Unenhanced Computed Tomography Images Research Articles

Extracellular volume-based scoring system for tracking tumor progression in pancreatic cancer patients receiving intraoperative radiotherapy

ObjectivesTo investigate the value of extracellular volume (ECV) derived from portal-venous phase (PVP) in predicting prognosis in locally advanced pancreatic cancer (LAPC) patients receiving intraoperative radiotherapy (IORT) with initial stable disease (SD) and to construct a risk-scoring system based on ECV and clinical-radiological features.Materials and methodsOne hundred and three patients with LAPC who received IORT demonstrating SD were enrolled and underwent multiphasic contrast-enhanced CT (CECT) before and after IORT. ECV maps were generated from unenhanced and PVP CT images. Clinical and CT imaging features were analyzed. The independent predictors of progression-free survival (PFS) determined by multivariate Cox regression model were used to construct the risk-scoring system. Time-dependent receiver operating characteristic (ROC) curve analysis and the Kaplan–Meier method were used to evaluate the predictive performance of the scoring system.ResultsMultivariable analysis revealed that ECV, rim-enhancement, peripancreatic fat infiltration, and carbohydrate antigen 19-9 (CA19-9) response were significant predictors of PFS (all p < 0.05). Time-dependent ROC of the risk-scoring system showed a satisfactory predictive performance for disease progression with area under the curve (AUC) all above 0.70. High-risk patients (risk score ≥ 2) progress significantly faster than low-risk patients (risk score < 2) (p < 0.001).ConclusionECV derived from PVP of conventional CECT was an independent predictor for progression in LAPC patients assessed as SD after IORT. The scoring system integrating ECV, radiological features, and CA19-9 response can be used as a practical tool for stratifying prognosis in these patients, assisting clinicians in developing an appropriate treatment approach.Critical relevance statementThe scoring system integrating ECV fraction, radiological features, and CA19-9 response can track tumor progression in patients with LAPC receiving IORT, aiding clinicians in choosing individual treatment strategies and improving their prognosis.Key PointsPredicting the progression of LAPC in patients receiving IORT is important.Our ECV-based scoring system can risk stratifying patients with initial SD.Appropriate prognostication can assist clinicians in developing appropriate treatment approaches.Graphical

CT-based radiomics model using stability selection for predicting the WHO/ISUP grade of clear cell renal cell carcinoma.

This study aimed to develop a model to predict World Health Organization/International Society of Urological Pathology (WHO/ISUP) low-grade or high-grade clear cell renal cell carcinoma (ccRCC) using 3D multi-phase enhanced computed tomography (CT) radiomics features (RFs). CT data of 138 low-grade and 60 high-grade ccRCC cases were included. RFs were extracted from four CT phases: non-contrast phase (NCP), cortico-medullary phase (CMP), nephrographic phase (NP), and excretory phase (EP). Models were developed using various combinations of RFs and subjected to cross-validation. There were 107 RFs extracted from each phase of the CT images. The NCP-EP model had the best overall predictive value (AUC = 0.78), but did not significantly differ from that of the NCP model (AUC = 0.76). By considering the predictive ability of the model, the level of radiation exposure, and model simplicity, the overall best model was the Conventional image and clinical features (CICFs)-NCP model (AUC = 0.77; sensitivity 0.75, specificity 0.69, positive predictive value 0.85, negative predictive value 0.54, accuracy 0.73). The second-best model was the NCP model (AUC = 0.76). Combining clinical features with unenhanced CT images of the kidneys seems to be optimal for prediction of WHO/ISUP grade of ccRCC. This noninvasive method may assist in guiding more accurate treatment decisions for ccRCC. This study innovatively employed stability selection for RFs, enhancing model reliability. The CICFs-NCP model's simplicity and efficacy mark a significant advancement, offering a practical tool for clinical decision-making in ccRCC management.

CT and MRI findings of small cell neuroendocrine carcinoma of the urinary bladder: comparison with urothelial carcinoma.

This study aimed to evaluate the efficacy of CT and MRI findings to differentiate small cell neuroendocrine carcinoma (SCNEC) from urothelial carcinoma (UC) of the urinary bladder. This study included 90 patients with histopathologically confirmed bladder cancer (10 SCNECs and 80 UCs). Eight patients with bladder SCNEC and 80 with UC underwent CT and MRI, whereas the remaining two patients with SCNEC underwent CT alone before treatment. CT and MRI findings were retrospectively evaluated and compared between the two pathologies. The maximum diameter (36.5mm vs. 19.0mm, p < 0.01) and height (22.0mm vs. 14.0mm, p < 0.01) of the tumor in bladder SCNEC were higher than in UC. The pedunculated configuration (20% vs. 61%, p < 0.05) and irregular tumor margins (20% vs. 76%, p < 0.01) in bladder SCNEC were less common than in UC. The CT attenuation of the solid component in unenhanced CT images was higher in bladder SCNEC than in UC (37 Hounsfield unit [HU] vs. 34 HU, p < 0.01). The apparent diffusion coefficient (ADC) of the solid component in bladder SCNEC was lower than in UC (0.49 × 10-3 mm2/s vs. 1.02 × 10-3 mm2/s, p < 0.01). In comparison with UC, bladder SCNEC was larger, had higher unenhanced CT attenuation, and had a lower ADC value. The pedunculated configuration and irregular tumor margins were typical of bladder UC.

CT-based radiomics research for discriminating the risk stratification of pheochromocytoma using different machine learning models: a multi-center study.

The purpose of this study was to explore and verify the value of various machine learning models in preoperative risk stratification of pheochromocytoma. A total of 155 patients diagnosed with pheochromocytoma through surgical pathology were included in this research (training cohort: n = 105; test cohort: n = 50); the risk stratification scoring system classified a PASS score of < 4 as low risk and a PASS score of ≥ 4 as high risk. From CT images captured during the non-enhanced, arterial, and portal venous phase, radiomic features were extracted. After reducing dimensions and selecting features, Logistic Regression (LR), Extra Trees, and K-Nearest Neighbor (KNN) were utilized to construct the radiomics models. By adopting ROC curve analysis, the optimal radiomics model was selected. Univariate and multivariate logistic regression analyses of clinical radiological features were used to determine the variables and establish a clinical model. The integration of radiomics and clinical features resulted in the creation of a combined model. ROC curve analysis was used to evaluate the performance of the model, while decision curve analysis (DCA) was employed to assess its clinical value. 3591 radiomics features were extracted from the region of interest in unenhanced and dual-phase (arterial and portal venous phase) CT images. 13 radiomics features were deemed to be valuable. The LR model demonstrated the highest prediction efficiency and robustness among the tested radiomics models, with an AUC of 0.877 in the training cohort and 0.857 in the test cohort. Ultimately, the composite of clinical features was utilized to formulate the clinical model. The combined model demonstrated the best discriminative ability (AUC, training cohort: 0.887; test cohort: 0.874). The DCA of the combined model showed the best clinical efficacy. The combined model integrating radiomics and clinical features had an outstanding performance in differentiating the risk of pheochromocytoma and could offer a non-intrusive and effective approach for making clinical decisions.

Predictive Value of Dual-Energy CT-Derived Metrics for the Use of Bone Substitutes in Distal Radius Fracture Surgery.

(1) Background: Low bone mineral density (BMD) is a significant risk factor for complicated surgery and leads to the increased use of bone substitutes in patients with distal radius fractures (DRFs). No accepted model has yet been established to predict the use of bone substitutes to facilitate preoperative planning. (2) Methods: Unenhanced dual-energy CT (DECT) images of DRFs were retrospectively acquired between March 2016 and September 2020 using the internal PACS system. Available follow-up imaging and medical health records were reviewed to determine the use of bone substitutes. DECT-based BMD, trabecular Hounsfield units (HU), cortical HU, and cortical thickness ratio were measured in non-fractured segments of the distal radius. Diagnostic accuracy parameters were calculated for all metrics using receiver-operating characteristic (ROC) curves and associations of all metrics with the use of bone substitutes were evaluated using logistic regression models. (3) The final study population comprised 262 patients (median age 55 years [IQR 43-67 years]; 159 females, 103 males). According to logistic regression analysis, DECT-based BMD was the only metric significantly associated with the use of bone substitutes (odds ratio 0.96, p = 0.003). However, no significant associations were found for cortical HU (p = 0.06), trabecular HU (p = 0.33), or cortical thickness ratio (p = 0.21). ROC-curve analysis revealed that a combined model of all four metrics had the highest diagnostic accuracy with an area under the curve (AUC) of 0.76. (4) Conclusions: DECT-based BMD measurements performed better than HU-based measurements and cortical thickness ratio. The diagnostic performance of all four metrics combined was superior to that of the individual parameters.

CT-based radiomics for differentiating peripherally located pulmonary sclerosing pneumocytoma from carcinoid.

Pulmonary sclerosing pneumocytoma (PSP) and pulmonary carcinoid (PC) are difficult to distinguish based on conventional imaging examinations. In recent years, radiomics has been used to discriminate benign from malignant pulmonary lesions. However, the value of radiomics based on computed tomography (CT) images to differentiate PSP from PC has not been well explored. We aimed to investigate the feasibility of radiomics in the differentiation between PSP and PC. Fifty-three PSP and fifty-five PC were retrospectively enrolled and then were randomly divided into the training and test sets. Univariate and multivariable logistic analyses were carried to select clinical predictor related to differential diagnosis of PSP and PC. A total of 1316 radiomics features were extracted from the unenhanced CT (UECT) and contrast-enhanced CT (CECT) images, respectively. The minimum redundancy maximum relevance and the least absolute shrinkage and selection operator were used to select the most significant radiomics features to construct radiomics models. The clinical predictor and radiomics features were integrated to develop combined models. Two senior radiologists independently categorized each patient into PSP or PC group based on traditional CT method. The performances of clinical, radiomics, and combined models in differentiating PSP from PC were investigated by the receiver operating characteristic (ROC) curve. The diagnostic performance was also compared between the combined models and radiologists. In regard to differentiating PSP from PC, the area under the curves (AUCs) of the clinical, radiomics, and combined models were 0.87, 0.96, and 0.99 in the training set UECT, and were 0.87, 0.97, and 0.98 in the training set CECT, respectively. The AUCs of the clinical, radiomics, and combined models were 0.84, 0.92, and 0.97 in the test set UECT, and were 0.84, 0.93, and 0.98 in the test set CECT, respectively. In regard to the differentiation between PSP and PC, the combined model was comparable to the radiomics model, but outperformed the clinical model and the two radiologists, whether in the test set UECT or CECT. Radiomics approaches show promise in distinguishing between PSP and PC. Moreover, the integration of clinical predictor (gender) has the potential to enhance the diagnostic performance even further.

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

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

Gold nanoshells for prostate cancer treatment: evidence for deposition in abdominal organs.

Gold-silica nanoshell therapy [AuroShells with subsequent focal laser therapy (AuroLase)] is an emerging targeted treatment modality for prostate cancer. We reviewed pre- and post-treatment unenhanced CT imaging to assess for retained gold-silica nanoshells in the abdomen and pelvis. This single-institution retrospective study identified patients in the AuroLase pilot who underwent pre- and post-treatment unenhanced abdominopelvic CT. The attenuation, before and after gold-silica nanoshell administration, of the liver, spleen, pancreas, kidneys, prostate, blood pool, paraspinal musculature, and abnormal lymph nodes were manually measured by two readers. After inter-reader agreement was calculated using intraclass correlation (ICC), a permutation test was used to assess pre- and post-therapy attenuation differences. Four patients met the inclusion criteria. Mean age was 72.3 ± 5.9years. Median time interval between pre-treatment CT and treatment, and between treatment and post-treatment CT, was 232days and 236.5days, respectively. The two readers' attenuation measurements had very high agreement (ICC = 0.99, p < 0.001). The highest differences in organ attenuation between pre- and post-therapy scans were seen in all four patients in the liver and spleen (liver increased by an average of 28.9 HU, p = 0.010; spleen increased by an average of 63.7 HU, p = 0.012). A single measured lymph node increased by an average of 58.9 HU. In the remainder of the measured sites, the change in attenuation from pre- to post-therapy scans ranged from -0.1 to 3.8 HU (p > 0.05). Increased attenuation of liver and spleen at CT can be an expected finding in patients who have received gold-silica nanoshell therapy.

Prediction of bowel necrosis by reduced bowel wall enhancement in closed-loop small bowel obstruction: Quantitative methods

PurposeTo assess diagnostic performance and reproducibility of reduced bowel wall enhancement evaluated by quantitative methods using CT to identify bowel necrosis among closed-loop small bowel obstruction (CL-SBO) patients. MethodsThis retrospective single-center study included patients who diagnosed with CL-SBO caused by adhesion or internal hernia during January 2016 and May 2022. Patients were divided into necrotic group (n = 41) and non-necrotic group (n = 67) according to surgical exploration and postoperative pathology. Two doctors independently measured the attenuation of bowel wall and consensus was reached through panel discussion with a third gastrointestinal radiologist. Reduced bowel wall enhancement was assessed by four quantitative methods. Univariate analyses were used to evaluate the association between each method and bowel necrosis, and kappa/intraclass correlation coefficient values were used to assess interobserver agreement. Diagnostic performance parameters were calculated for each method. ResultsReduced bowel wall enhancement in arterial phase (OR 8.98, P < 0.0001), reduced bowel wall enhancement in portal phase (OR 16.84, P < 0.001), adjusted reduced bowel wall enhancement in arterial phase (OR 29.48, P < 0.001), adjusted reduced bowel wall enhancement in portal phase (OR 145.69, P < 0.001) were significantly associated with bowel necrosis. Adjusted reduced bowel wall enhancement in portal phase had the best diagnostic performance (AUC: 0.92; Youden index: 0.84; specificity: 94.03 %) and interobserver agreement (kappa value of 0.59–0.73) to predict bowel necrosis. ConclusionWhen assessing reduced bowel enhancement to predict bowel necrosis among CL-SBO patients, using unenhanced CT images and proximal dilated loop as standard references in portal phase is the most accurate quantitative method among those tested.

LNAS: a clinically applicable deep-learning system for mediastinal enlarged lymph nodes segmentation and station mapping without regard to the pathogenesis using unenhanced CT images.

The accurate identification and evaluation of lymph nodes by CT images is of great significance for disease diagnosis, treatment, and prognosis. To assess the lymph nodes' segmentation, size, and station by artificial intelligence (AI) for unenhanced chest CT images and evaluate its value in clinical scenarios. This retrospective study proposed an end-to-end Lymph Nodes Analysis System (LNAS) consisting of three models: the Lymph Node Segmentation model (LNS), the Mediastinal Organ Segmentation model (MOS), and the Lymph Node Station Registration model (LNR). We selected a healthy chest CT image as the template image and annotated 14 lymph node station masks according to the IASLC to build the lymph node station mapping template. The exact contours and stations of the lymph nodes were annotated by two junior radiologists and reviewed by a senior radiologist. Patients aged 18 and above, who had undergone unenhanced chest CT and had at least one suspicious enlarged mediastinal lymph node in imaging reports, were included. Exclusions were patients who had thoracic surgeries in the past 2weeks or artifacts on CT images affecting lymph node observation by radiologists. The system was trained on 6725 consecutive chest CTs that from Tianjin Medical University General Hospital, among which 6249 patients had suspicious enlarged mediastinal lymph nodes. A total of 519 consecutive chest CTs from Qilu Hospital of Shandong University (Qingdao) were used for external validation. The gold standard for each CT was determined by two radiologists and reviewed by one senior radiologist. The patient-level sensitivity of the LNAS system reached of 93.94% and 92.89% in internal and external test dataset, respectively. And the lesion-level sensitivity (recall) reached 89.48% and 85.97% in internal and external test dataset. For man-machine comparison, AI significantly apparently shortened the average reading time (p < 0.001) and had better lesion-level and patient-level sensitivities. AI improved the sensitivity lymph node segmentation by radiologists with an advantage in reading time.

Development and validation of a radiomics nomogram for diagnosis of malignant pleural effusion

ObjectiveWe aimed to develop a radiomics nomogram based on computed tomography (CT) scan features and high-throughput radiomics features for diagnosis of malignant pleural effusion (MPE).MethodsIn this study, 507 eligible patients with PE (207 malignant and 300 benign) were collected retrospectively. Patients were divided into training (n = 355) and validation cohorts (n = 152). Radiomics features were extracted from initial unenhanced CT images. CT scan features of PE were also collected. We used the variance threshold algorithm and least absolute shrinkage and selection operator (LASSO) to select optimal features to build a radiomics model for predicting the nature of PE. Univariate and multivariable logistic regression analyzes were used to identify significant independent factors associated with MPE, which were then included in the radiomics nomogram.ResultsA total of four CT features were retained as significant independent factors, including massive PE, obstructive atelectasis or pneumonia, pleural thickening > 10 mm, and pulmonary nodules and/or masses. The radiomics nomogram constructed from 13 radiomics parameters and four CT features showed good predictive efficacy in training cohort [area under the curve (AUC) = 0.926, 95% CI 0.894, 0.951] and validation cohort (AUC = 0.916, 95% CI 0.860, 0.955). The calibration curve and decision curve analysis showed that the nomogram helped differentiate MPE from benign pleural effusion (BPE) in clinical practice.ConclusionThis study presents a nomogram model incorporating CT scan features and radiomics features to help physicians differentiate MPE from BPE.

Deep learning reconstruction for improving the visualization of acute brain infarct on computed tomography.

This study aimed to investigate the impact of deep learning reconstruction (DLR) on acute infarct depiction compared with hybrid iterative reconstruction (Hybrid IR). This retrospective study included 29 (75.8 ± 13.2years, 20 males) and 26 (64.4 ± 12.4years, 18 males) patients with and without acute infarction, respectively. Unenhanced head CT images were reconstructed with DLR and Hybrid IR. In qualitative analyses, three readers evaluated the conspicuity of lesions based on five regions and image quality. A radiologist placed regions of interest on the lateral ventricle, putamen, and white matter in quantitative analyses, and the standard deviation of CT attenuation (i.e., quantitative image noise) was recorded. Conspicuity of acute infarct in DLR was superior to that in Hybrid IR, and a statistically significant difference was observed for two readers (p ≤ 0.038). Conspicuity of acute infarct with time from onset to CT imaging at < 24h in DLR was significantly improved compared with Hybrid IR for all readers (p ≤ 0.020). Image noise in DLR was significantly reduced compared with Hybrid IR with both the qualitative and quantitative analyses (p < 0.001 for all). DLR in head CT helped improve acute infarct depiction, especially those with time from onset to CT imaging at < 24h.

Abstract 256: Quantitative Assessment and Visualization of Hypoxic Ischemic Brain Injury in Survivors of Cardiac Arrest Using Unenhanced Computed Tomography Images

Aim: This study aimed to investigate whether the distribution of Hounsfield units (HU) values in head computed tomography (HCT) images can serve as an effective tool for assessing the severity of hypoxic ischemic brain injury (HIBI) in survivors of out-of-hospital cardiac arrest (OHCA). Methods: This retrospective study included adult comatose OHCA survivors who underwent first HCT scans within 6 hours and second HCT scans between 72-96 hours on the same CT machine after the return of spontaneous circulation. Automated quantitative analysis was used on CT images to eliminate artifacts and identify differences in the proportion of HU values (HUp) at specific intervals across the intracranial area based on neurological outcome. Additionally, correlation analysis was conducted between serum levels of neuron-specific enolase (NSE) at 48 hours post-ROSC and the sum of HUp values in the aforementioned intervals. Poor outcome was defined as cerebral performance categories 3-5 at 6 months post-ROSC. Results: Of the 55 included patients (84% male), 23 (42%) had poor outcome. Significant differences in HU values were observed between outcomes in the second HCT scan at HU=1-14, 23-35, and 39-56 (P&lt;0.05). Within the same HU interval, the poor outcome group showed a greater change between the first and second HCT scans compared to the good outcome group (Figure). The correlations between the sum of HUp and NSE levels within the range of HU=1-14, 23-35, and 39-56 in the second HCT were 0.34, 0.60, and 0.49, respectively (Figure). The second HCT scans demonstrated better prognostic performance than the first HCT scan for HU=1-14 (area under the curve 0.75 vs. 0.60), 23-35 (0.85 vs. 0.62), and 39-56 (0.75 vs. 0.59), while no significant difference was observed compared to GWR (0.78) (all P≥0.05). Conclusions: The proportional sum of HU values within specific intervals in the entire cerebral area on CT images can effectively assess and visualize the extent of HIBI in survivors of OHCA.

96 Quantitative assessment of hypoxic ischemic brain injury in survivors of cardiac arrest using unenhanced computed tomography images

96 Quantitative assessment of hypoxic ischemic brain injury in survivors of cardiac arrest using unenhanced computed tomography images

4D-DSA for Assessment of the Angioarchitecture and Grading of Cranial Dural AVF.

Time-resolved 3D rotational angiography (4D-DSA) has been used to demonstrate details of the angioarchitecture of AVM, whereas it has rarely been used to describe features of dural AVF. In this exploratory study, we analyzed dural AVFs with a novel 4D software prototype, developed and provided by Siemens, to determine whether identification of the location of the fistulous point, grading, and treatment planning were feasible. 4D-DSA volumes were calculated from existing 3D rotational angiography data sets of patients with dural AVFs. The 4D-DSA volumes were displayed in a virtual DSA mode and MPR or MIP in 3 orthogonal planes and compared with 2D-DSA by 2 experienced neuroradiologists. Fusions with unenhanced CT or MR images were used to improve visualization of adjacent anatomic structures. Comparison with 2D-DSA showed that evaluation of the fistulous point and grading according to the classification of Borden, Cognard, or Barrow was feasible in 26 of 27 cases. In 8 of 27 cases, 4D-DSA was considered advantageous for determining the fistulous point and the course of the draining vein in the dural AVF with cortical venous drainage, especially in the frontoethmoidal and frontoparietal regions. In 6 cases, the display of angioarchitecture was considered inferior to that of 2D-DSA due to motion artifacts, suboptimal selection of the injected vessel, and lack of temporal resolution. Detailed analysis of dural AVFs according to the standardized display of 4D-DSA volumes was feasible and helpful in understanding the angioarchitecture in selected cases. Further improvement and validation of the 4D software should solidify the complementary role of 4D-DSA to conventional 2D-DSA series.

GA-Net: A geographical attention neural network for the segmentation of body torso tissue composition

PurposeBody composition analysis (BCA) of the body torso plays a vital role in the study of physical health and pathology and provides biomarkers that facilitate the diagnosis and treatment of many diseases, such as type 2 diabetes mellitus, cardiovascular disease, obstructive sleep apnea, and osteoarthritis. In this work, we propose a body composition tissue segmentation method that can automatically delineate those key tissues, including subcutaneous adipose tissue, skeleton, skeletal muscle tissue, and visceral adipose tissue, on positron emission tomography/computed tomography scans of the body torso. MethodsTo provide appropriate and precise semantic and spatial information that is strongly related to body composition tissues for the deep neural network, first we introduce a new concept of the body area and integrate it into our proposed segmentation network called Geographical Attention Network (GA-Net). The body areas are defined following anatomical principles such that the whole body torso region is partitioned into three non-overlapping body areas. Each body composition tissue of interest is fully contained in exactly one specific minimal body area. Secondly, the proposed GA-Net has a novel dual-decoder schema that is composed of a tissue decoder and an area decoder. The tissue decoder segments the body composition tissues, while the area decoder segments the body areas as an auxiliary task. The features of body areas and body composition tissues are fused through a soft attention mechanism to gain geographical attention relevant to the body tissues. Thirdly, we propose a body composition tissue annotation approach that takes the body area labels as the region of interest, which significantly improves the reproducibility, precision, and efficiency of delineating body composition tissues. ResultsOur evaluations on 50 low-dose unenhanced CT images indicate that GA-Net outperforms other architectures statistically significantly based on the Dice metric. GA-Net also shows improvements for the 95% Hausdorff Distance metric in most comparisons. Notably, GA-Net exhibits more sensitivity to subtle boundary information and produces more reliable and robust predictions for such structures, which are the most challenging parts to manually mend in practice, with potentially significant time-savings in the post hoc correction of these subtle boundary placement errors. Due to the prior knowledge provided from body areas, GA-Net achieves competitive performance with less training data. Our extension of the dual-decoder schema to TransUNet and 3D U-Net demonstrates that the new schema significantly improves the performance of these classical neural networks as well. Heatmaps obtained from attention gate layers further illustrate the geographical guidance function of body areas for identifying body tissues. Conclusions(i) Prior anatomic knowledge supplied in the form of appropriately designed anatomic container objects significantly improves the segmentation of bodily tissues. (ii) Of particular note are the improvements achieved in the delineation of subtle boundary features which otherwise would take much effort for manual correction. (iii) The method can be easily extended to existing networks to improve their accuracy for this application.

Radiomics Model for Predicting Pulmonary Metastasis in Sarcoma Patients

The lung is the most common site of metastasis in localized extremity soft tissue sarcomas (STS). Indeterminant lung nodules leads to challenges for both patients and providers. As such, accuracy in the diagnosis of pulmonary metastasis (PM) is important for clinical decision making. The purpose of this is study is to develop and evaluate a CT-based radiomic model to predict PM in sarcoma patients with indeterminate pulmonary nodules. Fifty-seven stage I-III sarcoma patients with indeterminate lung nodules on unenhanced CT that were approved for this study by our institution's internal review board (IRB). Twenty-two developed and categorized as PM and 35 without PM. A total number of 322 indeterminate nodules were investigated, where 165 & 157 nodules were from patients with and without PM, respectively. A total of 96 radiomic features (RF) (including first order statistics, 2d shape features, gray level co-occurrence, gray level size zone, gray level run length, neighboring gray tone difference & gray level dependence matrices) were taken from unenhanced CT images of each patient. We included 2D shape features since the size and shape of indeterminate lung nodules may have some correlation with PM. Spearman correlation were used to rule out redundant features (r2>0.9). T-test statistics was used to determine feature with significant differences that can differentiate responders with and without PM. Support Vector Machine (SVM) model was fitted to combinations of groups of 2-3 radiomic features. The AUC of the ROC curve was used to evaluate the model performance. Pointwise confidence intervals (CI) for AUC were calculated using 100 bootstrap replicates. Four features were found significant after Spearman correlation and T-test statistical testing: 1. Maximum (MAX); 2. Minor Axis Length (MAL); 3. Gray Level Non-uniformity (GLNU); and 4. Interquartile Range (IR). The best performing 2-feature radiomic SVM model had an AUC of 0.69 (95% CI 0.61,0.74) for the MAX and GLNU features. The best performing 3-feature radiomic SVM model had an AUC of 0.74 (95% CI 0.70,0.80) for the MAX, MAL and IR features. CT-based radiomics model has the potential to be used to predict PM in sarcoma patients presenting with indeterminate lung nodules. Additional research will focus on investigating derived parameters for existing RF, such as the ratio of the max RF to mean RF, to see if they improve the predictive performance of the model, as well as testing the models on external patient cohort. These radiomic models could be useful tools in guiding PM screening and providing a patient specific monitoring of sarcoma patients.

Enhanced and unenhanced: Radiomics models for discriminating between benign and malignant cystic renal masses on CT images: A multi-center study.

Machine learning algorithms used to classify cystic renal masses (CRMs) nave not been applied to unenhanced CT images, and their diagnostic accuracy had not been compared against radiologists. This retrospective study aimed to develop radiomics models that discriminate between benign and malignant CRMs in a triple phase computed tomography (CT) protocol and compare the diagnostic accuracy of the radiomics approach with experienced radiologists. Predictive models were established using a training set and validation set of unenhanced and enhanced (arterial phase [AP] and venous phase [VP]) CT images of benign and malignant CRMs. The diagnostic capabilities of the models and experienced radiologists were compared using Receiver Operating Characteristic (ROC) curves. On unenhanced, AP and VP CT images in the validation set, the AUC, specificity, sensitivity and accuracy for discriminating between benign and malignant CRMs were 90.0 (95%CI: 81-98%), 90.0%, 90.5% and 90.2%; 93.0% (95%CI: 86-99%), 86.7%, 95.2% and 88.3%; and 95.0% (95%CI: 90%-100%), 93.3%, 90.5% and 92.1%, respectively, for the radiomics models. Diagnostic accuracy of the radiomics models differed significantly on unenhanced images in the training set vs. each radiologist (p = 0.001 and 0.003) but not in the validation set (p = 0.230 and 0.590); differed significantly on AP images in the validation set vs. each radiologist (p = 0.007 and 0.007) but not in the training set (p = 0.663 and 0.663); and there were no differences on VP images in the training or validation sets vs. each radiologist (training set: p = 0.453 and 0.051, validation set: p = 0.236 and 0.786). Radiomics models may have clinical utility for discriminating between benign and malignant CRMs on unenhanced and enhanced CT images. The performance of the radiomics model on unenhanced CT images was similar to experienced radiologists, implying it has potential as a screening and diagnostic tool for CRMs.

Computed tomography-determined skeletal muscle density predicts 3-year mortality in initial-dialysis patients in China.

Skeletal muscle mass and quality assessed by computed tomography (CT) images of the third lumbar vertebra (L3) level have been established as risk factors for poor clinical outcomes in several illnesses, but the relevance for dialysis patients is unclear. A few studies have suggested a correlation between CT-determined skeletal muscle mass and quality at the first lumbar vertebra (L1) level and adverse outcomes. Generally, chest CT does not reach beyond L1. We aimed to determine whether opportunistic CT scan (chest CT)-determined skeletal muscle mass and quality at L1 are associated with mortality in initial-dialysis patients. This 3-year multicentric retrospective study included initial-dialysis patients from four centres between 2014 and 2017 in China. Unenhanced CT images of the L1 and L3 levels were obtained to assess skeletal muscle mass [by skeletal muscle index, (SMI), cm2 /m2 ] and quality [by skeletal muscle density (SMD), HU]. Skeletal muscle measures at L1 were compared with those at L3. The sex-specific optimal cutoff values of L1 SMI and L1 SMD were determined in relation to all-cause mortality. The outcomes were all-cause death and cardiac death. Cox regression models were applied to investigate the risk factors for death. A total of 485 patients were enrolled, of whom 257 had both L1 and L3 images. Pearson's correlation coefficient between L1 and L3 SMI was 0.84 (P<0.001), and that between L1 and L3 SMD was 0.90 (P<0.001). No significant association between L1 SMI and mortality was observed (P>0.05). Low L1 SMD (n=280, 57.73%) was diagnosed based on the optimal cutoff value (<39.56 HU for males and <33.06 HU for females). Multivariate regression analysis revealed that the low L1 SMD group had higher risks of all-cause death (hazard ratio 1.80; 95% confidence interval 1.05-3.11, P=0.034) and cardiac death (hazard ratio 3.74; 95% confidence interval 1.43-9.79, P=0.007). In initial-dialysis patients, there is high agreement between the L1 and L3 measures for SMI and SMD. Low SMD measured at L1, but not low SMI, is an independent predictor of both all-cause death and cardiac death.

Development and external validation of the multichannel deep learning model based on unenhanced CT for differentiating fat-poor angiomyolipoma from renal cell carcinoma: a two-center retrospective study.

There are undetectable levels of fat in fat-poor angiomyolipoma. Thus, it is often misdiagnosed as renal cell carcinoma. We aimed to develop and evaluate a multichannel deep learning model for differentiating fat-poor angiomyolipoma (fp-AML) from renal cell carcinoma (RCC). This two-center retrospective study included 320 patients from the First Affiliated Hospital of Sun Yat-Sen University (FAHSYSU) and 132 patients from the Sun Yat-Sen University Cancer Center (SYSUCC). Data from patients at FAHSYSU were divided into a development dataset (n = 267) and a hold-out dataset (n = 53). The development dataset was used to obtain the optimal combination of CT modality and input channel. The hold-out dataset and SYSUCC dataset were used for independent internal and external validation, respectively. In the development phase, models trained on unenhanced CT images performed significantly better than those trained on enhanced CT images based on the fivefold cross-validation. The best patient-level performance, with an average area under the receiver operating characteristic curve (AUC) of 0.951 ± 0.026 (mean ± SD), was achieved using the "unenhanced CT and 7-channel" model, which was finally selected as the optimal model. In the independent internal and external validation, AUCs of 0.966 (95% CI 0.919-1.000) and 0.898 (95% CI 0.824-0.972), respectively, were obtained using the optimal model. In addition, the performance of this model was better on large tumors (≥ 40mm) in both internal and external validation. The promising results suggest that our multichannel deep learning classifier based on unenhanced whole-tumor CT images is a highly useful tool for differentiating fp-AML from RCC.