Performance of artificial intelligence-based software for the automatic detection of lung lesions on chest radiographs of patients with suspected lung cancer.

PURPOSE: This study aimed to evaluate the performance of the commercially available artificial intelligence-based software CXR-AID for the automatic detection of pulmonary nodules on the chest radiographs of patients suspected of having lung cancer. MATERIALS AND METHODS: This retrospective study included 399 patients with clinically suspected lung cancer who underwent CT and chest radiography within 1 month between June 2020 and May 2022. The candidate areas on chest radiographs identified by CXR-AID were categorized into target (properly detected areas) and non-target (improperly detected areas) areas. The non-target areas were further divided into non-target normal areas (false positives for normal structures) and non-target abnormal areas. The visibility score, characteristics and location of the nodules, presence of overlapping structures, and background lung score and presence of pulmonary disease were manually evaluated and compared between the nodules detected or undetected by CXR-AID. The probability indices calculated by CXR-AID were compared between the target and non-target areas. RESULTS: Among the 450 nodules detected in 399 patients, 331 nodules detected in 313 patients were visible on chest radiographs during manual evaluation. CXR-AID detected 264 of these 331 nodules with a sensitivity of 0.80. The detection sensitivity increased significantly with the visibility score. No significant correlation was observed between the background lung score and sensitivity. The non-target area per image was 0.85, and the probability index of the non-target area was lower than that of the target area. The non-target normal area per image was 0.24. Larger and more solid nodules exhibited higher sensitivities, while nodules with overlapping structures demonstrated lower detection sensitivities. CONCLUSION: The nodule detection sensitivity of CXR-AID on chest radiographs was 0.80, and the non-target and non-target normal areas per image were 0.85 and 0.24, respectively. Larger, solid nodules without overlapping structures were detected more readily by CXR-AID.

Deep learning-based cardiothoracic ratio measurement on chest radiograph: accuracy improvement without self-annotation.

Background: A reproducible and accurate automated approach to measuring cardiothoracic ratio on chest radiographs is warranted. This study aimed to develop a deep learning-based model for estimating the cardiothoracic ratio on chest radiographs without requiring self-annotation and to compare its results with those of manual measurements. Methods: The U-net architecture was designed to segment the right and left lungs and the cardiac shadow, from chest radiographs. The cardiothoracic ratio was then calculated using these labels by a mathematical algorithm. The initial model of deep learning-based cardiothoracic ratio measurement was developed using open-source 247 chest radiographs that had already been annotated. The advanced model was developed using a training dataset of 729 original chest radiographs, the labels of which were generated by the initial model and then screened. The cardiothoracic ratio of the two models was estimated in an independent test set of 120 original cases, and the results were compared to those obtained through manual measurement by four radiologists and the image-reading reports. Results: The means and standard deviations of the cardiothoracic ratio were 52.4% and 9.8% for the initial model, 51.0% and 9.3% for the advanced model, and 49.8% and 9.4% for the total of four manual measurements, respectively. The intraclass correlation coefficients (ICCs) of the cardiothoracic ratio ranged from 0.91 to 0.93 between the advanced model and the manual measurements, whereas those for the initial model and the manual measurements ranged from 0.77 to 0.82. Conclusions: Deep learning-based cardiothoracic ratio estimation on chest radiographs correlated favorably with the results obtained through manual measurements by radiologists. When the model was trained on additional local images generated by the initial model, the correlation with manual measurement improved even more than the initial model alone.

Comparison of 18 F-FDG PET Findings of Pegfilgrastim-Induced Aortitis With Other Types of Large-Vessel Vasculitis : A Retrospective Observational Study.

PURPOSE OF THE REPORT: To elucidate the PET/CT findings of pegfilgrastim-induced aortitis (PFIA) and compare them with those of other large-vessel vasculitis. METHODS: We enrolled 45 patients diagnosed with the following: PFIA, n = 8; Takayasu arteritis (TA), n = 12; giant cell arteritis (GCA), n = 6; and immunoglobulin G4-related aortitis (IgG4-A), n = 19. Records of PET/CT performed before treatment initiation were collected. The aorta and its branches were divided into 16 anatomic regions. Presence of abnormal 18 F-FDG uptake in each region was determined and measured. RESULTS: The 18 F-FDG-positive areas of PFIA were distributed in the regions of the ascending aorta to the suprarenal abdominal aorta, cervical branches of the aorta, and external iliac arteries, similar to those of TA. However, TA had a higher proportion of 18 F-FDG-positive areas than PFIA in almost all anatomic regions. These areas of GCA were widespread throughout the entire aorta and the upper and lower limbs, whereas those of IgG4-A were observed from the abdominal aorta to iliac arteries. SUV max , SUV peak , metabolic volume, and total lesion glycolysis were higher in GCA than in PFIA, TA, and IgG4-A. CONCLUSIONS: Pegfilgrastim-induced aortitis distribution on PET/CT was frequently observed in the aorta, cervical branches, and extra iliac arteries. The low proportion of 18 F-FDG-positive areas in PFIA was different from that of TA, GCA, and IgG4-A. These findings may help identify and differentiate various aortitis types in clinical practice.

Fat fraction and R2 * values of various liver masses: Initial experience with 6-point Dixon method on a 3T MRI system.

Purpose: To assess the feasibility of the 6-point Dixon method for evaluating liver masses. We also report our initial experience with the quantitative values in various liver masses on a 3T system. Materials and methods: Of 251 consecutive patients for whom 6-point Dixon was employed in abdominal magnetic resonance imaging scans between October 2020 and October 2021, 117 nodules in 117 patients with a mass diameter of more than 1 cm were included in the study. Images for measuring the proton density fat fraction (PDFF) and R2 * values were obtained using the iterative decomposition of water and fat with echo asymmetry and least-squares estimation-quantitative technique for liver imaging. Two radiologists independently measured PDFF (%) and R2 * (Hz). Inter-reader agreement and the differences between readers were examined using intra-class correlation coefficient (ICC) and the Bland-Altman method, respectively. PDFF and R2 * values in differentiating liver masses were examined. Results: The masses included hepatocellular carcinoma (n = 59), cyst (n = 20), metastasis (n = 14), hemangioma (n = 8), and others (n = 16). The ICCs for the region of interest (mm2), PDFF, and R2 * were 0.988 (95 % confidence interval (CI): 0.983, 0.992), 0.964 (95 % CI: 0.949, 0.975), and 0.962 (95 % CI: 0.941, 0.975), respectively. The differences of measurements between the readers showed that 5.1 % (6/117) and 6.0% (7/117) for PDFF and R2 * , respectively, were outside the 95 % CI. Conclusion: Our observation indicates that the 6-point Dixon method is applicable to liver masses.

Deep learning-based automatic detection for pulmonary nodules on chest radiographs: The relationship with background lung condition, nodule characteristics, and location.

PURPOSE: Computer-aided diagnosis (CAD), which assists in the interpretation of chest radiographs, is becoming common. However, few studies have evaluated the benefits and pitfalls of CAD in the real world. This study aimed to evaluate the independent performance of commercially available deep learning-based automatic detection (DLAD) software, EIRL Chest X-ray Lung Nodule, in a cohort that included patients with background pulmonary abnormalities often encountered in clinical situations. METHODS: Patients with clinically suspected lung cancer for whom chest radiography was performed within a month before or after CT scan between June 2020 and May 2022 in our institution were enrolled. The reference standard was created using a bounding box annotated by two radiologists with reference to the CT. The visibility score, characteristics, location of the pulmonary nodules, presence of overlapping structures or pulmonary disease, and background lung score were manually determined. RESULTS: We included 388 patients. The DLAD software detected 222 of the 322 nodules visible on manual evaluation, with a sensitivity of 0.689 and a false-positive rate of 0.168. The detectability of the DLAD software was significantly lower for small and subsolid and nodules with overlapping structures. The visibility score and sensitivity of detection by the DLAD software were positively correlated. The relationship between the background lung score and detection by the DLAD software was unclear. CONCLUSION: The standalone performance of DLAD in detecting pulmonary nodules exhibited a sensitivity of 0.689 and a false-positive rate of 0.168. Understanding the characteristics of DLAD is crucial when interpreting chest radiographs with the assistance of the DLAD.

Computer-Aided Detection of Subsolid Nodules on Chest Computed Tomography: Assessment of Visualization on Vessel-Suppressed Images.

OBJECTIVES: This study aimed to clarify the performance of automatic detection of subsolid nodules by commercially available software on computed tomography (CT) images of various slice thicknesses and compare it with visualization on the accompanying vessel-suppression CT (VS-CT) images. METHODS: A total of 95 subsolid nodules from 84 CT examinations of 84 patients were included. The reconstructed CT image series of each case with 3-, 2-, and 1-mm slice thicknesses were loaded into a commercially available software application (ClearRead CT) for automatic detection of subsolid nodules and generation of VS-CT images. Automatic nodule detection sensitivity was assessed for 95 nodules on each series of images acquired at 3 slice thicknesses. Four radiologists subjectively evaluated visual assessment of the nodules on VS-CT. RESULTS: ClearRead CT automatically detected 69.5% (66/95 nodules), 68.4% (65/95 nodules), and 70.5% (67/95 nodules) of all subsolid nodules in 3-, 2-, and 1-mm slices, respectively. The detection rate was higher for part-solid nodules than for pure ground-glass nodules at all slice thicknesses. In the visualization assessment on VS-CT, 3 nodules at each slice thickness (3.2%) were judged as invisible, while 26 of 29 (89.7%), 27 of 30 (90.0%), and 25 of 28 (89.3%) nodules, which were missed by computer-aided detection, were judged as visible in 3-, 2-, and 1-mm slices, respectively. CONCLUSIONS: The automatic detection rate of subsolid nodules by ClearRead CT was approximately 70% at all slice thicknesses. More than 95% of subsolid nodules were visualized on VS-CT, including nodules undetected by the automated software. Computed tomography acquisition at slices thinner than 3 mm did not confer any benefits.

Can solid pseudopapillary neoplasm of the pancreas without degeneration be diagnosed with imaging? a comparison study of the solid component of solid pseudopapillary neoplasm, neuroendocrine neoplasm, and ductal adenocarcinoma.

PURPOSE: To investigate the MR findings of the solid components within pancreatic solid pseudopapillary neoplasms (SPNs) to characterize solid SPN without degeneration. METHODS: After case matching, 23 patients with SPNs, 23 with pancreatic neuroendocrine neoplasms (PNENs), and 46 pancreatic ductal adenocarcinomas (PDACs) were included in this retrospective comparative study. The MR findings of the solid components within the pancreatic tumors were assessed qualitatively and semi-quantitatively. RESULTS: In the qualitative assessment, significant differences were noted in T2-weighted imaging and MR cholangiopancreatography (MRCP). SPNs with a score of 4-5 (iso- to hyper-intense compared with the renal cortex) were observed in 18/19 (94.7%) by reader 1 and 15/19 (78.9%) by reader 2 (score 5, 52.6% and 47.4%) on fast spin-echo (FSE) T2-weighted imaging. On MRCP, the two readers identified 12 (63.2%) and 8 (42.1%) SPNs, respectively. The semi-quantitative signal-intensity ratio (SIR, signal intensity of tumor/signal intensity of the pancreatic parenchyma) of SPNs on FSE T2-weighted imaging was significantly higher (mean, 1.99-2.01) than that of PNENs (1.30-1.31) or PDACs (1.26-1.28). The sensitivity/specificity of 'hyper' on T2-weighted imaging (qualitative score of 4-5, or SIR of ≥ 1.5) were 78.9-100.0%/63.8-79.7%. The sensitivity/specificity of 'remarkably hyper' (score of 5, SIR of ≥ 2.0, or visible on MRCP) or salt-and-pepper pattern were 36.8-68.4%/85.5-98.6%. CONCLUSION: T2-weighted imaging may be the key sequence for solid SPN. Solid tumors with hyper-intensity on T2-weighted imaging (especially, more hyper-intense than the renal cortex, more than twice the signal of the pancreatic parenchyma, depicted on MRCP, or salt-and-pepper appearance) may be suspected to be SPNs.

Renal artery-based kidney segmentation on CT for patients with renal cell carcinoma: Feasibility of segmental artery clamping simulation.

Purpose: To evaluate the feasibility of renal artery-based segmentation of kidneys with renal cell carcinoma (RCC), based on three-dimensional (3D) software for the simulation of segmental artery clamping (SAC), and to correlate it with RENAL nephrometry score. Methods: Fifty RCCs (< 4 cm) identified from a pathological database search between January 2015 and January 2018 were included retrospectively. On computed tomography (CT) images, the relevant kidney, tumor, and renal artery were annotated semi-automatically on the commercial workstation, and renal artery-based segmentation was performed using 3D Voronoi diagrams. Simulation of SAC was performed by a radiologist and urologist in consensus. The volume of the whole kidney and tumor and estimated rescued volume for possible SAC cases were calculated. The correlation between possible SAC and RENAL nephrometry score was investigated. The reproducibility of the calculation of each volume and the interrater reliability of SAC simulation were assessed. Results: In the anatomical analysis, 44 patients had a single main renal artery and six had two main renal arteries, and of these, an early division pattern was observed in 11 cases. In the 3D simulation software, 22 out of 50 cases (44 %) were determined as possible SAC. The agreement of the SAC simulation was excellent (kappa = 0.96). RENAL nephrometry score was significantly different in the anterior/posterior and exophytic/endophytic components between possible and impossible SAC groups. Conclusions: Renal artery-based segmentation of kidneys with RCC on CT images using 3D simulation software is feasible for effectively estimating the possibility of SAC with high reproducibility.

Development of a Prognostic Model of Overall Survival for Metastatic Hormone-Naïve Prostate Cancer in Japanese Men.

BACKGROUND: Treatment strategies have changed dramatically in recent years with the development of a variety of agents for metastatic hormone-naïve prostate cancer (mHNPC). There is a need to identify prognostic factors for the appropriate choice of treatment for patients with mHNPC, and we retrospectively examined these factors. METHODS: Patients with mHNPC treated at our institution from 2000 to 2019 were included in this study. Overall survival (OS) was estimated retrospectively using the Kaplan-Meier method, and factors associated with OS were identified using univariate and multivariate analyses. A prognostic model was then developed based on the factors identified. Follow-up was terminated on 24 October 2021. RESULTS: The median follow-up duration was 44.2 months, whereas the median OS was 85.2 months, with 88 patients succumbing to their disease. Multivariate analysis identified Gleason pattern (GP) 5 content, bone scan index (BSI) ≥ 1.5, and lactate dehydrogenase (LDH) levels ≥ 300 IU/L as prognostic factors associated with OS. We also developed a prognostic model that classified patients with mHNPC as low risk with no factor, intermediate risk with one factor, and high risk with two or three factors. CONCLUSIONS: Three prognostic factors for OS were identified in patients with mHNPC, namely GP5 inclusion, BSI ≥ 1.5, and LDH ≥ 300. Using these three factors, we developed a new prognostic model for OS that can more objectively predict patient prognosis.

Is It Me or the Robot? A Critical Evaluation of Human Affective State Recognition in a Cognitive Task.

A key goal in human-robot interaction (HRI) is to design scenarios between humanoid robots and humans such that the interaction is perceived as collaborative and natural, yet safe and comfortable for the human. Human skills like verbal and non-verbal communication are essential elements as humans tend to attribute social behaviors to robots. However, aspects like the uncanny valley and different technical affinity levels can impede the success of HRI scenarios, which has consequences on the establishment of long-term interaction qualities like trust and rapport. In the present study, we investigate the impact of a humanoid robot on human emotional responses during the performance of a cognitively demanding task. We set up three different conditions for the robot with increasing levels of social cue expressions in a between-group study design. For the analysis of emotions, we consider the eye gaze behavior, arousal-valence for affective states, and the detection of action units. Our analysis reveals that the participants display a high tendency toward positive emotions in presence of a robot with clear social skills compared to other conditions, where we show how emotions occur only at task onset. Our study also shows how different expression levels influence the analysis of the robots' role in HRI. Finally, we critically discuss the current trend of automatized emotion or affective state recognition in HRI and demonstrate issues that have direct consequences on the interpretation and, therefore, claims about human emotions in HRI studies.

Single-Center Analysis of Pegfilgrastim-induced Aortitis Using a Drug Prescription Database and CT Findings.

Background Pegfilgrastim-induced aortitis is a rare but serious adverse event in patients undergoing anticancer therapy with granulocyte colony-stimulating factor analogs. Despite previous case series and systemic reviews, the exact incidence, clinical presentation, and CT manifestations of pegfilgrastim-induced aortitis remain unclear. Purpose To clarify the incidence and clinicoradiologic characteristics of pegfilgrastim-induced aortitis. Materials and Methods Pegfilgrastim administration records from January 2015 to March 2021 were retrospectively collected from the drug prescription database of a single center and were matched with the relevant findings in the CT database. Corresponding CT images within 6 months were available for a total of 1462 doses of pegfilgrastim in 674 patients. Four radiologists reviewed the CT images for the presence of aortitis in two steps. Clinical information and the distribution of aortitis on CT images were examined for patients with a diagnosis of pegfilgrastim-induced aortitis. Results Pegfilgrastim-induced aortitis was observed in 18 of 674 patients (mean age, 62 years ± 13 [SD]; 424 men), resulting in incidence rates of 2.7% per patient (95% CI: 1.6, 4.2) and 1.2% per dose (95% CI: 0.7, 1.9). The most common original primary malignancies were esophageal cancer (n = 10, 9%), breast cancer (n = 3, 4%), and pancreatic cancer (n = 2, 2%). The most common anticancer drugs used at onset were 5-fluorouracil, cisplatin, and docetaxel. Seven cases were symptomatic, while the remaining 11 (61%) were asymptomatic. CT findings indicated that aortitis involved branches of the aortic arch in 13 cases (72%), aortic arch in 10 cases (56%), and abdominal aorta in two cases (11%). Conclusion Pegfilgrastim-induced aortitis may be more prevalent than previously reported and may be more common in patients with esophageal cancer and those who received 5-fluorouracil, cisplatin, and docetaxel as anticancer drugs. The findings also suggest that pegfilgrastim-induced aortitis is often characterized by aortic arch and proximal branch involvement at CT. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Krinsky in this issue.

Prediction of pathological complete response to neoadjuvant chemotherapy in breast cancer using radiomics of pretreatment dynamic contrast-enhanced MRI.

PURPOSE: To investigate if the pretreatment dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI)-based radiomics machine learning predicts the pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer patients. METHODS: Seventy-eight breast cancer patients who underwent DCE-MRI before NAC and confirmed as pCR or non-pCR were enrolled. Early enhancement mapping images of pretreatment DCE-MRI were created using subtraction formula as follows: Early enhancement mapping = (Signal 1 min - Signal pre)/Signal pre. Images of the whole tumors were manually segmented and radiomics features extracted. Five prediction models were built using five scenarios that included clinical information, subjective radiological findings, first order texture features, second order texture features, and their combinations. In texture analysis workflow, the corresponding variables were identified by mutual information for feature selection and random forest was used for model prediction. In five models, the area under the receiver operating characteristic curves (AUC) to predict the pCR and several metrics for model evaluation were analyzed. RESULTS: The best diagnostic performance based on F-score was achieved when both first and second order texture features with clinical information and subjective radiological findings were used (AUC = 0.77). The second best diagnostic performance was achieved with an AUC of 0.76 for first order texture features followed by an AUC of 0.76 for first and second order texture features. CONCLUSIONS: Pretreatment DCE-MRI can improve the prediction of pCR in breast cancer patients when all texture features with clinical information and subjective radiological findings are input to build the prediction model.

Urinary collecting system invasion on multiphasic CT in renal cell carcinomas: prevalence, characteristics, and clinical significance.

PURPOSE: The aim of this study was to determine the prevalence of collecting system invasion (CSI) on multiphasic CT, validate the pathological findings, and investigate the relationship between CSI and clinical outcomes in patients with renal cell carcinomas (RCC). METHODS: Patients pathologically diagnosed with RCC between January 2008 and December 2017 were retrospectively enrolled in this study. They were divided into two groups according to the presence of CSI on multiphasic CT images. Patients' clinical characteristics, radiological findings, and overall survival (OS) and recurrence-free survival (RFS) rates were analyzed and compared between the groups. In addition, the correlation of radiological findings with pathological findings was investigated. RESULTS: Among the included 347 kidneys of 340 patients, CSI was observed in 11 kidneys (3%; 95% confidence interval, 1.3-5.0%). In all the 11 kidneys, the tumors were pathologically diagnosed as clear cell RCC, and in one kidney, the tumor also had sarcomatoid features. When pathological CSI served as the standard of reference, the sensitivity, specificity, and accuracy of CSI on CT were 50%, 99.7%, and 97.1%, respectively. The OS and RFS rates were not significantly different between patients with CSI on CT and those without CSI. CONCLUSION: This study found that the prevalence of RCC-related CSI was 3%. Because of the low prevalence, we cannot exclude the possibility that CSI on CT would be associated with the OS and RFS. Further studies are needed to determine whether CSI on CT can be an independent prognostic factor for survival in patients with RCC.

[A case of body lateropulsion to the left in acute cerebral infarction of the right medial parietal lesion].

A 68-year-old right-handed woman with acute-onset inability to stand was admitted to our department. Although left hemiparesis was minor, the neurological examination on admission showed marked body lateropulsion (BL) to the left when she stood or stepped with eyes open and feet closed. Neither ataxia nor sensory disturbance was present. Brain MRI and 3D-CT angiography revealed infarction of the right posterior cingulate and the precuneus due to dissection of the right anterior cerebral artery. BL improved on day 10 and she was discharged without sequelae on day 26. BL caused by cerebral lesions is rare, and we should recognize that infarction of the posterior cingulate and/or the precuneus can cause BL.