Prospective Comparison of Standard and Deep Learning-reconstructed Turbo Spin-Echo MRI of the Shoulder.

Background Deep learning (DL)-based MRI reconstructions can reduce imaging times for turbo spin-echo (TSE) examinations. However, studies that prospectively use DL-based reconstructions of rapidly acquired, undersampled MRI in the shoulder are lacking. Purpose To compare the acquisition time, image quality, and diagnostic confidence of DL-reconstructed TSE (TSEDL) with standard TSE in patients indicated for shoulder MRI. Materials and Methods This prospective single-center study included consecutive adult patients with various shoulder abnormalities who were clinically referred for shoulder MRI between February and March 2023. Each participant underwent standard TSE MRI (proton density- and T1-weighted imaging; conventional TSE sequence was used as reference for comparison), followed by a prospectively undersampled accelerated TSEDL examination. Six musculoskeletal radiologists evaluated images using a four-point Likert scale (1, poor; 4, excellent) for overall image quality, perceived signal-to-noise ratio, sharpness, artifacts, and diagnostic confidence. The frequency of major pathologic features and acquisition times were also compared between the acquisition protocols. The intergroup comparisons were performed using the Wilcoxon signed rank test. Results Overall, 135 shoulders in 133 participants were evaluated (mean age, 47.9 years ± 17.1 [SD]; 73 female participants). The median acquisition time of the TSEDL protocol was lower than that of the standard TSE protocol (288 seconds [IQR, 288-288 seconds] vs 926 seconds [IQR, 926-950 seconds], respectively; P < .001), achieving a 69% lower acquisition time. TSEDL images were given higher scores for overall image quality, perceived signal-to-noise ratio, and artifacts (all P < .001). Similar frequency of pathologic features (P = .48 to > .99), sharpness (P = .06), or diagnostic confidence (P = .05) were noted between images from the two protocols. Conclusion In a clinical setting, TSEDL led to reduced examination time and higher image quality with similar diagnostic confidence compared with standard TSE MRI in the shoulder. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Chang and Chow in this issue.

CT-based deep learning model for predicting hospital discharge outcome in spontaneous intracerebral hemorrhage.

OBJECTIVES: To predict the functional outcome of patients with intracerebral hemorrhage (ICH) using deep learning models based on computed tomography (CT) images. METHODS: A retrospective, bi-center study of ICH patients was conducted. Firstly, a custom 3D convolutional model was built for predicting the functional outcome of ICH patients based on CT scans from randomly selected ICH patients in H training dataset collected from H hospital. Secondly, clinical data and radiological features were collected at admission and the Extreme Gradient Boosting (XGBoost) algorithm was used to establish a second model, named the XGBoost model. Finally, the Convolution model and XGBoost model were fused to build the third "Fusion model." Favorable outcome was defined as modified Rankin Scale score of 0-3 at discharge. The prognostic predictive accuracy of the three models was evaluated using an H test dataset and an external Y dataset, and compared with the performance of ICH score and ICH grading scale (ICH-GS). RESULTS: A total of 604 patients with ICH were included in this study, of which 450 patients were in the H training dataset, 50 patients in the H test dataset, and 104 patients in the Y dataset. In the Y dataset, the areas under the curve (AUCs) of the Convolution model, XGBoost model, and Fusion model were 0.829, 0.871, and 0.905, respectively. The Fusion model prognostic performance exceeded that of ICH score and ICH-GS (p = 0.043 and p = 0.045, respectively). CONCLUSIONS: Deep learning models have good accuracy for predicting functional outcome of patients with spontaneous intracerebral hemorrhage. CLINICAL RELEVANCE STATEMENT: The proposed deep learning Fusion model may assist clinicians in predicting functional outcome and developing treatment strategies, thereby improving the survival and quality of life of patients with spontaneous intracerebral hemorrhage. KEY POINTS: • Integrating clinical presentations, CT images, and radiological features to establish deep learning model for functional outcome prediction of patients with intracerebral hemorrhage. • Deep learning applied to CT images provides great help in prognosing functional outcome of intracerebral hemorrhage patients. • The developed deep learning model performs better than clinical prognostic scores in predicting functional outcome of patients with intracerebral hemorrhage.

Intravoxel incoherent motion MRI for rectal cancer: correlation of diffusion and perfusion characteristics with clinical-pathologic factors.

BACKGROUND: Colorectal cancer is the most common cause of cancer-related death worldwide. Magnetic resonance imaging (MRI) has become a promising alternative method for staging the cancer. PURPOSE: To evaluate parameters of intravoxel incoherent motion (IVIM) and their relationships with clinical-pathologic factors in rectal cancers. MATERIAL AND METHODS: A total of 51 patients with histopathologically proven rectal cancer who underwent preoperative pelvic MRI were prospectively enrolled. Parameters (ADC, D, D*, and f) derived from IVIM-diffusion-weighted imaging (DWI) were independently measured by two radiologists. Student's t-test, receiver operating characteristic curves, and Spearman correlation were used for statistical analysis. RESULTS: ADC, D, and D* were significantly higher in pT1-2 tumors than in pT3-4 tumors (1.108 ± 0.233 vs. 0.950 ± 0.176, 0.796 ± 0.199 vs. 0.684 ± 0.114, 0.013 ± 0.005 vs. 0.008 ± 0.003, respectively; P < 0.05). D* exhibited a strong correlation with the tumor stage (r = -0.675, P < 0.001). In poorly differentiated cluster (PDC) grading, ADC, D*, and f were significantly lower in high-grade tumors than in low-grade tumors (0.905 ± 0.148 vs. 1.064 ± 0.200, 0.008 ± 0.002 vs. 0.011 ± 0.005, and 0.252 ± 0.032 vs. 0.348 ± 0.058, respectively; P < 0.05). The f value exhibited a significantly strong correlation with the PDC grades (r = -0.842, P < 0.001), and higher sensitivity and specificity (95.2% and 75.9%) than those shown by the ADC, D, and D* values. CONCLUSION: IVIM parameters, especially f, demonstrated a strong correlation with histologic grades and showed a better performance in differentiating between high- and low-grade rectal cancers. These parameters would be helpful in predicting tumor aggressiveness and prognosis.

Tracking intrahepatically transplanted islets labeled with Feridex-polyethyleneimine complex using a clinical 3.0-T magnetic resonance imaging scanner.

OBJECTIVES: To evaluate the images acquired with a clinical 3.0-T magnetic resonance imaging machine as the quantification of transplanted and surviving islets in vivo. METHODS: Polyethyleneimine (PEI) was introduced to increase the labeling efficiency of Feridex, a dextran-coated superparamagnetic iron oxide. Allogeneic (Lewis-to-Wistar) and syngeneic (Wistar-to-Wistar) intraheptatic islet transplantations were performed to study the relationship among magnetic resonance imaging, metabolic monitoring, and pathological examination. RESULTS: After receiving Feridex-PEI-labeled islets, dark voids could be observed in the livers of both groups, accompanied with a significant decrease in liver/muscle intensity ratio from 1.25 +/- 0.03 to 1.09 +/- 0.05 (P < 0.01). One week after transplantation, islet grafts were rejected in the allogeneic group. Rapid disappearance of dark voids and a significant increase of liver/muscle ratio were observed. No islet grafts could be found in the paraffin sections of livers by that time. Meanwhile, in the syngeneic group, islet grafts survived indefinitely. Dark voids persisted and low liver/muscle ratios retained. The fact that the dark voids represented the labeled islets was confirmed by combined staining of insulin activity and Prussian blue. CONCLUSIONS: Either spot counting or signal intensity measurement provides a perfect quantification of transplanted and surviving islets in vivo. Feridex-PEI provides an effective and safe way to label islets.