The implementation of artificial intelligence in radiology: a narrative review of patient perspectives.

Aim: To synthesise research on the view of the public and patients of the use of artificial intelligence (AI) in radiology investigations. Methods: A literature review of narrative synthesis of qualitative and quantitative studies that reported views of the public and patients toward the use of AI in radiology. Results: Only seven studies related to patient and public views were retrieved, suggesting that this is an underexplored area of research. Two broad themes, of confidence in the capabilities of AI, and the accountability and transparency of AI, were identified. Conclusions: Both optimism and concerns were expressed by participants. Transparency in the implementation of AI, scientific validation, clear regulation and accountability were expected. Combined human and AI interpretation of imaging was strongly favoured over AI acting autonomously. The review highlights the limited engagement of the public in the adoption of AI in a radiology setting. Successful implementation of AI in this field will require demonstrating not only adequate accuracy of the technology, but also its acceptance by patients.

The radiology quality assurance process in the Yorkshire Lung Screening Trial, and findings from the baseline round of low dose CT screening for lung cancer.

OBJECTIVE: As lung cancer screening is rolled-out, there is a need to develop an effective quality assurance (QA) framework around radiology reporting to ensure optimal implementation. Here, we report a structured QA process for low-dose CT (LDCT) scans performed in the Yorkshire Lung Screening Trial. METHODS: Negative LDCT scans were single read after using computer-aided detection software. The radiology QA process included reviewing 5% of negative scans selected at random, and all cases with a subsequent diagnosis of extrapulmonary cancer or interval lung cancer not detected on the baseline scan. Radiologists were not informed of the reason for review and original radiology reports were scored as either "satisfactory", "satisfactory with learning points", or "unsatisfactory". RESULTS: From 6650 participants undergoing LDCT screening, 208 negative scans were reviewed alongside 11 cases with subsequent extrapulmonary cancer and 10 cases with interval lung cancer. Overall, only three reports were ultimately judged "unsatisfactory", 1% of randomly selected negative scans (n = 2/208) and one interval lung cancer scan (n = 1/10). Four out of a total of five cases judged "satisfactory with learning points" were related to oesophageal abnormalities where the participant was subsequently diagnosed with oesophageal cancer. CONCLUSION: The described process attempts to minimise bias in retrospective review of screening scans, and may represent a framework for future QA of national screening programmes. ADVANCES IN KNOWLEDGE: This study describes a structured QA process for a lung cancer screening programme, involving blinded second-read of LDCT screening scans to ensure fair, constructive audit of clinical performance.

Diagnoses and treatments for participants with interstitial lung abnormalities detected in the Yorkshire Lung Screening Trial.

INTRODUCTION: Interstitial lung abnormalities (ILA) are relatively common incidental findings in participants undergoing low-dose CT screening for lung cancer. Some ILA are transient and inconsequential, but others represent interstitial lung disease (ILD). Lung cancer screening therefore offers the opportunity of earlier diagnosis and treatment of ILD for some screening participants. METHODS: The prevalence of ILA in participants in the baseline screening round of the Yorkshire Lung Screening Trial is reported, along with the proportion referred to a regional ILD service, eventual diagnoses, outcomes and treatments. RESULTS: Of 6650 participants undergoing screening, ILA were reported in 169 (2.5%) participants. Following review in a screening review meeting, 56 participants were referred to the ILD service for further evaluation (0.8% of all screening participants). 2 participants declined referral, 1 is currently awaiting review and the remaining 53 were confirmed as having ILD. Eventual diagnoses were idiopathic pulmonary fibrosis (n=14), respiratory bronchiolitis ILD (n=4), chronic hypersensitivity pneumonitis (n=2), connective tissue disease/rheumatoid arthritis-related ILD (n=4), asbestosis (n=1), idiopathic non-specific interstitial pneumonia (n=1), sarcoidosis (n=1) and pleuroparenchymal fibroelastosis (n=1). Twenty five patients had unclassifiable idiopathic interstitial pneumonia. Overall, 10 people received pharmacotherapy (7 antifibrotics and 3 prednisolone) representing 18% of those referred to the ILD service and 0.15% of those undergoing screening. 32 people remain under surveillance in the ILD service, some of whom may require treatment in future. DISCUSSION: Lung cancer screening detects clinically significant cases of ILD allowing early commencement of disease-modifying treatment in a proportion of participants. This is the largest screening cohort to report eventual diagnoses and treatments and provides an estimate of the level of clinical activity to be expected by ILD services as lung cancer screening is implemented. Further research is needed to clarify the optimal management of screen-detected ILD. TRIAL REGISTRATION NUMBER: ISRCTN42704678.

A prospective cohort evaluation of the sensitivity and specificity of the chest X-ray for the detection of lung cancer in symptomatic adults.

BACKGROUND: The accuracy of the chest x-ray (CXR) in the identification of lung cancer amongst symptomatic individuals is uncertain. PURPOSE: To determine the diagnostic accuracy of the CXR for the detection of non-small cell carcinomas (NSCLC) and all primary intrathoracic malignancies. METHODS: A prospective cohort study of consecutive CXR reports obtained within a primary care open access initiative. Eligibility criteria were symptoms specified by National Institute for Clinical Excellence as indicative of possible lung cancer and age over 50-yrs. A positive test was a CXR which led directly or indirectly to investigation with CT. The reference standards were malignancies observed within a one- or two-year post-test period. RESULTS: 8,948 CXR outcomes were evaluated. 496 positive studies led to a diagnosis of 101 patients with primary intrathoracic malignancy including 80 with NSCLC. Within two-years, a cumulative total of 168 patients with primary intrathoracic malignancies including 133 NSCLC were observed. The sensitivity and specificity for NSCLC were 76% (95 %CI 68-84) and 95% (95 %CI 95-96) within 1-year and 60% (95 %CI 52-69) and 95% (95 %CI 95-96) within 2-years. The 2-yr positive and negative likelihood ratios were 12.8 and 0.4. The results did not differ for NSCLC compared to all primary malignancies. Within this symptomatic population a negative test reduced the 2-year risk of lung cancer to 0.8%. CONCLUSIONS: A positive test strongly increases the probability of malignancy whereas a negative test does not conclusively exclude the disease. The findings allow the risk of malignancy following a negative test to be estimated.

Chest X-ray sensitivity and lung cancer outcomes: a retrospective observational study.

BACKGROUND: Chest X-ray (CXR) is the first-line investigation for lung cancer in many healthcare systems. An understanding of the consequences of false-negative CXRs on time to diagnosis, stage, and survival is limited. AIM: To determine the sensitivity of CXR for lung cancer and to compare stage at diagnosis, time to diagnosis, and survival between those with CXR that detected, or did not detect, lung cancer. DESIGN AND SETTING: Retrospective observational study using routinely collected healthcare data. METHOD: All patients diagnosed with lung cancer in Leeds Teaching Hospitals NHS Trust during 2008-2015 who had a GP-requested CXR in the year before diagnosis were categorised based on the result of the earliest CXR performed in that period. The sensitivity of CXR was calculated and analyses were performed with respect to time to diagnosis, survival, and stage at diagnosis. RESULTS: CXR was negative for 17.7% of patients (n = 376/2129). Median time from initial CXR to diagnosis was 43 days for those with a positive CXR and 204 days for those with a negative CXR. Of those with a positive CXR, 29.8% (95% confidence interval [CI] = 27.9% to 31.8%) were diagnosed at stage I or II, compared with 33.5% (95% CI = 28.8% to 38.6%) with a negative CXR. CONCLUSION: GPs should consider lung cancer in patients with persistent symptoms even when CXR is negative. Despite longer duration to diagnosis for those with false-negative CXRs, there was no evidence of an adverse impact on stage at diagnosis or survival; however, this comparison is likely to be affected by confounding variables.

Estimating lung cancer risk from chest X-ray and symptoms: a prospective cohort study.

BACKGROUND: Chest X-ray (CXR) is the first-line investigation for lung cancer in many countries but previous research has suggested that the disease is not detected by CXR in approximately 20% of patients. The risk of lung cancer, with particular symptoms, following a negative CXR is not known. AIM: To establish the sensitivity and specificity of CXR requested by patients who are symptomatic; determine the positive predictive values (PPVs) of each presenting symptom of lung cancer following a negative CXR; and determine whether symptoms associated with lung cancer are different in those who had a positive CXR result compared with those who had a negative CXR result. DESIGN AND SETTING: A prospective cohort study was conducted in Leeds, UK, based on routinely collected data from a service that allowed patients with symptoms of lung cancer to request CXR. METHOD: Symptom data were combined with a diagnostic category (positive or negative) for each CXR, and the sensitivity and specificity of CXR for lung cancer were calculated. The PPV of lung cancer associated with each symptom or combination of symptoms was estimated for those patients with a negative CXR. RESULTS: In total, 114 (1.3%) of 8996 patients who requested a CXR were diagnosed with lung cancer within 1 year. Sensitivity was 75.4% and specificity was 90.2%. The PPV of all symptoms for a diagnosis of lung cancer within 1 year of CXR was <1% for all individual symptoms except for haemoptysis, which had a PPV of 2.9%. PPVs for a diagnosis of lung cancer within 2 years of CXR was <1.5% for all single symptoms except for haemoptysis, which had a PPV of 3.9%. CONCLUSION: CXR has limited sensitivity; however, in a population with a low prevalence of lung cancer, its high specificity and negative predictive value means that lung cancer is very unlikely to be present following a negative result. Findings also support guidance that unexplained haemoptysis warrants urgent referral, regardless of CXR result.

Correction to: Long-Term Outcomes in Percutaneous Radiofrequency Ablation for Histologically Proven Colorectal Lung Metastasis.

In the introduction section on line 7, the following sentence "A large epidemiological study of colorectal cancer patients with lung metastases found 3 and 5-year survival rates of 1.3% and 6.9%" should actually be "A large epidemiological study of colorectal cancer patients with lung metastases found 3 and 5-year survival rates of 11.3% and 6.9%".

Urinothorax following percutaneous image-guided renal cryoablation.

A 69-year-old lady with 2 renal cell carcinomas, one sited at the upper pole of her solitary right kidney, underwent percutaneous image-guided cryoablation and developed urinothorax as a complication. This was diagnosed from pleural fluid analysis and radiology imaging with computed tomography (CT). Management included image-guided chest drain and retrograde ureteric stent insertion to divert the urine from entering the pleural cavity. CT images demonstrated a fistula between the site of renal puncture and the pleural cavity, indicating that the cryoprobes traversed the diaphragm during the procedure. This case highlights urinothorax as an unusual complication of cryoablation of renal cell carcinoma. Prompt diagnosis by interventional radiologists is crucial to avert from this potentially life-threatening complication.

Long-Term Outcomes in Percutaneous Radiofrequency Ablation for Histologically Proven Colorectal Lung Metastasis.

INTRODUCTION: To evaluate the long-term outcome of image-guided radiofrequency ablation (RFA) when treating histologically confirmed colorectal lung metastasis in terms of overall survival (OS), progression-free survival (PFS) and local tumour control (LTC). MATERIALS AND METHODS: Retrospective single-centre study. Consecutive RFA treatments of histologically proven lung colorectal metastases between 01/01/2008 and 31/12/14. The primary outcome was patient survival (OS and PFS). Secondary outcomes were local tumour progression (LTP) and complications. Prognostic factors associated with OS/ PFS were determined by univariate and multivariate analyses. RESULTS: Sixty patients (39 males: 21 females; median age 69 years) and 125 colorectal lung metastases were treated. Eighty percent (n = 48) also underwent lung surgery for lung metastases. Mean metastasis size (cm) was 1.4 ± 0.6 (range 0.3-4.0). Median number of RFA sessions was 1 (1-4). During follow-up (median 45.5 months), 45 patients died (75%). The estimated OS and PFS survival rates at 1, 3, 5, 7, 9 years were 96.7%, 74.7%, 44.1%, 27.5%, 16.3% (median OS, 52 months) and 66.7%, 31.2%, 25.9%, 21.2% and 5.9% (median PFS, 19 months). The LTC rate was 90% with 6 patients developing LTP with 1-, 2-, 3- and 4-year LTP rates of 3.3%, 8.3%, 10.0% and 10.0%. Progression-free interval < 1 year (P = 0.002, HR = 0.375) and total number of pulmonary metastases (≥ 3) treated (P = 0.037, HR = 0.480) were independent negative prognostic factors. Thirty-day mortality rate was 0% with no intra-procedural deaths. CONCLUSION: The long-term OS and PFS following RFA for the treatment of histologically confirmed colorectal lung metastases demonstrate comparable oncological durability to surgery.

Lung cancer stage-shift following a symptom awareness campaign.

BACKGROUND: Lung cancer outcomes in the UK are worse than in many other developed nations. Symptom awareness campaigns aim to diagnose patients at an earlier stage to improve cancer outcomes. METHODS: An early diagnosis campaign for lung cancer commenced in Leeds, UK in 2011 comprising public and primary-care facing components. Rates of community referral for chest X-ray and lung cancer stage (TNM seventh edition) at presentation were collected from 2008 to 2015. Linear trends were assessed by χ2 test for trend in proportions. Headline figures are presented for the 3 years pre-campaign (2008-2010) and the three most recent years for which data are available during the campaign (2013-2015). FINDINGS: Community-ordered chest X-ray rates per year increased from 18 909 in 2008-2010 to 34 194 in 2013-2015 (80.8% increase). A significant stage shift towards earlier stage lung cancer was seen (χ2(1)=32.2, p<0.0001). There was an 8.8 percentage point increase in the proportion of patients diagnosed with stage I/II lung cancer (26.5% pre-campaign vs 35.3% during campaign) and a 9.3% reduction in the absolute number of patients diagnosed with stage III/IV disease (1254 pre-campaign vs 1137 during campaign). INTERPRETATION: This is the largest described lung cancer stage-shift in association with a symptom awareness campaign. A causal link between the campaign and stage-shift cannot be proven but appears plausible. Limitations of the analysis include a lack of contemporary control population.

Hepatocellular carcinoma in cirrhotic livers: double-contrast thin-section MR imaging with pathologic correlation of explanted tissue.

OBJECTIVE: The aim of our study was to determine the sensitivity of double-contrast MR imaging in the detection of hepatocellular carcinomas in patients with a cirrhotic liver. SUBJECTS AND METHODS: Thirty-one patients underwent double-contrast MR imaging and subsequent liver transplantation. Breath-hold T1- and T2-weighted MR images were obtained before and after administration of superparamagnetic iron oxide, and three-dimensional T1-weighted gradient-recalled echo MR images were obtained 10, 40, and 120 sec after a bolus injection of gadolinium. Hypervascular lesions that failed to take up superparamagnetic iron oxide were regarded as showing typical characteristics of hepatocellular carcinoma; lesions that had only one of these two characteristics (either hypervascularity or failure to take up superparamagnetic iron oxide) were regarded as highly suspicious for hepatocellular carcinoma. Radiology reports were correlated with pathology reports for the explanted livers. RESULTS: Thirty-two hepatocellular carcinomas were found in 14 of the 31 patients. Combining the number of MR imaging reports citing lesions that were "typical of hepatocellular carcinoma" with the number of those citing lesions that were "highly suspicious," we found that for 25 of 32 lesions, an accurate MR imaging diagnosis of hepatocellular carcinoma was made (overall sensitivity, 78%). These lesions included 10 of the 11 lesions that were larger than 20 mm (sensitivity, 91%), 12 of the 13 lesions that were 11-20 mm (sensitivity, 92%), and three of the eight lesions that were 10 mm or less (sensitivity, 38%). Nineteen (76%) of 25 lesions had characteristics considered typical of hepatocellular carcinoma; the remaining six lesions either failed to take up superparamagnetic iron oxide and were hypovascular or were hypervascular but showed some uptake of superparamagnetic iron oxide. CONCLUSION: In patients with a cirrhotic liver, double-contrast MR imaging is highly sensitive in the diagnosis of hepatocellular carcinomas of 10 mm or larger, but success in the identification of tumors smaller than 10 mm is still limited.