BJR open最新文献

The use of artificial intelligence (AI) holds great promise for radiation oncology, with many applications being reported in the literature, including some of which are already in clinical use. These are mainly in areas where AI provides benefits in efficiency (such as automatic segmentation and treatment planning). Prediction models that directly impact patient decision-making are far less mature in terms of their application in clinical practice. Part of the limited clinical uptake of these models may be explained by the need for broader knowledge, among practising clinicians within the medical community, about the processes of AI development. This lack of understanding could lead to low commitment to AI research, widespread scepticism, and low levels of trust. This attitude towards AI may be further negatively impacted by the perception that deep learning is a "black box" with inherently low transparency. Thus, there is an unmet need to train current and future clinicians in the development and application of AI in medicine. Improving clinicians' AI-related knowledge and skills is necessary to enhance multidisciplinary collaboration between data scientists and physicians, that is, involving a clinician in the loop during AI development. Increased knowledge may also positively affect the acceptance and trust of AI. This paper describes the necessary steps involved in AI research and development, and thus identifies the possibilities, limitations, challenges, and opportunities, as seen from the perspective of a practising radiation oncologist. It offers the clinician with limited knowledge and experience in AI valuable tools to evaluate research papers related to an AI model application.

Objectives: To apply CT-based foundational artificial intelligence (AI) and radiomics models for predicting overall survival (OS) for patients with locally advanced non-small cell lung cancer (NSCLC).

Methods: Data for 449 patients retrospectively treated on the NRG Oncology/Radiation Therapy Oncology Group (RTOG) 0617 clinical trial were analyzed. Foundational AI, radiomics, and clinical features were evaluated using univariate cox regression and correlational analyses to determine independent predictors of survival. Several models were fit using these predictors and model performance was evaluated using nested cross-validation and unseen independent test datasets via area under receiver-operator-characteristic curves, AUCs.

Results: For all patients, the combined foundational AI and clinical models achieved AUCs of 0.67 for the Random Forest (RF) model. The combined radiomics and clinical models achieved RF AUCs of 0.66. In the low-dose arm, foundational AI alone achieved AUC of 0.67, while AUC for the ensemble radiomics and clinical models was 0.65 for the support vector machine (SVM). In the high-dose arm, AUC values were 0.67 for combined radiomics and clinical models and 0.66 for the foundational AI model.

Conclusions: This study demonstrated encouraging results for application of foundational AI and radiomics models for prediction of outcomes. More research is warranted to understand the value of ensemble models toward improving performance via complementary information.

Advances in knowledge: Using foundational AI and radiomics-based models we were able to identify significant signatures of outcomes for NSCLC patients retrospectively treated on a national cooperative group clinical trial. Associated models will be important for application toward prospective patients.

Objectives: To gather and synthesize evidence regarding diagnostic accuracy of perfusion imaging by CT (CTP) or MR (MRP) for brain death (BD) diagnosis.

Methods: A systematic review and meta-analysis was prospectively registered with PROSPERO (CRD42022336353) and conducted in accordance with the PRISMA guidelines and independently by 3 reviewers. PubMed/MEDLINE, EMBASE and Cochrane Database were searched for relevant studies. Quality Assessment of Diagnostic Accuracy Studies-2 was used to assess studies' quality. Meta-analysis was performed using univariate random-effects models.

Results: Ten studies (328 patients) were included. Perfusion imaging (most commonly CTP, n = 8 studies) demonstrated a high sensitivity of 96.1% (95% CI, 89.5-98.6) for BD, consistent in subgroup analysis at 95.5% (95% CI, 86.5-98.6). Unfortunately, it was not feasible to calculate other metrics. Additionally, evidence of publication bias was identified in our findings.

Conclusions: The sensitivity of CTP or MRP for BD diagnosis is very high, comparable to CTA and TCD. However, considering most studies were retrospective, and lacked control groups and unambiguous criteria for perfusion imaging in BD assessment, results should be interpreted with caution. Future studies, ideally prospective, multi-centre, and with control groups are of utmost importance for validation of these methods, particularly with standardized technical parameters.

Advances in knowledge: Cerebral perfusion imaging using CT or MRI demonstrates high sensitivity in diagnosing BD, on par with CTA and TCD. Recommended by the World Brain Death group, this method holds promise for further investigation in this area.

Prospero registration number: CRD42022336353.

Post-mortem CT (PMCT) is increasingly used in adult post-mortem investigations as a non-invasive alternative to traditional autopsies. Using PMCT supports death investigations in the face of severe pathologist workforce shortages and the less invasive nature maintains respect for cultural sensitivities. This article reviews the diverse service structures of PMCT, highlighting the importance of customizing these structures to meet the specific needs of various coronial jurisdictions. These jurisdictions often face challenges such as limited access to imaging facilities and logistical issues with geographically dispersed mortuaries. We outline options for leading and operating PMCT services, including models led by pathologists, radiologist, or a hybrid of the two; use of static, relocatable, or mobile CT scanning units; as well as making the most of existing resources such as NHS or private scanning facility scanners already in place. We also explore different PMCT reporting structures through in-house NHS radiologists, combined in-house and teleradiology, or fully outsourced teleradiology services. Each of these offerings provides different levels of efficiency, cost-effectiveness, data security and challenges to set-up. Where applicable, we present and describe real-world examples as case studies for readers interested in replicating existing models.

Objectives: Coronary CT angiography (CCTA) is becoming increasingly important in the workup of coronary artery disease. Imaging of stents and in-stent stenoses remains a challenge. This work investigates the assessability of in-stent stenoses in photon counting CT (PCCT) using ultra-high-resolution (UHR) imaging and optimized reconstruction kernels.

Methods: In an established phantom, 6 stents with inserted hypodense stenoses were scanned in both standard resolution (SRM) and UHR in a clinical PCCT scanner (NAEOTOM Alpha, Siemens Healthineers, Germany). Reconstructions were made both with the clinically established and optimized kernels. The visible stent lumen and the extent of stenosis were quantitatively measured and compared with the angiographic reference standard. Also, region-of-interest (ROI)-based measurements and a qualitative assessment of image quality were performed.

Results: The visible stent lumen and the extent of stenosis were measured more precisely in UHR compared to SRM (0.11 ± 0.19 vs 0.41 ± 0.22 mm, P < .001). The optimized kernel further improved the accuracy of the measurements and image quality in UHR (0.35 ± 0.23 vs 0.47 ± 0.19 mm, P < .001). Compared to angiography, stenoses were overestimated in PCCT, on average with an absolute difference of 18.20% ± 4.11%.

Conclusions: Photon counting CCTA allows improved imaging of in-stent stenoses in a phantom using UHR imaging and optimized kernels. These results support the use of UHR and optimized kernels in clinical practice and further studies.

Advances in knowledge: UHR imaging and optimized reconstruction kernels should be used in CCTA in the presence of cardiac stents.

Objectives: The aim of this study was to systematically review the literature to assess the application of AI-based interventions in lung cancer screening, and its future implications.

Methods: Relevant published literature was screened using PRISMA guidelines across three databases: PubMed, Scopus, and Web of Science. Search terms for article selection included "artificial intelligence," "radiology," "lung cancer," "screening," and "diagnostic." Included studies evaluated the use of AI in lung cancer screening and diagnosis.

Results: Twelve studies met the inclusion criteria. All studies concerned the role of AI in lung cancer screening and diagnosis. The AIs demonstrated promising ability across four domains: (1) detection, (2) characterization and differentiation, (3) augmentation of the work of human radiologists, (4) AI implementation of the LUNG-RADS framework and its ability to augment this framework. All studies reported positive results, demonstrating in some cases AI's ability to perform these tasks to a level close to that of human radiologists.

Conclusions: The AI systems included in this review were found to be effective screening tools for lung cancer. These findings hold important implications for the future use of AI in lung cancer screening programmes as they may see use as an adjunctive tool for lung cancer screening that would aid in making early and accurate diagnosis.

Advances in knowledge: AI-based systems appear to be powerful tools that can assist radiologists with lung cancer screening and diagnosis.

Background: Wilms tumour, a common paediatric cancer, is difficult to treat in low- and middle-income countries due to limited access to imaging. Artificial intelligence (AI) has been introduced for staging, detecting, and classifying tumours, aiding physicians in decision-making. However, challenges include algorithm accuracy, translation into conventional diagnosis, reproducibility, and reliability. As AI technology advances, radiomics, an AI tool, emerges to extract tumour morphology and stage information.

Objectives: This review explores the application of radiomics in Wilms tumour management, including its potential in diagnosis, prognosis, and treatment. Additionally, it discusses the future prospects of AI in this field and potential directions for automation-aided Wilms tumour treatment.

Methods: The review analyses various research studies and articles on the use of radiomics in Wilms tumour management. This includes studies on automated deep learning-based classification, interobserver variability in histopathological analysis, and the application of AI in staging, detecting, and classifying Wilms tumours.

Results: The review finds that radiomics offers several promising applications in Wilms tumour management, including improved diagnosis: it helps in classifying Wilms tumours from other paediatric kidney tumours, prognosis prediction: radiomic features can be used to predict both staging and response to preoperative chemotherapy, Treatment response assessment: Radiomics can be used to monitor the response of Wilms and to predict the feasibility of nephron-sparing surgery.

Conclusions: This review concludes that radiomics has the potential to significantly improve the diagnosis, prognosis, and treatment of Wilms tumours. Despite some challenges, such as the need for further research and validation, AI integration in Wilms tumour management offers promising opportunities for improved patient care.

Advances in knowledge: This review provides a comprehensive overview of the potential applications of radiomics in Wilms tumour management and highlights the significant role AI can play in improving patient outcomes. It contributes to the growing body of knowledge on AI-assisted diagnosis and treatment of paediatric cancers.

Objectives: Artificial intelligence (AI) algorithms have the potential to assist radiologists in the reporting of head computed tomography (CT) scans. We investigated the performance of an AI-based software device used in a large teleradiology practice for intracranial haemorrhage (ICH) detection.

Methods: A randomly selected subset of all non-contrast CT head (NCCTH) scans from patients aged ≥18 years referred for urgent teleradiology reporting from 44 different hospitals within the United Kingdom over a 4-month period was considered for this evaluation. Thirty auditing radiologists evaluated the NCCTH scans and the AI output retrospectively. Agreement between AI and auditing radiologists is reported along with failure analysis.

Results: A total of 1315 NCCTH scans from as many distinct patients (median age, 73 years [IQR 53-84]; 696 [52.9%] females) were evaluated. One hundred twelve (8.5%) scans had ICH. Overall agreement, positive percent agreement, negative percent agreement, and Gwet's AC1 of AI with radiologists were found to be 93.5% (95% CI, 92.1-94.8), 85.7% (77.8-91.6), 94.3% (92.8-95.5) and 0.92 (0.90-0.94), respectively, in detecting ICH. 9 out of 16 false negative outcomes were due to missed subarachnoid haemorrhages and these were predominantly subtle haemorrhages. The most common reason for false positive results was due to motion artefacts.

Conclusions: AI demonstrated very good agreement with the radiologists in the detection of ICH.

Advances in knowledge: Real-world evaluation of an AI-based CT head interpretation device is reported. Knowledge of scenarios where false negative and false positive results are possible will help reporting radiologists.

Objective: Diagnostic reference levels (DRLs) for CT in PET-CT are limited, and published DRLs from other countries may not be directly applicable to the State of Kuwait (KW). The authors aimed to carry out the final phase of a 3-year study on DRLs in KW, supporting optimization and dose reduction as imaging technology advances.

Methods: In this cohort study, 400 adult oncology patients from 8 PET-CT centres were included, following the same procedures as in the first (2018) and second (2020) years, in accordance with the MOH-KW Ethical Committee's recommendations. The CT dose index (CTDIvol), dose-length product (DLP), and scan length were recorded, and the median, mean, standard deviation, as well as the 75th and 25th percentiles, along with the whole-body (WB) effective dose (ED), were calculated. Comparative studies were conducted to track implementation and identify any shortfalls.

Results: In this study, half-body (HB) and WB scans accounted for 66% and 34% of the total 400 cases, respectively. The proposed local DRL practice among the 8 centres in the 2022 study exhibited a maximum variation of 25%, showing a 30% improvement over 2020. The achievable local DRL remained consistent with 2020 levels. Comparative results of the third quartile DLP (476 mGy cm) and CTDIvol (4 mGy) values for 2022 indicated lower values for the third phase (400 entries) compared to 2020, with a 1.5-fold variation in DLP. The calculated ED for WB scans ranged from 2.6 to 7.1 mSv, with mean values of 4.7 ± 1.25 mSv, using a conversion factor (k = 0.0093 mSv/mGy/cm). The 2022 proposed national diagnostic reference levels (NDRLs) for HB (469 mGy cm, 4.0 mGy) were lower than the Swiss National Data (620 mGy cm, 6.0 mGy) and France (628 mGy cm, 6.6 mGy), but slightly higher than those of the United Kingdom (400 mGy cm, 4.3 mGy), despite the Swiss having about 5000 entries, France 1000 entries, and the United Kingdom 370 HB entries.

Conclusions: There was a 11.1% continuous improvement in NDRL for 2022 compared to 9.1% in 2020 and 13% in 2018, demonstrating a trend of enhanced optimization.

Advances in knowledge: The data established a trend of NDRL for WBCT (PET-CT) that can serve as a national databank for ongoing optimization. This promotes improvements in patient protection and quality care within the clinical environment of the State of Kuwait, aligning with the strategic goals of Kuwait Vision-2035.

Objectives: We aimed to analyse clinical outcomes of peripheral, early-stage non-small cell lung cancer (NSCLC) patients treated with stereotactic ablative body radiotherapy (SABR), and evaluate potential patient, tumour, and dosimetric variables influencing survival.

Methods: Data were collected retrospectively from patients treated between September 2012 and December 2016 and followed up until January 2021. Patient demographics, tumour characteristics, SABR dosimetric parameters, and survival data were collected from electronic patient medical records. Descriptive statistics were performed, and SPSS software was used for survival analysis.

Results: Eighty-nine patients were included of whom 49.5% were male and 50.5% female. Median age was 74 years. 98.8% of patients had T1-2 tumours and 89.9% underwent 55 Gy in 5 fractions. Median overall survival time was 58.7 months. On uni- and multi-variate analysis, neither patient nor tumour variables showed association with overall survival. However, planning target volume (PTV) and minimum dose to PTV correlated with overall survival. There was a signal for association between mean lung dose and overall survival on multivariate analysis.

Conclusions: Our long-term results show SABR is an effective treatment for peripheral, early-stage NSCLC with excellent overall survival, comparable to other series. Our study found only the PTV and minimum dose to PTV had an impact on overall survival, which demonstrates the importance of generating optimal SABR plans.

Advances in knowledge: Our work identified lung SABR dosimetric parameters that correlate with survival, which illustrates the importance of producing optimal lung SABR plans.