To determine intermodal agreement of AI-derived volumetric body composition metrics between [18 F]FDG-PET/CT-derived low-dose CT and whole-body MRI (wb-MRI). We retrospectively analyzed same-day whole-body low-dose CT, obtained from [18F]FDG-PET/CT, and wb-MRI from consecutive multiple myeloma patients from the prospective GMMG-HD7- and GMMG-HD8-/DSMM-XIX-trials. An automated deep learning pipeline based on the TotalSegmentator model, and its MRI variant were used to segment skeletal muscle (SM), subcutaneous- (SAT) and abdominal visceral adipose tissue (VAT) from CT and MRI, respectively. Volumetric measures and single-slice estimates were derived. Agreement between CT- and MRI-based metrics was assessed using Pearson r, intraclass correlation coefficients (ICC), and Bland-Altman analyses. Forty-two multiple myeloma patients with same-day PET/CT and wb-MRI were included. For AI-based volumetric body composition metrics, CT and MRI demonstrated high agreement (ICC > 0.95) across all SM, SAT and VAT. However, Bland-Altmann plots revealed systematic modality-related differences between MRI and CT: both SM (bias: -2582 ml, p < 0.0001) and VAT (bias: -211 ml, p < 0.0001) were significantly lower when extracted from MRI compared to CT, while for SAT, the difference was not statistically significant (bias: -263 ml, p = 0.1629). For AI-based single slice analysis, agreement remained high for SAT and VAT (ICC > 0.97), while being lower for SM (ICC 0.87). Bland-Altman plots showed that SM was systematically and significantly lower when obtained from MRI than from CT (bias: -18.4cm2, p < 0.0001). Automated body composition measures from (PET-)CT and wb-MRI are highly concordant but show consistent intermodal biases that should be considered when pooling multimodal data or applying modality-agnostic thresholds. Not applicable. Fully automated deep learning body composition analysis showed strong agreement between same-day low-dose CT from PET/CT and whole-body MRI in newly diagnosed multiple myeloma patients. However, MRI produced systematically lower skeletal muscle and visceral adipose tissue estimates than CT, indicating modality-dependent bias. 1. Whole-body volumetric quantification of skeletal muscle, subcutaneous adipose tissue, and abdominal visceral adipose tissue can be performed in a fully automated manner from both low-dose CT (from PET/CT) and whole-body MRI acquired on the same day. 2. While correlation was high, tissue-specific, modality-related biases were observed, indicating that values from CT and MRI are not directly interchangeable without accounting for systematic differences in real world clinical practice, particularly for skeletal muscle and visceral adipose tissue. 3. Segmentation quality was predominantly sufficient-to-good on radiologist review with the main error patterns were linked to imaging artifacts and challenging anatomy
This letter to the Editor addresses a fundamental inconsistency in the reporting of study design and validation status in the article by Zhou et al. published in Cancer Imaging. The authors state in the Methods section that the study retrospectively analyzed 125 patients, yet the Introduction and Discussion claim that the study prospectively validated the prognostic value of early tumour shrinkage (ETS) and depth of response (DoR). Prospective validation requires a priori model specification and evaluation on subsequently collected data, whereas retrospective analyses are inherently susceptible to selection bias and cannot establish prospective validity. The authors also conclude that they developed and validated a comprehensive predictive model, but no external validation, temporal validation, or clearly described internal validation procedure is reported. The ETS and DoR thresholds were derived from ROC analysis within the same cohort and then applied to the same patients for survival analysis, risking overfitting and optimistic bias. Additionally, the Results report “responders (n = 109) and non-responders (n = 12),” while the corresponding results in the original article by Zhou et al. [1] list the ratio as 12/109, reversing the group labels. We respectfully suggest that the authors clarify the actual study design, remove any claim of prospective validation if the study is retrospective, provide evidence of validation or remove the claim of a validated model, and correct the numerical inconsistency. Such corrections would uphold the scientific integrity of the literature.
To compare the diagnostic performance of diffusion-weighted imaging (DWI), diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), mean apparent propagator (MAP), and neurite orientation dispersion and density imaging (NODDI) in grading pediatric intracranial tumors (PITs). Forty-four patients with PITs (mean age 8.39 ± 4.17 years; 18 males) who underwent DWI and diffusion spectrum imaging (DTI, DKI, MAP, and NODDI) were enrolled, and classified into 23 low-grade and 21 high-grade PITs. Diffusion metrics were compared between the low- and high-grade PITs using the Mann-Whitney U test. Receiver operating characteristic (ROC) analysis and Spearman correlations were performed. Sensitivity analysis was performed in 27 pediatric gliomas. The MAP_MSD (mean squared diffusion), MAP_QIV (q-space inverse variance), and DWI_ADC (apparent diffusion coefficient) values were significantly lower, while MAP_NG (non-gaussianity), MAP_NGAx (axial non-gaussianity), MAP_NGRad (radial non-gaussianity), NODDI_ICVF (intracellular volume fraction), and NODDI_ODI (orientation dispersion index) were significantly higher in high-grade PITs than in low-grade PITs (p < 0.05). All parameters yielded AUCs (areas under the ROC curves) of 0.72–0.82. NODDI_ICVF achieved the highest diagnostic performance for distinguishing PIT grades (AUC = 0.82, p < 0.001), with significant integrated discrimination improvement over DWI (corrected p < 0.05). Sensitivity analysis showed the robustness of MAP parameters in pediatric glioma grading, with favorable efficiency (AUC = 0.75–0.82). Furthermore, MAP, NODDI, and DWI metrics exhibited significant correlations with the Ki-67 index. MAP and NODDI have comparable diagnostic performance to DWI in PIT grading, and may provide new approaches for selecting therapeutic strategies for PITs.
Lung cancer remains the leading cause of cancer-related mortality worldwide, and low-dose CT (LDCT) screening has been shown to significantly reduce lung cancer deaths in high-risk populations. The American College of Radiology (ACR) introduced the Lung CT Screening Reporting and Data System (Lung-RADS) in 2014 to standardize the reporting and management of screen-detected pulmonary nodules. Since its inception, Lung-RADS has undergone iterative updates reflecting advances in evidence, clinical practice, and the expanding scope of lung cancer screening. This review summarizes the major updates to Lung-RADS, highlights areas of active discussion within the Lung-RADS Committee, and outlines future research directions. Key topics include the ongoing evaluation of volumetric versus diameter-based nodule measurement, the management of subsolid nodules and the balance between surveillance and intervention, the implementation of Category 0 for infectious or inflammatory findings, clarification of juxtapleural nodule classification, the approach to incidental findings, and the development of a Lung-RADS Atlas. Future directions address the integration of risk-based and category-based assessment approaches, the role of artificial intelligence and deep learning in nodule stratification, the potential incorporation of incidental nodules into Lung-RADS, and the prospect of international harmonization for lung cancer screening reporting and management. Lung-RADS has become an essential reporting and quality tool for lung cancer screening in the United States and increasingly worldwide. Continued refinement, guided by emerging evidence, multidisciplinary collaboration, and technological innovation, will be critical to advancing precision screening strategies and reducing the global burden of lung cancer. Clinical trial number Not applicable.
