Validation study of machine-learning chest radiograph software in primary and emergency medicine.

AIM: To evaluate the performance of a machine learning based algorithm tool for chest radiographs (CXRs), applied to a consecutive cohort of historical clinical cases, in comparison to expert chest radiologists. MATERIALS AND METHODS: The study comprised 1,960 consecutive CXR from primary care referrals and the emergency department (992 and 968 cases respectively), obtained in 2015 at a UK hospital. Two chest radiologists, each with >20 years of experience independently read all studies in consensus to serve as a reference standard. A chest artificial intelligence (AI) algorithm, Lunit INSIGHT CXR, was run on the CXRs, and results were correlated with those by the expert readers. The area under the receiver operating characteristic curve (AUROC) was calculated for the normal and 10 common findings: atelectasis, fibrosis, calcification, consolidation, lung nodules, cardiomegaly, mediastinal widening, pleural effusion, pneumothorax, and pneumoperitoneum. RESULTS: The ground truth annotation identified 398 primary care and 578 emergency department datasets containing pathologies. The AI algorithm showed AUROC of 0.881-0.999 in the emergency department dataset and 0.881-0.998 in the primary care dataset. The AUROC for each of the findings between the primary care and emergency department datasets did not differ, except for pleural effusion (0.954 versus 0.988, p<0.001). CONCLUSIONS: The AI algorithm can accurately and consistently differentiate normal from major thoracic abnormalities in both acute and non-acute settings, and can serve as a triage tool.

Coronary artery calcification is associated with mortality independent of pulmonary embolism severity: a retrospective cohort study.

AIM: To assess coronary artery calcification (CAC) and vascular calcification in patients with pulmonary embolism (PE) and correlate this with mortality. MATERIALS AND METHODS: PE severity was quantified using computed tomography pulmonary angiography (CTPA) in 400 consecutive cases using the modified Miller score (1-5, mild; 6-11, moderate; 12-16, severe). Right ventricle strain was assessed using the right/left ventricle diameter (RV/LV) ratio. CAC score (CACS) was assessed using a four-point scale (CACS mild 1-3, moderate 4-8, severe 9-12) for each vessel and summed to give the total CACS. Follow-up for mortality was obtained at 3 years. RESULTS: PE severity was classified as mild in 48%, moderate in 21%, and severe in 32% of cases. The median modified Miller score was 6 (Interquartile range [IQR] 2, 14) and median total CACS was 2 (IQR 0, 7). All-cause mortality occurred in 128 (32%) patients. Patients with CAC were three times more likely to die than patients without CAC (Hazard ratio [HR] 2.96; 95% CI 1.84, 4.77; p<0.001), and patients with severe CAC were at the highest risk (HR 4.62; 95% CI 2.73, 7.83, p<0.001). Gender, modified Miller score and RV/LV ratio were not predictive of mortality. In multivariate analysis both CACS and age were independent predictors of 3-year all-cause mortality. Of the patients with CAC who died, the presence of coronary artery disease was only documented in 34 (32%). CONCLUSION: CACS is an independent predictor of all-cause mortality in patients with PE, and has important implications for subsequent patient management.

Long-term adverse effects associated with isolated below-knee deep-vein thrombosis: a 10-year follow-up study.

AIM: To assess the effect of the presence and locality of symptomatic lower-limb deep vein thrombosis (DVT) on mortality and morbidity following contrast venography (CV), the reference standard for diagnosing below-knee DVT, with a view to determining the prevalence of recurrent episodes of DVT and post-thrombotic syndrome (PTS). MATERIALS AND METHODS: Patients with clinical DVT undergoing investigation using CV were prospectively recorded. By retrospective case note examination and mortality data evaluation, 347 patients with DVT were matched with negative controls for mortality follow-up. Long-term complications were recorded. RESULTS: Fifty-one (14.7%) of the DVT patients were diagnosed with PTS and 43 (12.4%) with possible PTS in the 10 years following presentation. The relative risk for developing definite PTS was 0.544 for below- versus above-knee DVT; 9.9% with below-knee DVT had PTS, and 9% had probable PTS. Recurrent DVT occurred in 23.3% of patients with proximal DVT as opposed to 12.6% of patients with isolated below-knee DVT. CONCLUSIONS: Morbidity is greater in patients with proximal DVT; however, a significant, albeit smaller, proportion of patients with isolated below-knee DVT develop recurrent DVT and PTS. Below-knee DVT carries sufficient morbidity and mortality to warrant vigilance in diagnosis and management of this condition.

Ten years of imaging for pulmonary embolism: too many scans or the tip of an iceberg?

AIM: To examine the number and nature of investigations performed for suspected pulmonary embolism (PE) in a large teaching hospital and the change in incidence and severity of PE over a decade. MATERIALS AND METHODS: In this retrospective study, all patients investigated for suspected PE using computed tomography pulmonary angiography (CTPA) or lung scintigraphy during 10 years to March 2012 were identified and their records reviewed. In the final year, all reportedly positive CTPA cases were reviewed and PE severity calculated, for comparison with similar historical data. RESULTS: From 2002 to 2012, total annual investigations for suspected acute PE increased by 163% (805 to 2121). CTPA increased by 325% (475 to 2019). Detection of PE increased by 121% (193 to 426 per annum), with stable distribution of severity scores. The positive scan rate decreased from 24% to 20%. The mean age of patients being investigated for PE increased from 56 to 63 years. CONCLUSIONS: Increased detection of PE is not due to disproportionate increase in small PEs, but to increased detection of PE of all severities. This finding supports the hypothesis that PE is more common in the general population than previously appreciated, which may represent an iceberg phenomenon of previously undetected disease.

'Presenting CXR phenotype of H1N1' flu compared with contemporaneous non-H1N1, community acquired pneumonia, during pandemic and post-pandemic outbreaks'.

AIMS: To review, phenotype and assess potential prognostic value of initial chest X-ray findings in patients with H1N1 influenza during seasonal outbreaks of 2009 and 2010, in comparison with non-H1N1, community acquired pneumonia (CAP). METHODS: We retrospectively identified 72 patients admitted to hospital with pneumonia during the seasons of 2009 and 2010. H1N1 cases were confirmed by virology PCR. Presenting chest X-rays were jointly read by 2 radiologists, who were 'blinded' to further patient details and divided into 6 zones. Total number of opacified zones, the pattern and distribution of changes and length of hospital stay were recorded. RESULTS: Patients with H1N1 demonstrated more opacified zones (mean of 2.9 compared with 2.0; p=0.006), which were bilateral in two-thirds compared with a quarter of those with non-H1N1 CAP (p=0.001). H1N1 radiographs were more likely to be 'patchy' versus 'confluent' changes of non-H1N1 CAP (p=0.03) and more often demonstrated peripheral distribution (p=0.01). H1N1 patients tended to stay in hospital longer (not significant; p=0.08). A positive correlation existed between number of affected zones and length of inpatient stay, which was statistically significant for the cohorts combined (p=0.02). The findings were the same for the two evaluated seasons. CONCLUSION: H1N1 patients demonstrated more extensive disease, which was more likely bilateral, 'patchy', and peripheral in distribution. With increasing global cases of H1N1, knowledge of the typical findings of the H1N1 presenting chest X-ray may assist with early triage of patients, particularly where rapid viral testing is not available.

Feasibility of radiation dose reduction using AIDR-3D in dynamic pulmonary CT perfusion.

AIM: To assess the feasibility of radiation dose reduction with adaptive iterative dose reduction (AIDR-6 3D) reconstruction in dynamic pulmonary CT perfusion. MATERIALS AND METHODS: CTP examinations of 10 patients acquired at 100 kVp/50 mAs were reconstructed with filtered back projection (FBP) and AIDR-3D. Artificial noise was added to raw data (pre-reconstruction projection data) to simulate lower tube current scanning. Radiodensity (in Hounsfield units), noise, and perfusion values were compared. RESULTS: There was no significant difference in noise between the full and simulated reduced tube current with AIDR-3D reconstruction (p = 1). There was significantly lower noise in lung tissue with AIDR-3D images when compared to reconstructions without AIDR-3D (p = 0.005) and no significant change in the radiodensity (p = 1; mean difference <6 HU). Mean perfusion values increased significantly at lower tube currents (25 and 12.5 mAs), compared to 50 mAs (p = 0.005). This effect was significantly greater in larger patients compared to thin patients. CONCLUSION: AIDR-3D produced significantly lower noise images than FBP-based algorithms and maintained consistent noise levels in lung at 12.5 mAs, indicating this algorithm is suitable for reduced dose lung perfusion imaging. Iterative reconstruction allows significant radiation dose reduction of up to fourfold in smaller patients, and up to twofold in the medium/large size patients. The increase in perfusion values at 25% simulated tube currents is attributed to attenuation bias.

Radiology smartphone applications; current provision and cautions.

OBJECTIVES: Medical smartphone applications are increasingly popular amongst doctors. However, the quality of their content is variable. We assessed contemporary radiology-related smartphone applications, focussing on the level of advertised medical involvement in application development. METHODS: Six major application stores were searched between 18-30 June 2012 using the terms radiology, radiation, x-ray(s), computed tomography/CT, magnetic resonance imaging/MRI, ultrasound, nuclear medicine, fluoroscopy and mammography/mammogram. Application ratings, cost and medical input in development were recorded. RESULTS: 321 applications were identified. One hundred fifty-eight were "teaching" and 96 "reference". Three of the 29 DICOM viewing applications had FDA approval for primary diagnosis, while 62 % stated they should not be used for primary diagnosis; 24 % of applications stated named medical professional involvement, 12 % had unnamed medical involvement and 4 % acknowledged guidelines or papers; 42 % did not disclose authorship. CONCLUSIONS: A large variety of radiology-related smartphone applications are available with many potential benefits. Advertised medical involvement in application design is variable, making assessment of their accuracy difficult prior to purchase. Additional measures are required to ensure the accuracy of such applications. The limitations of image interpretation using smartphones are a major drawback of DICOM viewing applications. Further research into the accuracy of primary diagnosis using such applications is needed. MAIN MESSAGES: • A large variety of radiology smartphone applications are available with many potential benefits • Variable medical involvement in application design limits assessment of accuracy before purchase • Limitations of image interpretation using smartphones are a drawback of DICOM viewing applications • Further work on the accuracy of primary diagnosis using these DICOM viewing applications is needed.

Thrombus load and acute right ventricular failure in pulmonary embolism: correlation and demonstration of a "tipping point" on CT pulmonary angiography.

OBJECTIVES: The aim of this study was to determine the correlation between increasing pulmonary embolism thrombus load and right ventricular (RV) dilatation as demonstrated by CT pulmonary angiography (CTPA) and to assess the thrombus load threshold which indicates impending RV decompensation. METHODS: 2425 consecutive CTPAs were retrospectively analysed. Thrombus load using a modified Miller score (MMS), RV to left ventricular (RV:LV) ratio, presence of septal shift, and pulmonary artery and aorta size were analysed in 504 positive CTPA scans and a representative cohort of 100 negative scans. Results were correlated using non-parametric analysis (two-tailed t-test or χ(2) test) and Pearson's rank correlation. RESULTS: Increasing thrombus load correlated with a higher RV:LV ratio, with a statistically significant difference in RV:LV ratios between the negative and positive pulmonary embolism (PE) cohorts. Larger thrombus loads (MMS ≥12 vs MMS <12) were strongly correlated with RV strain (mean RV:LV ratio, 1.323 vs 0.930; p<0.0001). Smaller thrombus loads had no significant influence on RV strain. Septal shift was also more likely with an MMS of ≥12, as was an increase in pulmonary artery diameter (r=0.221, p<0.001). CONCLUSION: With increasing thrombus load in PE, there is CTPA evidence of RV decompensation with an MMS threshold of 12. This suggests a "tipping point" beyond which RV decompensation is more likely to occur. This is the first study to describe this tipping point between a thrombus load of MMS >12 and an increase in RV:LV ratio. This finding may help to improve risk stratification in patients with acute PE diagnosed by CTPA.

Prognostic significance of the lung scintigraphy scan result and corresponding chest X-ray in patients with suspected pulmonary embolism.

AIM: To determine whether the survival of patients with suspected acute pulmonary embolism (PE) relates to radiological probability of acute PE assessed using lung scintigraphy scans (LSS). METHODS: Lung scintigraphy scan results from a venous thromboembolism database were categorised as high, indeterminate or low probability using the modified PIOPED criteria and corresponding chest X-rays (CXRs) as normal or abnormal. Mortality data on these cases were obtained from the General Register Office for Scotland, and survival was analysed using the Kaplan-Meier method. RESULTS: Of the 1,818 LSS analysed, 941 (51.8%) were normal, 532 (29.3%) indeterminate and 345 (19.0%) high probability. After an adjustment for age and gender, no significant survival difference was found between patients with normal and high probability LSS (p=0.182). However, patients with indeterminate LSS had significantly lower survival than patients in the other groups. This difference persisted after adjustment for CXR result. CONCLUSIONS: Indeterminate LSS results are associated with a poor prognosis. Careful follow-up of patients with inderminate LSS would appear to be justified.

Unsuspected pulmonary embolism identified using multidetector computed tomography in hospital outpatients.

AIM: To evaluate the incidence of unsuspected pulmonary embolism (PE) in an unselected population of outpatients undergoing contrast-enhanced multidetector computed tomography (MDCT) for indications other than the investigation of PE. MATERIALS AND METHODS: Outpatients undergoing CT of the chest over a 6-month period were retrospectively identified and images reviewed. Inpatients and patients undergoing unenhanced CT of the chest were excluded. Data, including referring specialty, patient age and sex, reasons for examination, level of embolism, image quality, and section thickness were recorded. Radiology reports were reviewed with respect to whether or not the embolism was noted at the time of initial reporting. RESULTS: Following exclusions 440 patients were reviewed (195 women and 245 men). PE was identified in 10 of the 440 patients, an incidence of 2.23%. One pulmonary embolus was in the main pulmonary artery, three were in lobar arteries, three in segmental arteries, and three in subsegmental arteries. Patients over the age of 60 years were more likely to have an embolism (9/300, 2.9%) compared with those under 60 years (1/140, 0.7%). Seven of the 10 positive examinations were carried out in patients who were known or later shown to have malignancy. Seven of the 10 emboli were reported at the time of initial reporting. CONCLUSION: The outpatient population has a significant incidence of unsuspected PE. PE should be actively sought when reporting examinations performed for alternative indications, particularly where cancer is a known or suspected diagnosis.

Comparison between conventional interrupted high-resolution CT and volume multidetector CT acquisition in the assessment of bronchiectasis.

The aim of this study was to determine whether there is superior diagnostic accuracy for the detection and exclusion of bronchiectasis using 16-slice CT of the chest (1 mm) compared with conventional high-resolution CT (HRCT) of the chest (10 mm). A prospective study was carried out in patients who were referred for chest CT by a chest physician for the investigation of bronchiectasis over a 1-year period. All scans were performed using a 16-slice CT scanner. In addition to contiguous 1 mm slices, conventional HRCT images (1 mm slice every 10 mm) were prepared. Both datasets were dual read. There were 53 patients with a median age of 62 years (range, 51.5-71.5 years), comprising 14 males and 39 females. 10 of 53 scans had no bronchiectasis in either dataset. 36 patients had bronchiectasis diagnosed on both HRCT and 1 mm scans. Two patients had tubular bronchiectasis on the HRCT scans, which was not confirmed on the 1 mm scans. Five patients had confirmed tubular bronchiectasis on the 1 mm scans, which was not identified on HRCT scans. 40 extra lobes demonstrated bronchiectasis on the 1 mm vs the HRCT scans; of these, half were labelled as definite bronchiectasis on the 1 mm scan. There was a 32% increased confidence with the 1 mm scans compared with conventional HRCT of the chest in the diagnosis of bronchiectasis (p < 0.001). In conclusion, there is improved diagnostic accuracy and confidence for diagnosis and exclusion of bronchiectasis using 16-slice chest CT (1 mm cuts) compared with conventional HRCT of the chest.

Idiopathic venous thromboembolic disease is associated with a poorer prognosis from subsequent malignancy.

METHODS: We carried out a retrospective study of prognosis in Scottish patients diagnosed with cancer within 5 years after a venous thromboembolism (VTE). RESULTS AND CONCLUSIONS: Prognosis was significantly poorer if a VTE occurred up to 2 years before cancer diagnosis, most notably if the cancer was diagnosed in the 6 months after a VTE.

Clinically suspected acute pulmonary embolism: a comparison of presentation, radiological features and outcome in patients with and without PE.

BACKGROUND: Relatively little is known about prognosis in patients for whom suspected pulmonary embolism (PE) is refuted by imaging. AIM: This prospective study of suspected PE therefore compared clinico-radiological features and outcome in patients with and without PE. DESIGN AND METHODS: Computed tomographic pulmonary angiography (CTPA) confirmed or refuted PE in consecutive patients. Clinical, laboratory and radiological features were recorded at baseline, and mortality at 1 year determined. Univariate and multivariate analyses identified variables associated with PE. RESULTS: PE was diagnosed in 45 patients and refuted in 141. The PE and 'non-PE' groups were similar with regard to extravascular radiology (though consolidation was significantly more common in the PE group [present in 24 (53%) of the PE group and 42 (30%) of the non-PE group, P < 0.01)], comorbidities (no significant differences), and baseline characteristics (only serum D-dimer concentrations were independently associated with PE by multivariate analysis, P = 0.001). Right ventricular dimensions were significantly higher in the PE group, [right ventricular to left ventricular ratio was 0.98 (range 0.64-2.48) in the PE group and 0.92 (range 0.66-1.95) in the non-PE group, P < 0.05]. In the PE group, right ventricular dimensions rose sharply when 10 or more segmental pulmonary arteries were occluded. One year all-cause mortality was 6.7% in the PE group and 13.5% in the non-PE group (no significant difference, P = 0.218). CONCLUSION: Among a cohort of patients presenting with clinically suspected PE, clinical characteristics, co-morbidities and radiological features were similar when comparing groups with CTPA-proven or CTPA-refuted PE. However RV dimensions, radiological consolidation on imaging and D-dimer levels were significantly higher in the PE group. Patients with suspected PE have a poor prognosis irrespective of whether PE is confirmed. This appears accentuated in patients without PE, a finding possibly under-recognized in clinical practice.

Non-small-cell lung cancer dimensions: CT-pathological correlation and interobserver variation.

The aim of the study was to identify the most accurate CT window level setting for the measurement of non-small-cell lung cancer to optimise CT planning for radiotherapy treatment. 27 patients who underwent resection for non-small-cell lung cancer in a single institution were studied. The maximal superior-inferior, anteroposterior and mediolateral dimensions of the resected tumours were measured by a consultant pathologist. Two radiologists made corresponding measurements using pre-operative CT scans independently of each other and of the pathologist's findings. The measurements were obtained using four different CT window settings. The mean pathological size of the superior-inferior tumours, the anteroposterior tumours and the mediolateral tumours was 32 mm, 28 mm and 25 mm, respectively. A total of 648 CT measurements were taken, of which 321 were within +/-5 mm of the pathological size (49.5%). There was significant interobserver variability between the two radiologists. There was poor correlation between the pathological and radiological measurements of tumour size. Significant interobserver variability was noted between the two radiologists and no window setting could be identified as being superior in accurately assessing the tumour size.

Computed tomographic emphysema distribution: relationship to clinical features in a cohort of smokers.

Computed tomography (CT) scanning allows precise assessment of both the extent and distribution of emphysema. There has been little work on the relationship between the distribution of emphysema and clinical features of the disease. The current study investigated the association between clinical features and distribution of emphysema. A total of 129 patients with smoking-related chronic obstructive pulmonary disease underwent CT assessment of the extent and distribution of their emphysema (core/rind and upper/lower zone predominance). Emphysema was found predominantly in the upper/core zone and this distribution was related to the extent of disease. Core predominance was associated with lower forced expiratory volume in one second (FEV(1)), FEV(1)/forced vital capacity ratio and body mass index (BMI); and with higher BODE (BMI, airflow obstruction, dyspnoea and exercise capacity) index and Medical Research Council dyspnoea score. Upper-zone predominance was associated with female sex and an increased total St George's Respiratory Questionnaire score. Using multiple linear regression age, sex and whole lung emphysema severity were independently associated with core/rind distribution, while sex and whole lung emphysema severity were independently related to upper/lower distribution. Distribution of emphysema related best to clinical features when divided into core/rind predominance. However, the effects were not independent of the extent of emphysema. Increased age and female sex were related to disease distribution independent of emphysema severity. These findings may be related to differences in development of emphysema.

A profile of lower-limb deep-vein thrombosis: the hidden menace of below-knee DVT.

AIMS: To describe the anatomical site and laterality of deep-vein thrombosis (DVT) in symptomatic patients using contrast venography (CV), and to assess age, sex distribution, and accuracy of pre-test clinical suspicion of DVT. METHODS: One thousand, five hundred and seventy-two patients undergoing CV because of a clinical suspicion of DVT at a large teaching hospital from October 1995 to March 2003 were prospectively studied. RESULTS: Thrombi were demonstrated in 511 (32.5%) of all CV studies. Isolated, below-knee thrombi were identified in 29.4% of positive studies. There was a left-sided predominance of DVT (ratio 1.24:1) that was most evident in the elderly and in more proximal veins. CONCLUSION: Almost a third of positive cases were shown to be isolated, below-knee thrombi. These are thrombi that are more difficult to detect by non-invasive means. A left-sided predominance of DVT is evident.

Endobronchial and endoscopic ultrasound-guided real-time fine-needle aspiration for mediastinal staging.

Accurate staging of the mediastinum in lung cancer is essential for optimising treatment strategies. Conventional transbronchial needle aspiration (TBNA) is a blind procedure, reliant upon prior computed tomography (CT) or ultrasound imaging, but has low sensitivity. The current study reports the initial experience of using a prototype endobronchial ultrasound (EBUS) probe that allows TBNA under real-time imaging. In 20 patients selected by CT scanning, a linear-array ultrasound bronchoscope was used to visualise paratracheal and hilar lymph nodes, and TBNA was performed under direct ultrasonic control. In seven cases, sequential endoscopic ultrasound (EUS) was used to assess postero-inferior mediastinal lymph nodes. All procedures were performed under conscious sedation. EBUS-TBNA was undertaken in 18 out of 20 cases and EUS-guided fine-needle aspiration in six out of seven cases. Cytology showed node (N)2/N3 disease in 11 out of 18 EBUS-TBNA cases and provided a primary diagnosis for eight patients. EBUS-TBNA cytology was negative in six cases, which was confirmed by mediastinoscopy or clinical follow-up in four. EUS provided additional information in all cases. There were no procedural complications. Sensitivity, specificity and accuracy for EBUS-TBNA were 85%, 100% and 89%, respectively. In conclusion, endobronchial ultrasound with real-time transbronchial needle aspiration offers improved sensitivity and accuracy for staging of the middle mediastinum, and, combined with endoscopic ultrasound, should allow investigation of the majority of the mediastinum.

Effect of the introduction of helical CT on radiation dose in the investigation of pulmonary embolism.

The aim of this study was to assess the change in patient radiation dose in the radiological investigation of pulmonary embolism since the introduction of helical CT pulmonary angiography (CTPA) in a large teaching hospital. All radiological investigations performed as an integral part of the imaging protocol in the investigation of clinically suspected pulmonary embolism (PE) were retrospectively reviewed. The protocol for the investigation of PE changed in our institution after the introduction of CTPA. Protocols 1 and 2 were the protocols in place before and after the introduction of CTPA, respectively. An in-depth evaluation was made of the imaging records and radiation dose for 30 consecutive patients investigated for clinically suspected PE in 1995 (protocol 1) and 2002 (protocol 2). Radiation doses were then extrapolated for the total number of patients investigated in each year. The number of radiological investigations performed per patient decreased from a mean of 1.17 in protocol 1 to 1.06 in protocol 2. There was a 44% increase in the total number of patients investigated. The effective dose per patient increased from 1.30 mSv to 1.35 mSv with the introduction of CTPA into the imaging protocol, an increase of only 4%. First line investigations showed a significant decrease in indeterminate examinations from 25.7% to 8.5%. Two different imaging protocols are reviewed with respect to type and number of procedures required for the investigation of PE and the resulting patient effective dose incurred. Results demonstrate an increase in the number of patients being investigated for suspected PE and a small increase in effective dose per patient since the introduction of helical CTPA. Although CTPA in itself incurs a higher effective dose, this is offset by the significant decrease in the number of non-diagnostic and total number of investigations per patient. In addition the ventilation component of lung scintigraphy was not required in protocol 2, thus reducing the dose further. We believe this small increase in effective dose is justified by the decrease in non-diagnostic studies and the reduction in total number of investigations per patient. We hope this paper will serve as a stimulus for the radiology community to examine current protocols in all areas of diagnostic imaging. We stress the importance of assessing new and established imaging investigative protocols to maximize the benefit and reduce any risk to patients.