant differences in bias when comparing 9/55 pairs of teams (Marascuilo test; q &lt; 0.05). The range of absolute bias values was 20% to 100%. We did not find a correlation between bias and DR (p = 0.64). Additionally, when comparing the average bias value per reader, no differences were observed (Marascuilo test; q &lt; 0.05). We failed to predict teams DR and readers' bias. The endorsement rate of readers ranged [18%; 82%]. After adjudication, we found that 27% of the PFS were lower than the minimum value obtained from the single readers, in one case 20.6% shorter. Conclusions: Significant readers' bias has a notable impact on double readings, independent of the DR values. The performance of one reader cannot be generalized based on others. Additionally, adjudication significantly affects the PFS of clinical trials. These findings emphasize the importance of considering readers' bias and the potential consequences of adjudication in clinical trial assessments.

Promptable foundation model for automatic whole body RECIST measurement.

e13643 Background: RECIST 1.1 is the gold standard for tumor evaluation in clinical trials and patient care. However, it faces challenges due to the subjective selection and measurement of lesions, notably because of inter-observer variability in the 20-30% range (1), which can potentially result in inaccuracies in therapeutic response classification (2). The main objective of this study is to assess the added value of a radiological foundation model to assist the radiologist to measure RECIST across different lesion topographies, using visual prompting. Methods: We use a promptable segmentation algorithm, based on a visual foundation model, pre-trained on various CT-scans and then trained on a subset of the Medical Segmentation Decathlon (MSD) dataset, covering 538 patients with pancreas, liver, colon and lung lesions. It outputs a 3D segmentation mask of the lesion, using a visual prompt, i.e. a 3D bounding box (BBox) as input. We then measure the capacity of the model to measure RECIST on the MSD validation set and on the KiTS validation set (99 samples), which features a new lesion type (kidney). To simulate inter-reader variability of up to 23%, we provide as inputs BBoxes centered on the lesions but with a varying error (15% variability: +10 pixels, 23% variability: +15 pixels). Results: The model is able to correct variability in BBox input: with all BBox error ranges, it beats the input variability (15% and 23% thresholds in each column) across many organ types, except in the pancreas when using large error BBoxes. Moreover, the model is able to generalize to new lesion types not seen during training, on an external dataset (KiTS - Kidney). Conclusions: Unlike existing supervised machine learning models dedicated to lesion detections in specific organs, our approach is organ and lesion agnostic and offers a more reliable, precise tumor assessment. This is a first step before allowing the longitudinal evaluation of tumors while safeguarding the clinical intuition necessary for selecting the right target lesions. Moreover, we believe that these methods will be crucial in advancing beyond RECIST 1.1, facilitating the identification of new prognostic biomarkers and proxies for tumor burden derived from previously underexplored radiomics features, ultimately refining the efficacy of clinical trial assessments and elevating the standard of patient care. 1. Yoon, S et al. 2015. 2. Kuhl et al. 2018. [Table: see text]

Explainable Precision Medicine in Breast MRI: A Combined Radiomics and Deep Learning Approach for the Classification of Contrast Agent Uptake.

In DCE-MRI, the degree of contrast uptake in normal fibroglandular tissue, i.e., background parenchymal enhancement (BPE), is a crucial biomarker linked to breast cancer risk and treatment outcome. In accordance with the Breast Imaging Reporting & Data System (BI-RADS), it should be visually classified into four classes. The susceptibility of such an assessment to inter-reader variability highlights the urgent need for a standardized classification algorithm. In this retrospective study, the first post-contrast subtraction images for 27 healthy female subjects were included. The BPE was classified slice-wise by two expert radiologists. The extraction of radiomic features from segmented BPE was followed by dataset splitting and dimensionality reduction. The latent representations were then utilized as inputs to a deep neural network classifying BPE into BI-RADS classes. The network's predictions were elucidated at the radiomic feature level with Shapley values. The deep neural network achieved a BPE classification accuracy of 84 ± 2% (p-value < 0.00001). Most of the misclassifications involved adjacent classes. Different radiomic features were decisive for the prediction of each BPE class underlying the complexity of the decision boundaries. A highly precise and explainable pipeline for BPE classification was achieved without user- or algorithm-dependent radiomic feature selection.

Improving assessment of lesions in longitudinal CT scans: a bi-institutional reader study on an AI-assisted registration and volumetric segmentation workflow

PurposeAI-assisted techniques for lesion registration and segmentation have the potential to make CT-based tumor follow-up assessment faster and less reader-dependent. However, empirical evidence on the advantages of AI-assisted volumetric segmentation for lymph node and soft tissue metastases in follow-up CT scans is lacking. The aim of this study was to assess the efficiency, quality, and inter-reader variability of an AI-assisted workflow for volumetric segmentation of lymph node and soft tissue metastases in follow-up CT scans. Three hypotheses were tested: (H1) Assessment time for follow-up lesion segmentation is reduced using an AI-assisted workflow. (H2) The quality of the AI-assisted segmentation is non-inferior to the quality of fully manual segmentation. (H3) The inter-reader variability of the resulting segmentations is reduced with AI assistance.Materials and methodsThe study retrospectively analyzed 126 lymph nodes and 135 soft tissue metastases from 55 patients with stage IV melanoma. Three radiologists from two institutions performed both AI-assisted and manual segmentation, and the results were statistically analyzed and compared to a manual segmentation reference standard.ResultsAI-assisted segmentation reduced user interaction time significantly by 33% (222 s vs. 336 s), achieved similar Dice scores (0.80–0.84 vs. 0.81–0.82) and decreased inter-reader variability (median Dice 0.85–1.0 vs. 0.80–0.82; ICC 0.84 vs. 0.80), compared to manual segmentation.ConclusionThe findings of this study support the use of AI-assisted registration and volumetric segmentation for lymph node and soft tissue metastases in follow-up CT scans. The AI-assisted workflow achieved significant time savings, similar segmentation quality, and reduced inter-reader variability compared to manual segmentation.