Shortcut learning in medical AI hinders generalization: method for estimating AI model generalization without external data.

Healthcare datasets are becoming larger and more complex, necessitating the development of accurate and generalizable AI models for medical applications. Unstructured datasets, including medical imaging, electrocardiograms, and natural language data, are gaining attention with advancements in deep convolutional neural networks and large language models. However, estimating the generalizability of these models to new healthcare settings without extensive validation on external data remains challenging. In experiments across 13 datasets including X-rays, CTs, ECGs, clinical discharge summaries, and lung auscultation data, our results demonstrate that model performance is frequently overestimated by up to 20% on average due to shortcut learning of hidden data acquisition biases (DAB). Shortcut learning refers to a phenomenon in which an AI model learns to solve a task based on spurious correlations present in the data as opposed to features directly related to the task itself. We propose an open source, bias-corrected external accuracy estimate, PEst, that better estimates external accuracy to within 4% on average by measuring and calibrating for DAB-induced shortcut learning.

Chest CT at X-Ray Dose Using a Noise-Mitigating Weighted Projection: The Thoracic Tomogram. Diagnostic Performance for Pneumonia Detection in Hemato-Oncology Patients.

PURPOSE: To compare the diagnostic performance of a thick-slab reconstruction obtained from an ultra-low-dose CT (termed thoracic tomogram) with standard-of-care low-dose CT (SOC-CT) for rapid interpretation and detection of pneumonia in hemato-oncology patients. METHODS: Hemato-oncology patients with a working diagnosis of pneumonia underwent an SOC-CT followed by an ultra-low-dose CT, from which the thoracic tomogram (TT) was reconstructed. Three radiologists evaluated the TT and SOC-CT in the following categories: (I) infectious/inflammatory opacities, (II) small airways infectious/inflammatory changes, (III) atelectasis, (IV) pleural effusions, and (V) interstitial abnormalities. The TT interpretation time and radiation dose were recorded. Sensitivity, specificity, diagnostic accuracy, ROC, and AUC were calculated with the corresponding power analyses. The agreement between TT and SOC-CT was calculated by Correlation Coefficient for Repeated Measures (CCRM), and the Shrout-Fleiss intra-class correlations test was used to calculate interrater agreement. RESULTS: Forty-seven patients (mean age 58.7 ± 14.9 years; 29 male) were prospectively enrolled. Sensitivity, specificity, accuracy, AUC, and Power for categories I/II/III/IV/V were: 94.9/99/97.9/0.971/100, 78/91.2/86.5/0.906/100, 88.6/100/97.2/0.941/100, 100/99.2/99.3/0.995/100, and 47.6/100/92.2/0.746/87.3. CCRM between TT and SOC-CT for the same categories were .97/.81/.92/.96/.62 with an interobserver agreement of .93/.88/.82/.96/.61. Mean interpretation time was 18.6 ± 5.4 seconds. The average effective radiation dose of TT was similar to a frontal and lateral chest X-ray (0.27 ± 0.08 vs 1.46 ± 0.64 mSv for SOC-CT; P < .01). CONCLUSION: Thoracic tomograms provide comparable diagnostic information to SOC-CT for the detection of pneumonia in immunocompromised patients at one-fifth of the radiation dose with high interobserver agreement.

Manual aspiration of a pneumothorax after CT-guided lung biopsy: outcomes and risk factors.

OBJECTIVE: Quantify the outcomes following pneumothorax aspiration and influence upon chest drain insertion. METHODS: This was a retrospective cohort study of patients who underwent aspiration for the treatment of a pneumothorax following a CT percutaneous transthoracic lung biopsy (CT-PTLB) from January 1, 2010 to October 1, 2020 at a tertiary center. Patient, lesion and procedural factors associated with chest drain insertion were assessed with univariate and multivariate analyses. RESULTS: A total of 102 patients underwent aspiration for a pneumothorax following CT-PTLB. Overall, 81 patients (79.4%) had a successful pneumothorax aspiration and were discharged home on the same day. In 21 patients (20.6%), the pneumothorax continued to increase post-aspiration and required chest drain insertion with hospital admission. Significant risk factors requiring chest drain insertion included upper/middle lobe biopsy location [odds ratio (OR) 6.46; 95% CI 1.77-23.65, p = 0.003], supine biopsy position (OR 7.06; 95% CI 2.24-22.21, p < 0.001), emphysema (OR 3.13; 95% CI 1.10-8.87, p = 0.028), greater needle depth ≥2 cm (OR 4.00; 95% CI 1.44-11.07, p = 0.005) and a larger pneumothorax (axial depth ≥3 cm) (OR 16.00; 95% CI 4.76-53.83, p < 0.001). On multivariate analysis, larger pneumothorax size and supine position during biopsy remained significant for chest drain insertion. Aspiration of a larger pneumothorax (radial depths ≥3 cm and ≥4 cm) had a 50% rate of success. Aspiration of a smaller pneumothorax (radial depth 2-3 cm and <2 cm) had an 82.6% and 100% rate of success, respectively. CONCLUSION: Aspiration of pneumothorax after CT-PTLB can help reduce chest drain insertion in approximately 50% of patients with larger pneumothoraces and even more so with smaller pneumothoraces (>80%). ADVANCES IN KNOWLEDGE: Aspiration of pneumothoraces up to 3 cm was often associated with avoiding chest drain insertion and allowing for earlier discharge.

Utility of anti-neutrophil cytoplasmic antibody screening in idiopathic interstitial lung disease.

BACKGROUND: Interstitial lung disease (ILD) is an established manifestation of anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis. Autoimmune serologic screening is recommended by international consensus guidelines during the evaluation of idiopathic ILD, but ANCA testing only on a case-by-case basis. OBJECTIVE: We aimed to evaluate the role of ANCA screening in patients with idiopathic ILD. METHODS: We performed a retrospective review of patients seen between September 2015 and April 2017 in the ILD clinic at Toronto General Hospital. Patients referred with confirmed or suspected connective tissue disease were excluded. Patient demographics, symptoms, chest imaging, and pulmonary function testing was collected. We performed descriptive statistics based on the presence of ANCAs and estimated operating characteristics for ANCA screening. RESULTS: In total, 360 patients with idiopathic ILD were reviewed, 159 met study inclusion criteria and 4 (2.5%) tested positive for ANCAs. Two patients (1.2%) had elevated myeloperoxidase-ANCAs (MPO-ANCA) and 2 (1.2%) had elevated proteinase-3-ANCAs (PR3-ANCA). There were no significant associations between patient demographics and ANCAs. One patient (0.6%) with PR3-ANCAs was diagnosed with vasculitis following rheumatologic evaluation. Despite negative ANCA testing, 1 patient (0.6%) was diagnosed with vasculitis following rheumatologic evaluation. The sensitivity and specificity of ANCA screening for vasculitis in patients with ILD was calculated as 50% (95% CI, 1.3%-98.7%) and 98% (95%CI, 4.4-155.5) respectively. Negative and positive likelihood ratios were 0.5 (95%CI 0.1-2.0) and 26.2 (95%CI 4.4-155.5) respectively. CONCLUSION: ANCA screening in patients with idiopathic ILD rarely yields positive results. These results support an individualized approach to ANCA testing as opposed to widespread screening.

Comparison of conventional chest x ray with a novel projection technique for ultra-low dose CT.

PURPOSE: To compare a novel thick-slab projection technique for ultra-low dose computed tomography (CT; thoracic tomogram) with conventional chest x ray with respect to 13 diagnostic categories. METHODS: With the approval of the institutional ethics board, a dataset was retrospectively collected of 22 consecutive patients who had undergone a clinically requested emergency room conventional chest x ray (CXR) and a same-day standard-of-care non-contrast CT. Scanner specific noise was added to the CT images to simulate a target dose of 0.18 mSv. A novel algorithm was used to post-process CT images as coronal isotropic reformats by applying a voxel-based, locally normalized weighted-intensity projection to generate 2 cm thick slabs with 1 cm overlap. Three chest radiologists with no prior training for the study reviewed the CXR and thoracic tomogram for each case and assessed each diagnostic category (pneumonic infiltrates, pulmonary edema, interstitial lung disease, nodules > 5 mm, nodules < 5 mm, pleural effusion, pericardial effusion, heart size, acute bone fractures, foreign bodies, pneumothorax, mediastinal vessel diameter, free abdominal air) on a Likert scale from -4 (definitely absent/normal) to +4 (definitely present/abnormal). MRMC ROC curves were generated for each category. Time for interpretation and subjective image quality score (0-10) were also assessed. RESULTS: For focal lung disease (pneumonic infiltrates, nodules < 5 mm, nodules > 5mm), the area under the ROC curve (AUC) was significantly higher for thoracic tomograms than CXR (0.803 vs 0.648, respectively, P = 0.02). For non-focal lung disease (pulmonary edema, interstitial lung disease) and effusions (pulmonary, pericardial), the AUC was larger for thoracic tomograms than CXR but the difference did not reach significance (0.870 vs 0.833, P = 0.141; and 0.823 vs 0.752, P = 0.296, respectively). For acute bone fractures and foreign bodies, the AUC was smaller for thoracic tomograms than CXR, the difference was however not significant (0.491 vs 0.532, P = 0.42; and 0.871 vs 0.971, P = 0.39, respectively). Other diagnostic categories had no true positive cases in the dataset. The mean time for interpretation for each was 36.9 and 24.0 s with standard deviations of 0.857 and 5.977. The image quality score for each was 8.2 and 7.8 with standard deviations of 0.970 and 1.614. CONCLUSION: Thoracic tomograms were found to be diagnostically superior to CXR for focal lung disease, at no increased radiation dose. The thoracic tomogram presents an opportunity to improve the standard-of-care for patients who would otherwise receive a conventional CXR.

Screening for Myositis Antibodies in Idiopathic Interstitial Lung Disease.

PURPOSE: International guidelines recommend screening for connective tissue disease (CTD) with autoantibodies when evaluating patients with idiopathic interstitial lung disease (ILD). Idiopathic inflammatory myositis comprises of a subgroup of CTD diagnosed with myositis antibodies (MA), often presenting with ILD. Our aim was to evaluate the utility of MA screening in patients with idiopathic ILD. METHODS: A retrospective analysis was conducted on patients referred with idiopathic ILD to a tertiary centre ILD clinic who were screened for MA. Patients with known or suspected CTD were excluded. Descriptive statistics, univariate analysis and multivariable logistic regression were used to detect associations between MA and patient characteristics. RESULTS: Of 360 patients, 165 met inclusion criteria and 44 (26.7%) were identified to have MA. Fourteen patients (8.5%) had a change in diagnosis as a result of MA screening. Multivariable logistic regression identified the presence of MA to be associated with current smoking [OR 6.87 (1.65-28.64), p = 0.008] and a diffusing capacity of < 70% predicted [OR 2.55 (1.09-5.97), p = 0.03]. In patients with a change in diagnosis due to MA screening, 3 (1.8%) underwent a surgical lung biopsy and 2 (1.2%) were previously treated with antifibrotic therapy. CONCLUSIONS: Screening for MA in patients with idiopathic ILD can contribute to a change in patient diagnosis, and may prevent invasive testing and unproven use of antifibrotic therapy. These results support the addition of MA to CTD screening panels during the initial evaluation of idiopathic ILD.

Helical CT with variable target noise levels for dose reduction in chest, abdomen and pelvis CT.

OBJECTIVES: To evaluate a contiguous helical CT protocol with two different target noise levels in chest/abdomen/pelvis CT. METHODS: 41 patients (study group) underwent a helical scan (P1) with two different target noise levels (SDs), SD = 16 for chest and SD = 13 for abdomen/pelvis. Two further protocols were planned but not executed: a single helical scan with only one SD (SD = 13) for the entire scan range (P2), and two separate helical scans overlapping over the liver and same SD settings as for P1 (P3). All DLPs were recorded. Image quality was assessed qualitatively and quantitatively on all scans. The control group consisted of 40 patients, was scanned with protocol P3 and analysed using the same metrics. RESULTS: DLPs (mean/SD) for P1, P2 and P3 were 859.5/392.9, 1040.2/510.5 and 1027.4/469.4, respectively. P1 offered a mean dose reduction of 17.4% compared to P2, and 16.3% compared to P3 (both p < 0.001). There were no differences in image quality between both patient groups (p > 0.3). CONCLUSION: Contiguous helical scanning of the chest/abdomen/pelvis with variable target noise levels results in approximately 17% dose reduction if compared to a single acquisition with only abdominal dose settings or two separate acquisitions of the chest and abdomen/pelvis. KEY POINTS: • Low dose chest and standard abdomen CTs can be combined. • Variable SD CT scanning allows for radiation dose reduction. • Variable SD CT scanning maintains image quality.

Impact of Pretreatment Interstitial Lung Disease on Radiation Pneumonitis and Survival in Patients Treated With Lung Stereotactic Body Radiation Therapy (SBRT).

INTRODUCTION: The purpose of this study was to determine the impact of interstitial lung disease (ILD) on radiation pneumonitis (RP) and overall survival (OS) in lung stereotactic body radiation therapy (SBRT). METHODS: Patients treated with lung SBRT from 2004 to 2015 were included. Pretreatment computed tomography scans were reviewed and classified for interstitial changes by thoracic radiologists using American Thoracic Society guidelines and Washko and Kazerooni scores. RP was scored prospectively using Common Terminology Criteria for Adverse Events, version 3.0. Pretreatment imaging characteristics, clinical variables, and dosimetry were assessed by univariate (UVA) and multivariate analysis (MVA). OS was assessed by the log-rank test, and the impact of ILD on OS was assessed by Cox regression. RESULTS: Of the 537 patients assessed, 39 had interstitial changes (13 usual interstitial pneumonia [UIP], 24 possible UIP, and 2 inconsistent with UIP). RP was significantly higher in patients with ILD than in patients without ILD (grade ≥ 2, 20.5% vs. 5.8%; P < .01; grade ≥ 3, 10.3% vs. 1.0%; P < .01). Two of 3 grade 5 RP had imaging features of ILD. On UVA, ILD, Washko score, lung parameters performance status, and dose were significant predictors of grade ≥ 2 RP. On MVA, ILD (odds ratio, 5.81; 95% confidence interval, 2.28-14.83; P < .01) and mean lung dose (odds ratio, 1.40; 95% confidence interval, 1.14-1.71; P < .01) were predictors of RP. ILD did not significantly affect OS on UVA or MVA. Median survival was 27.4 months in the ILD cohort and 34.8 in the ILD-negative cohort (P = .17). DISCUSSION: ILD is a significant risk factor for RP in patients treated with lung SBRT. Computed tomography scans should be reviewed for evidence of ILD prior to SBRT.

Utility of transthoracic needle biopsy after lung transplantation.

The purpose of this study was to assess the diagnostic yield and complications of CT-guided transthoracic needle biopsy (TTNB) after lung transplantation. A database search identified all TTNB performed in lung transplant patients over a 14-year period. Forty-two biopsies in transplant patients (transplant group) were identified and matched to the next biopsy performed in native lungs by the same operator (nontransplant group) as a control. Primary outcomes recorded were diagnosis, diagnostic yield, pneumothorax requiring intervention, and symptomatic pulmonary hemorrhage. Biopsy outcomes were classified as diagnostic, not specifically diagnostic, and nondiagnostic. Patients in the transplant group were younger (P < .002). Emphysema along the biopsy trajectory was more commonly seen in the nontransplant group (P < .0006). Needle gauge, size of lesion, pleural punctures, lesion depth, and number of passes were not significantly different. Diagnostic yield was 71% in the transplant group and 91% in the nontransplant group. There were 20 of 42 (48%) malignant nodules in the transplant group compared to 31 of 44 (70%) nodules in the nontransplant group (P = .05). There were no complications in the transplant group. The nontransplant group had two pneumothoraces requiring intervention. TTNB after lung transplant is safe with a moderate diagnostic yield. Nonmalignant lesions are more common after lung transplantation.

Changes of Emphysema Parameters over the Respiratory Cycle During Free Breathing: Preliminary Results Using Respiratory Gated 4D-CT.

The purpose of this research was to evaluate respiratory gated CT of the lung in patients with COPD for analysis of parenchymal characteristics who were potential candidates for volume reduction surgery. Eleven patients with clinically known emphysematous disease underwent a respiratory gated, free-breathing 64-multislice-CT (Aquilion 64, Toshiba). Retrospective image reconstruction was performed similar to cardiac CT at every 10% of the respiratory loop, resulting in 10 complete volumetric datasets at 10 equidistant time points. All images were transferred onto a PC for calculation of the total lung volume, emphysema volume, emphysema index, and mean lung density. Complete datasets could be successfully reconstructed in all patients. The mean lung volume increased from 6.9 L to 7.5 L over the respiratory cycle. Emphysema volume increased from 1.6 L to 2.0 L and emphysema index from 22.6% to 26.5% from expiration to inspiration. In conclusion, respiratory gated chest 4D-CT allows for combined morphologic and functional image analysis, which can provide new insight into functional impairment and individual treatment planning.

Multi-dimensional scores to predict mortality in patients with idiopathic pulmonary fibrosis undergoing lung transplantation assessment.

BACKGROUND: The heterogeneous progression of idiopathic pulmonary fibrosis (IPF) makes prognostication difficult and contributes to high mortality on the waitlist for lung transplantation (LTx). Multi-dimensional scores (Composite Physiologic index [CPI], [Gender-Age-Physiology [GAP]; RIsk Stratification scorE [RISE]) demonstrated enhanced predictive power towards outcome in IPF. The lung allocation score (LAS) is a multi-dimensional tool commonly used to stratify patients assessed for LTx. We sought to investigate whether IPF-specific multi-dimensional scores predict mortality in patients with IPF assessed for LTx. METHODS: The study included 302 patients with IPF who underwent a LTx assessment (2003-2014). Multi-dimensional scores were calculated. The primary outcome was 12-month mortality after assessment. LTx was considered as competing event in all analyses. RESULTS: At the end of the observation period, there were 134 transplants, 63 deaths, and 105 patients were alive without LTx. Multi-dimensional scores predicted mortality with accuracy similar to LAS, and superior to that of individual variables: area under the curve (AUC) for LAS was 0.78 (sensitivity 71%, specificity 86%); CPI 0.75 (sensitivity 67%, specificity 82%); GAP 0.67 (sensitivity 59%, specificity 74%); RISE 0.78 (sensitivity 71%, specificity 84%). A separate analysis conducted only in patients actively listed for LTx (n = 247; 50 deaths) yielded similar results. CONCLUSIONS: In patients with IPF assessed for LTx as well as in those actually listed, multi-dimensional scores predict mortality better than individual variables, and with accuracy similar to the LAS. If validated, multi-dimensional scores may serve as inexpensive tools to guide decisions on the timing of referral and listing for LTx.

Protective Capnothorax During Transthoracic Needle Biopsy.

PURPOSE: Transthoracic needle biopsy (TTNB) is an established procedure in the management of pulmonary nodules. The most common complications are directly related to crossing the lung or visceral pleura during the biopsy. In this study, we describe the use of carbon dioxide instead of room air to create a protective "capnothorax" during TTNB. MATERIALS AND METHODS: Five patients underwent creation of a capnothorax during TTNB. Parameters recorded were location and size of target, distance from pleura, length of procedure, volume of carbon dioxide, periprocedural complications and biopsy result. RESULTS: Induction of capnothorax was successful in all cases. In two patients, a continuous infusion of carbon dioxide was required to maintain an adequate volume of intrapleural gas. In two patients, the carbon dioxide resolved spontaneously and in the remaining patients it was aspirated at the end of the procedure. All biopsies were diagnostic with no periprocedural or postprocedural complications. CONCLUSION: This study suggests that protective iatrogenic capnothorax is a safe and effective technique during TTNB. The intrinsic properties and availability of carbon dioxide make it an attractive alternative to room air.

Detection of Early Tumor Response to Axitinib in Advanced Hepatocellular Carcinoma by Dynamic Contrast Enhanced Ultrasound.

This study aimed to evaluate the utility of dynamic contrast-enhanced ultrasound (DCE-US) in measuring early tumor response of advanced hepatocellular carcinoma to axitinib. Twenty patients were enrolled (aged 18-78 y; median 65). DCE-US was performed with bolus injection and infusion/disruption replenishment. Median overall survival was 7.1 mo (1.8-27.3) and progression free survival was 3.6 mo (1.8-17.4). Fifteen patients completed infusion scans and 12 completed bolus scans at 2 wk. Among the perfusion parameters, fractional blood volume at infusion (INFBV) decreased at 2 wk in 10/15 (16%-81% of baseline, mean 47%) and increased in 5/15 (116%-535%, mean 220%). This was not significantly associated with progression free survival (p = 0.310) or progression at 16 wk (p = 0.849), but was borderline statistically significant (p = 0.050) with overall survival, limited by a small sample size. DCE-US is potentially useful in measuring early tumor response of advanced hepatocellular carcinoma to axitinib, but a larger trial is needed.

Classification and Reporting of Late Radiographic Changes After Lung Stereotactic Body Radiotherapy: Proposing a New System.

UNLABELLED: Radiation-induced parenchymal lung changes after stereotactic body radiotherapy are common, and can obscure the primary tumor site. In this study we propose a structured radiographic reporting tool for characterization of these changes, pilot its feasibility in a group of radiation oncologists, and test the interrater agreement. We could demonstrate the applicability of the scale, with a fair to moderate agreement. BACKGROUND: The purpose of the study was to design and pilot a synoptic scale for characterization of late radiographic changes after lung stereotactic body radiotherapy (SBRT). PATIENTS AND METHODS: A participatory design process involving 6 radiation oncologists and 2 thoracic radiologists was used in the scale's design. Seventy-seven early-stage non-small-cell lung cancer patients who were treated with SBRT were included, and after treatment their serial computed tomography (CT) images were scored by 6 radiation oncologists. Gwet's First-order Agreement Coefficient (AC1) and a leave-one-out (LOO) analysis was used to assess interrater reliability and variability among raters, respectively. RESULTS: The scale reports on 5 independent categories including "tumor in primary site," "tumor in involved lobe," "consolidation," "volume loss," and "ground-glass or interstitial changes." At each time point, each category is reported as "increased," "stable," "decreased," "obscured," or "not present," compared with the previous. The total number of rated images for the pilot ranged from 450 at 6 months to 84 at 48 months. The primary tumor site was scored as obscured in 38% to 40% of ratings from 12 months onward; 3% to 5% of primary tumors were scored as "increased." Consolidation, volume loss, and ground-glass or interstitial changes were increasingly marked as "stable" with time. At 24 months, AC1 was 0.28 (LOO, 0.22-0.42), 0.47 (LOO, 0.39-0.72), 0.45 (LOO, 0.42-0.50), 0.21 (LOO, 0.15-0.26), and 0.25 (LOO, 0.20-0.38) for the 5 categories listed, respectively. CONCLUSION: In a population of clinicians, this scale could be implemented to characterize evolving lung changes after SBRT, and had fair to moderate interrater agreement. Obscured tumor site is a common challenge of follow-up CT imaging, and new imaging techniques should be explored. This scale provides a tool for communicating changes after SBRT.

How influential is the duration of contrast material bolus injection in perfusion CT? evaluation in a swine model.

PURPOSE: To analyze the effect of the duration of contrast material bolus injection on perfusion values in a swine model by using the maximum slope method. MATERIALS AND METHODS: This study was approved by the institutional animal care committee. Twenty pigs (weight range, 63-77 kg) underwent dynamic volume computed tomography (CT) of the kidneys during suspended respiration. Before the CT examination, a miniature cuff-shaped ultrasonographic flow probe encircling the right renal artery was surgically implanted in each pig to obtain true perfusion values. Two sequential perfusion CT series were performed in 30 seconds, each comprising 30 volumes with identical parameters (100 kV, 200 mAs, 0.5 sec rotation time). The duration of contrast material bolus (0.5 mL/kg of body weight) was 3.8 seconds in the first series (short bolus series) and 11.5 seconds in the second series (long bolus series), and the injection flow rate was adapted accordingly. In each pig, cortical kidney volume was determined by using the volume with the highest cortical enhancement. CT perfusion values were calculated for both series by using the maximum slope method and were statistically compared and correlated with the true perfusion values from the flow probe by using linear regression analysis. RESULTS: Mean true perfusion and CT perfusion values (in minutes(-1)) for the short bolus series were 1.95 and 2.03, respectively (P = .22), and for the long bolus series, they were 2.02 and 1.92, respectively (P = .12). CT perfusion showed very good correlation with true perfusion in both the short (slope, 1.01; 95% confidence interval: 0.91, 1.11) and long (slope, 0.92; 95% confidence interval: 0.78, 1.04) series. On the basis of the regression analysis, CT perfusion values in the short bolus series were overestimated by 1% and those in the long bolus series were underestimated by 8%. CONCLUSION: Duration of contrast material bolus injection does not influence CT perfusion values substantially. The longer, clinically preferred intravenous injection scheme is sufficiently accurate for CT perfusion.

Comparison of free breathing versus breath-hold in perfusion imaging using dynamic volume CT.

OBJECTIVES: To compare two scanning protocols (free breathing versus breath-hold) for perfusion imaging using dynamic volume computed tomography (CT) and to evaluate their effects on image registration. MATERIAL AND METHODS: Forty patients underwent dynamic volume CT for pancreatic perfusion analysis and were randomly assigned to either a shallow-breathing (I) or breath-hold (II) group. Both dynamic CT protocols consisted of 17 low-dose volumetric scans. Rigid image registration was performed by using the volume with highest aortic attenuation as reference. All other volumes were visually matched with the pancreatic lesion serving as the volumetric region of interest. The overall demand for post-processing per patient was calculated as the median of three-dimensional vector lengths of all volumes in relation to the relative patient origin. The number of volumes not requiring registration was recorded per group. RESULTS: Registration mismatch for groups I and II was 2.61 mm (SD, 1.57) and 4.95 mm (SD, 2.71), respectively (P < 0.005). Twenty-eight volumes in group I (8.2%) and 47 volumes in group II (14.1%) did not require manual registration (P = 0.014). CONCLUSION: Shallow breathing during dynamic volume CT scanning reduces the overall demand for motion correction and thus may be beneficial in perfusion imaging of the pancreas MAIN MESSAGES: • Shallow breathing during perfusion CT scanning reduces the overall demand for motion correction. • Shallow breathing may be beneficial in perfusion imaging of the pancreas. • Image registration is crucial for CT perfusion imaging.

Radiation dose optimisation in dynamic volume CT of the heart: tube current adaptation based on anterior-posterior chest diameter.

To compare tube current adaptation based on 3 body mass index (BMI) categories versus anterior-posterior chest diameter (APD) for radiation dose optimisation in patients undergoing dynamic volume cardiac CT. Two cardiac imaging centres participated in the study. 20 patients underwent a prospectively triggered 320-slice single beat cardiac CT using the X-ray tube current [mA] manually adjusted to the patient's BMI (group I). In 20 subsequent patients, the tube current was adapted according to the patient's APD (group II). All other parameters were kept constant. Image noise was defined as the standard deviation of attenuation values and measured using a ROI in the descending aorta. Variation in image noise was statistically compared between both patient groups. Average and standard deviation of pixel noise were 29.1 HU and 14.8 HU in group I and 28.0 HU and 4.2 HU in group II. Inter-individual variation of pixel noise was significantly lower in group II compared to group I (p < 0.0001). Tube current adaptation based on APD is superior to stepwise adaptation based on BMI for optimising radiation dose in dynamic volume cardiac CT and therefore limits unnecessary radiation dose while ensuring diagnostic image quality in patients with diverse body habitus.

Computer-aided detection in computed tomography colonography with full fecal tagging: comparison of standalone performance of 3 automated polyp detection systems.

PURPOSE: We sought to compare the performance of 3 computer-aided detection (CAD) polyp algorithms in computed tomography colonography (CTC) with fecal tagging. METHODS: CTC data sets of 33 patients were retrospectively analysed by 3 different CAD systems: system 1, MedicSight; system 2, Colon CAD; and system 3, Polyp Enhanced View. The polyp database comprised 53 lesions, including 6 cases of colorectal cancer, and was established by consensus reading and comparison with colonoscopy. Lesions ranged from 6-40 mm, with 25 lesions larger than 10 mm in size. Detection and false-positive (FP) rates were calculated. RESULTS: CAD systems 1 and 2 could be set to have varying sensitivities with higher FP rates for higher sensitivity levels. Sensitivities for system 1 ranged from 73%-94% for all lesions (78%-100% for lesions > or =10 mm) and, for system 2, from 64%-94% (78%-100% for lesions > or =10 mm). System 3 reached an overall sensitivity of 76% (100% for lesions > or =10 mm). The mean FP rate per patient ranged from 8-32 for system 1, from 1-8 for system 2, and was 5 for system 3. At the highest sensitivity level for all polyps (94%), system 2 showed a statistically significant lower FP rate compared with system 1 (P = .001). When analysing lesions > or =10 mm, system 3 had significantly fewer FPs than systems 1 and 2 (P < .012). CONCLUSIONS: Standalone CTC-CAD analysis in the selected patient collective showed the 3 systems tested to have a variable but overall promising performance with respect to sensitivity and the FP rate.