BJR open最新文献

Objectives: High breast density is a risk factor for breast cancer and overlapping of glandular tissue can mask lesions thus lowering mammographic sensitivity. Also, dense breasts are more vulnerable to increase recall rate and false-positive results. New generations of artificial intelligence (AI) have been introduced to the realm of mammography. We aimed to assess the strengths and challenges of adopting artificial intelligence in reading mammograms of dense breasts.

Methods: This study included 6600 mammograms of dense patterns "c" and "d" and presented 4061 breast abnormalities. All the patients were subjected to full-field digital mammography, breast ultrasound, and their mammographic images were scanned by AI software (Lunit INSIGHT MMG).

Results: Diagnostic indices of the sono-mammography: a sensitivity of 98.71%, a specificity of 88.04%, a positive-predictive value of 80.16%, a negative-predictive value of 99.29%, and a diagnostic accuracy of 91.5%. AI-aided mammograms presented sensitivity of 88.29%, a specificity of 96.34%, a positive-predictive value of 92.2%, a negative-predictive value of 94.4%, and a diagnostic accuracy of 94.5% in its ability to read dense mammograms.

Conclusion: Dense breasts scanned with AI showed a notable reduction of mammographic misdiagnosis. Knowledge of such software challenges would enhance its application as a decision support tool to mammography in the diagnosis of cancer.

Advances in knowledge: Dense breast is challenging for radiologists and renders low sensitivity mammogram. Mammogram scanned by AI could be used to overcome such limitation, enhance the discrimination between benign and malignant breast abnormalities and the early detection of breast cancer.

Objective: To assess the diagnostic performance of morphological MRI features separately and in combination for distinguishing low- from high-grade soft tissue sarcoma (STS).

Methods and materials: We retrospectively analysed pre-treatment MRI examinations with T1, T2 with and without fat suppression (FS) and contrast-enhanced T1 obtained in 64 patients with STS categorized histologically as low (n = 21) versus high grade (n = 43). Two musculoskeletal radiologists blinded to histology evaluated MRI features. Diagnostic performance was calculated for each reader and for MRI features showing significant association with histology (p < 0.05). Logistic regression analysis was performed to develop a diagnostic model to identify high-grade STS.

Results: Among all evaluated MRI features, only six features had adequate interobserver reproducibility (kappa>0.5). Multivariate logistic regression analysis revealed a significant association with tumour grade for lesion heterogeneity on FS images, intratumoural enhancement≥51% of tumour volume and peritumoural enhancement for both readers (p < 0.05). For both readers, the presence of each of the three features yielded odds ratios for high grade versus low grade from 4.4 to 9.1 (p < 0.05). The sum of the positive features for each reader independent of reader expertise yielded areas under the curve (AUCs) > 0.8. The presence of ≥2 positive features indicated a high risk for high-grade sarcoma, whereas ≤1 positive feature indicated a low-to-moderate risk.

Conclusion: A diagnostic MRI score based on tumour heterogeneity, intratumoural and peritumoural enhancement enables identification of lesions that are likely to be high-grade as opposed to low-grade STS.

Advances in knowledge: Tumour heterogeneity in Fat Suppression sequence, intratumoural and peritumoural enhancement is identified as signs of high-grade sarcoma.

Artificial intelligence (AI) plays a crucial role in the future development of all healthcare sectors ranging from clinical assistance of physicians by providing accurate diagnosis, prognosis and treatment to the development of vaccinations and aiding in the combat against the Covid-19 global pandemic. AI has an important role in diagnostic radiology where the algorithms can be trained by large datasets to accurately provide a timely diagnosis of the radiological images given. This has led to the development of several AI algorithms that can be used in regions of scarcity of radiologists during the current pandemic by simply denoting the presence or absence of Covid-19 pneumonia in PCR positive patients on plain chest radiographs as well as in helping to levitate the over-burdened radiology departments by accelerating the time for report delivery. Plain chest radiography is the most common radiological study in the emergency department setting and is readily available, fast and a cheap method that can be used in triaging patients as well as being portable in the medical wards and can be used as the initial radiological examination in Covid-19 positive patients to detect pneumonic changes. Numerous studies have been done comparing several AI algorithms to that of experienced thoracic radiologists in plain chest radiograph reports measuring accuracy of each in Covid-19 patients. The majority of studies have reported performance equal or higher to that of the well-experienced thoracic radiologist in predicting the presence or absence of Covid-19 pneumonic changes in the provided chest radiographs.

Millions of breast imaging exams are performed each year in an effort to reduce the morbidity and mortality of breast cancer. Breast imaging exams are performed for cancer screening, diagnostic work-up of suspicious findings, evaluating extent of disease in recently diagnosed breast cancer patients, and determining treatment response. Yet, the interpretation of breast imaging can be subjective, tedious, time-consuming, and prone to human error. Retrospective and small reader studies suggest that deep learning (DL) has great potential to perform medical imaging tasks at or above human-level performance, and may be used to automate aspects of the breast cancer screening process, improve cancer detection rates, decrease unnecessary callbacks and biopsies, optimize patient risk assessment, and open up new possibilities for disease prognostication. Prospective trials are urgently needed to validate these proposed tools, paving the way for real-world clinical use. New regulatory frameworks must also be developed to address the unique ethical, medicolegal, and quality control issues that DL algorithms present. In this article, we review the basics of DL, describe recent DL breast imaging applications including cancer detection and risk prediction, and discuss the challenges and future directions of artificial intelligence-based systems in the field of breast cancer.

Objectives: To investigate the relationship between magnetization transfer (MT) imaging and tissue macromolecules in high-grade serous ovarian cancer (HGSOC) and whether MT ratio (MTR) changes following neoadjuvant chemotherapy (NACT).

Methods: This was a prospective observational study. 12 HGSOC patients were imaged before treatment. MTR was compared to quantified tissue histology and immunohistochemistry. For a subset of patients (n = 5), MT imaging was repeated after NACT. The Shapiro-Wilk test was used to assess for normality of data and Spearman's rank-order or Pearson's correlation tests were then used to compare MTR with tissue quantifications. The Wilcoxon signed-rank test was used to assess for changes in MTR after treatment.

Results: Treatment-naïve tumour MTR was 21.9 ± 3.1% (mean ± S.D.). MTR had a positive correlation with cellularity, rho = 0.56 (p < 0.05) and a negative correlation with tumour volume, ρ = -0.72 (p = 0.01). MTR did not correlate with the extracellular proteins, collagen IV or laminin (p = 0.40 and p = 0.90). For those patients imaged before and after NACT, an increase in MTR was observed in each case with mean MTR 20.6 ± 3.1% (median 21.1) pre-treatment and 25.6 ± 3.4% (median 26.5) post-treatment (p = 0.06).

Conclusion: In treatment-naïve HGSOC, MTR is associated with cellularity, possibly reflecting intracellular macromolecular concentration. MT may also detect the HGSOC response to NACT, however larger studies are required to validate this finding.

Advances in knowledge: MTR in HGSOC is influenced by cellularity. This may be applied to assess for cell changes following treatment.

Objectives: Coronary and cardiac calcification are frequent incidental findings on non-gated thoracic computed tomography (CT). However, radiologist opinions and practices regarding the reporting of incidental calcification are poorly understood.

Methods: UK radiologists were invited to complete this online survey, organised by the British Society of Cardiovascular Imaging (BSCI). Questions included anonymous information on subspecialty, level of training and reporting practices for incidental coronary artery, aortic valve, mitral and thoracic aorta calcification.

Results: The survey was completed by 200 respondents: 10% trainees and 90% consultants. Calcification was not reported by 11% for the coronary arteries, 22% for the aortic valve, 35% for the mitral valve and 37% for the thoracic aorta. Those who did not subspecialise in cardiac imaging were less likely to report coronary artery calcification (p = 0.005), aortic valve calcification (p = 0.001) or mitral valve calcification (p = 0.008), but there was no difference in the reporting of thoracic aorta calcification. Those who did not subspecialise in cardiac imaging were also less likely to provide management recommendations for coronary artery calcification (p < 0.001) or recommend echocardiography for aortic valve calcification (p < 0.001), but there was no difference for mitral valve or thoracic aorta recommendations.

Conclusion: Incidental coronary artery, valvular and aorta calcification are frequently not reported on thoracic CT and there are differences in reporting practices based on subspeciality.

Advances in knowledge: On routine thoracic CT, 11% of radiologists do not report coronary artery calcification. Radiologist reporting practices vary depending on subspeciality but not level of training.

Objectives: To assess the frequency and intensity of [¹⁸F]-prostate-specific membrane antigen (PSMA)-1007 axillary uptake in lymph nodes ipsilateral to COVID-19 vaccination with BNT162b2 (Pfizer-BioNTech) or mRNA-1273 (Moderna) in patients with prostate cancer referred for oncological [¹⁸F]-PSMA positron emission tomography (PET)/CT or PET/MR imaging.

Methods: 126 patients undergoing [¹⁸F]-PSMA PET/CT or PET/MR imaging were retrospectively included. [¹⁸F]-PSMA activity (maximum standardized uptake value) of ipsilateral axillary lymph nodes was measured and compared with the non-vaccinated contralateral side and with a non-vaccinated negative control group. [¹⁸F]-PSMA active lymph node metastases were measured to serve as quantitative reference.

Results: There was a significant difference in maximum standardized uptake value in ipsilateral and compared to contralateral axillary lymph nodes in the vaccination group (n = 63, p < 0.001) and no such difference in the non-vaccinated control group (n = 63, p = 0.379). Vaccinated patients showed mildly increased axillary lymph node [¹⁸F]-PSMA uptake as compared to non-vaccinated patients (p = 0.03). [¹⁸F]-PSMA activity of of lymph node metastases was significantly higher (p < 0.001) compared to axillary lymph nodes of vaccinated patients.

Conclusion: Our data suggest mildly increased [¹⁸F]-PSMA uptake after COVID-19 vaccination in ipsilateral axillary lymph nodes. However, given the significantly higher [¹⁸F]-PSMA uptake of prostatic lymph node metastases compared to "reactive" nodes after COVID-19 vaccination, no therapeutic and diagnostic dilemma is to be expected.

Advances in knowledge: No specific preparations or precautions (e.g. adaption of vaccination scheduling) need to be undertaken in patients undergoing [¹⁸F]-PSMA PET imaging after COVID-19 vaccination.

Objectives: To assess the impact of noise-optimised virtual monoenergetic imaging (VMI+) on image quality and diagnostic evaluation in abdominal dual-energy CT scans with impaired portal-venous contrast.

Methods: We screened 11,746 patients who underwent portal-venous abdominal dual-energy CT for cancer staging between 08/2014 and 11/2019 and identified those with poor portal-venous contrast.Standard linearly-blended image series and VMI+ image series at 40, 50, and 60 keV were reconstructed. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of abdominal organs and vascular structures were calculated. Image noise, image contrast and overall image quality were rated by three radiologists using 5-point Likert scale.

Results: 452 of 11,746 (4%) exams were poorly opacified. We excluded 190 cases due to incomplete datasets or multiple exams of the same patient with a final study group of 262. Highest CNR values in all abdominal organs (liver, 6.4 ± 3.0; kidney, 17.4 ± 7.5; spleen, 8.0 ± 3.5) and vascular structures (aorta, 16.0 ± 7.3; intrahepatic vein, 11.3 ± 4.7; portal vein, 15.5 ± 6.7) were measured at 40 keV VMI+ with significantly superior values compared to all other series. In subjective analysis, highest image contrast was seen at 40 keV VMI+ (4.8 ± 0.4), whereas overall image quality peaked at 50 keV VMI+ (4.2 ± 0.5) with significantly superior results compared to all other series (p < 0.001).

Conclusions: Image reconstruction using VMI+ algorithm at 50 keV significantly improves image contrast and image quality of originally poorly opacified abdominal CT scans and reduces the number of non-diagnostic scans.

Advances in knowledge: We validated the impact of VMI+ reconstructions in poorly attenuated DECT studies of the abdomen in a big data cohort.

Objective: The aim of this study was to explore opinions and views towards radiology AI among Saudi Arabian radiologists including both consultants and trainees.

Methods: A qualitative approach was adopted, with radiologists working in radiology departments in the Western region of Saudi Arabia invited to participate in this interview-based study. Semi-structured interviews (n = 30) were conducted with consultant radiologists and trainees. A qualitative data analysis framework was used based on Miles and Huberman's philosophical underpinnings.

Results: Several factors, such as lack of training and support, were attributed to the non-use of AI-based applications in clinical practice and the absence of radiologists' involvement in AI development. Despite the expected benefits and positive impacts of AI on radiology, a reluctance to use AI-based applications might exist due to a lack of knowledge, fear of error and concerns about losing jobs and/or power. Medical students' radiology education and training appeared to be influenced by the absence of a governing body and training programmes.

Conclusion: The results of this study support the establishment of a governing body or national association to work in parallel with universities in monitoring training and integrating AI into the medical education curriculum and residency programmes.

Advances in knowledge: An extensive debate about AI-based applications and their potential effects was noted, and considerable exceptions of transformative impact may occur when AI is fully integrated into clinical practice. Therefore, future education and training programmes on how to work with AI-based applications in clinical practice may be recommended.

Objectives: Muscle volume may reflect both strength and functional capability and hence is a parameter often measured to assess the effect of various interventions. The aim of the current study was to determine the sensitivity of muscle volume calculations on participant postural position and hence gauge possible errors that may arise in longitudinal studies, especially those where an intervention leads to large muscle changes and potentially the degree of spinal curvature.

Methods: Twenty healthy participants (22-49 years, 10 male and 10 female), were recruited and MRI images acquired with them lying in four different positions; neutral spine (P1), decreased lordosis (P2), increased lordosis (P3) and neutral spine repeated (P4). Images were analysed in Simpleware ScanIP, and lumbar muscle volume and Cobb's angle, as an indicator of spine curvature, determined.

Results: After comparing volume determinations, no statistically significant differences were found for P1 - P2 and P1 - P4, whereas significant changes were determined for P2 - P3 and P1 - P3. P2 and P3 represent the two extremes of spinal curvature with a difference in Cobb's angle of 17°. However, the mean difference between volume determinations was only 29 cm³. These results suggest the differences in muscle volume determinations are generally greater with increasing differences in curvature between measurements, but that overall the effects are small.

Conclusions: Thus, generally, spinal muscle volume determinations are robust in terms of participant positioning.

Advances in knowledge: Differences in muscle volume calculations appear to become larger the greater the difference in spinal curvature between positions. Thus, spinal curvature should not have a major impact on the results of spinal muscle volume determinations following interventions in longitudinal studies.