Search

BACKGROUND: Pancreatic cancer is a malignancy with one of the poorest prognoses amongst all cancers. Patients with unresectable tumours either receive palliative care or undergo various chemoradiotherapy regimens. Conventional techniques are often associated with acute gastrointestinal toxicities, as adjacent critical structures such as the duodenum ultimately limits delivered doses. Stereotactic body radiotherapy (SBRT) is an advanced radiation technique that delivers highly ablative radiation split into several fractions, with a steep dose fall-off outside target volumes. AIM: To discuss the latest data on SBRT and whether there is a role for magnetic resonance-guided techniques in multimodal management of locally advanced, unresectable pancreatic cancer. METHODS: We conducted a search on multiple large databases to collate the latest records on radiotherapy techniques used to treat pancreatic cancer. Out of 1229 total records retrieved from our search, 36 studies were included in this review. RESULTS: Studies indicate that SBRT is associated with improved clinical efficacy and toxicity profiles compared to conventional radiotherapy techniques. Further dose escalation to the tumour with SBRT is limited by the poor soft-tissue visualisation of computed tomography imaging during radiation planning and treatment delivery. Magnetic resonance-guided techniques have been introduced to improve imaging quality, enabling treatment plan adaptation and re-optimisation before delivering each fraction. CONCLUSION: Therefore, SBRT may lead to improved survival outcomes and safer toxicity profiles compared to conventional techniques, and the addition of magnetic resonance-guided techniques potentially allows dose escalation and conversion of unresectable tumours to operable cases.

BACKGROUND: Non-alcoholic fatty liver disease (NAFLD) is highly prevalent in people with diabetes with no available treatment. AIM: To explore the effect of testosterone treatment on liver. Testosterone therapy improves insulin resistance and reduces total body fat, but its impact on the liver remains poorly studied. METHODS: This secondary analysis of a 40 wk, randomised, double-blinded, placebo-controlled trial of intramuscular testosterone undecanoate in men with type 2 diabetes and lowered serum testosterone concentrations evaluated the change in hepatic steatosis as measured by liver fat fraction on magnetic resonance imaging (MRI). RESULTS: Of 88 patients enrolled in the index study, 39 had liver MRIs of whom 20 received testosterone therapy and 19 received placebo. All patients had > 5% hepatic steatosis at baseline and 38 of 39 patients met diagnostic criteria for NAFLD. Median liver fat at baseline was 15.0% (IQR 11.5%-21.1%) in the testosterone and 18.4% (15.0%-28.9%) in the placebo group. Median ALT was 34units/L (26-38) in the testosterone and 32units/L (25-52) in the placebo group. At week 40, patients receiving testosterone had a median reduction in absolute liver fat of 3.5% (IQR 2.9%-6.4%) compared with an increase of 1.2% in the placebo arm (between-group difference 4.7% P < 0.001). After controlling for baseline liver fat, testosterone therapy was associated with a relative reduction in liver fat of 38.3% (95% confidence interval 25.4%-49.0%, P < 0.001). CONCLUSION: Testosterone therapy was associated with a reduction in hepatic steatosis in men with diabetes and low serum testosterone. Future randomised studies of testosterone therapy in men with NAFLD focusing on liver-related endpoints are therefore justified.

Purpose: Delivery of endovascular clot retrieval (ECR) in acute stroke requires complex coordination of the patient and many stroke team members across different locations in the hospital. Not knowing each individual's location throughout the progress of a stroke call can cause miscommunication and inefficiencies in the delivery of ECR. 1 Machine learning (ML) can be used to develop a real time location system (RTLS) that determines a person's location by analysing their interaction with surrounding ubiquitous communications technology signals such as WiFi. 2, 3 This is called WiFi fingerprinting. We propose a WiFi fingerprinting model of RTLS to streamline delivery of ECR. Methods and Materials: In this proof-of-concept study, we collected WiFi signals from different hospital zones relevant to the ECR workflow, including the emergency department, CT scanner, angiography suite, intensive care units and stroke ward. We trained several ML algorithms including K nearest neighbors, decision tree, random forest, support vector machine and 2-ensemble models with labelled WiFi data. The same ML algorithms were then used to predict one investigator's location with unlabeled WiFi data. The accuracies of the different ML algorithms, in percentage of correct hospital zones prediction, were measured. Result(s): ML-based WiFi fingerprinting could accurately predict the investigator's location in different hospital zones relevant to the ECR workflow with up to 98% accuracy using random forest and support vector machine models. Conclusion(s): WiFi fingerprinting, a machine learning-based location tracking technology, has the potential to streamline delivery of acute stroke endovascular intervention by efficiently tracking patient and staff movement during a stroke call.

BACKGROUND: Delivery of acute stroke endovascular intervention can be challenging because it requires complex coordination of patient and staff across many different locations. In this proof-of-concept paper we (a) examine whether WiFi fingerprinting is a feasible machine learning (ML)-based real-time location system (RTLS) technology that can provide accurate real-time location information within a hospital setting, and (b) hypothesize its potential application in streamlining acute stroke endovascular intervention. METHOD(S): We conducted our study in a comprehensive stroke care unit in Melbourne, Australia that offers a 24-hour mechanical thrombectomy service. ML algorithms including K-nearest neighbors, decision tree, random forest, support vector machine and ensemble models were trained and tested on a public WiFi dataset and the study hospital WiFi dataset. The hospital dataset was collected using the WiFi explorer software (version 3.0.2) on a MacBook Pro (AirPort Extreme, Broadcom BCM43xx1.0). Data analysis was implemented in the Python programming environment using the scikit-learn package. The primary statistical measure for algorithm performance was the accuracy of location prediction. RESULT(S): ML-based WiFi fingerprinting can accurately predict the different hospital zones relevant in the acute endovascular intervention workflow such as emergency department, CT room and angiography suite. The most accurate algorithms were random forest and support vector machine, both of which were 98% accurate. The algorithms remain robust when new data points, which were distinct from the training dataset, were tested. CONCLUSION(S): ML-based RTLS technology using WiFi fingerprinting has the potential to streamline delivery of acute stroke endovascular intervention by efficiently tracking patient and staff movement during stroke calls.Copyright © Author(s) (or their employer(s)) 2021. No commercial re-use. See rights and permissions. Published by BMJ.

Artificial intelligence (AI) is making a profound impact in healthcare, with the number of AI applications in medicine increasing substantially over the past five years. In acute stroke, it is playing an increasingly important role in clinical decision-making. Contemporary advances have increased the amount of information - both clinical and radiological - which clinicians must consider when managing patients. In the time-critical setting of acute stroke, AI offers the tools to rapidly evaluate and consolidate available information, extracting specific predictions from rich, noisy data. It has been applied to the automatic detection of stroke lesions on imaging and can guide treatment decisions through the prediction of tissue outcomes and long-term functional outcomes. This review examines the current state of AI applications in stroke, exploring their potential to reform stroke care through clinical decision support, as well as the challenges and limitations which must be addressed to facilitate their acceptance and adoption for clinical use.Copyright © 2021 The Royal Australian and New Zealand College of Radiologists

Artificial intelligence is a rapidly evolving field, with modern technological advances and the growth of electronic health data opening new possibilities in diagnostic radiology. In recent years, the performance of deep learning (DL) algorithms on various medical image tasks have continually improved. DL algorithms have been proposed as a tool to detect various forms of intracranial hemorrhage on non-contrast computed tomography (NCCT) of the head. In subtle, acute cases, the capacity for DL algorithm image interpretation support might improve the diagnostic yield of CT for detection of this time-critical condition, potentially expediting treatment where appropriate and improving patient outcomes. However, there are multiple challenges to DL algorithm implementation, such as the relative scarcity of labeled datasets, the difficulties in developing algorithms capable of volumetric medical image analysis, and the complex practicalities of deployment into clinical practice. This review examines the literature and the approaches taken in the development of DL algorithms for the detection of intracranial hemorrhage on NCCT head studies. Considerations in crafting such algorithms will be discussed, as well as challenges which must be overcome to ensure effective, dependable implementations as automated tools in a clinical setting.Copyright ©

BACKGROUND: Delivery of acute stroke endovascular intervention can be challenging because it requires complex coordination of patient and staff across many different locations. In this proof-of-concept paper we (a) examine whether WiFi fingerprinting is a feasible machine learning (ML)-based real-time location system (RTLS) technology that can provide accurate real-time location information within a hospital setting, and (b) hypothesize its potential application in streamlining acute stroke endovascular intervention. METHOD(S): We conducted our study in a comprehensive stroke care unit in Melbourne, Australia that offers a 24-hour mechanical thrombectomy service. ML algorithms including K-nearest neighbors, decision tree, random forest, support vector machine and ensemble models were trained and tested on a public WiFi dataset and the study hospital WiFi dataset. The hospital dataset was collected using the WiFi explorer software (version 3.0.2) on a MacBook Pro (AirPort Extreme, Broadcom BCM43xx1.0). Data analysis was implemented in the Python programming environment using the scikit-learn package. The primary statistical measure for algorithm performance was the accuracy of location prediction. RESULT(S): ML-based WiFi fingerprinting can accurately predict the different hospital zones relevant in the acute endovascular intervention workflow such as emergency department, CT room and angiography suite. The most accurate algorithms were random forest and support vector machine, both of which were 98% accurate. The algorithms remain robust when new data points, which were distinct from the training dataset, were tested. CONCLUSION(S): ML-based RTLS technology using WiFi fingerprinting has the potential to streamline delivery of acute stroke endovascular intervention by efficiently tracking patient and staff movement during stroke calls.Copyright © Author(s) (or their employer(s)) 2021. No commercial re-use. See rights and permissions. Published by BMJ.

Purpose: Delivery of endovascular clot retrieval (ECR) in acute stroke requires complex coordination of the patient and many stroke team members across different locations in the hospital. Not knowing each individual's location throughout the progress of a stroke call can cause miscommunication and inefficiencies in the delivery of ECR. 1 Machine learning (ML) can be used to develop a real time location system (RTLS) that determines a person's location by analysing their interaction with surrounding ubiquitous communications technology signals such as WiFi. 2, 3 This is called WiFi fingerprinting. We propose a WiFi fingerprinting model of RTLS to streamline delivery of ECR. Methods and Materials: In this proof-of-concept study, we collected WiFi signals from different hospital zones relevant to the ECR workflow, including the emergency department, CT scanner, angiography suite, intensive care units and stroke ward. We trained several ML algorithms including K nearest neighbors, decision tree, random forest, support vector machine and 2-ensemble models with labelled WiFi data. The same ML algorithms were then used to predict one investigator's location with unlabeled WiFi data. The accuracies of the different ML algorithms, in percentage of correct hospital zones prediction, were measured. Result(s): ML-based WiFi fingerprinting could accurately predict the investigator's location in different hospital zones relevant to the ECR workflow with up to 98% accuracy using random forest and support vector machine models. Conclusion(s): WiFi fingerprinting, a machine learning-based location tracking technology, has the potential to streamline delivery of acute stroke endovascular intervention by efficiently tracking patient and staff movement during a stroke call.

Artificial intelligence (AI) is making a profound impact in healthcare, with the number of AI applications in medicine increasing substantially over the past five years. In acute stroke, it is playing an increasingly important role in clinical decision-making. Contemporary advances have increased the amount of information - both clinical and radiological - which clinicians must consider when managing patients. In the time-critical setting of acute stroke, AI offers the tools to rapidly evaluate and consolidate available information, extracting specific predictions from rich, noisy data. It has been applied to the automatic detection of stroke lesions on imaging and can guide treatment decisions through the prediction of tissue outcomes and long-term functional outcomes. This review examines the current state of AI applications in stroke, exploring their potential to reform stroke care through clinical decision support, as well as the challenges and limitations which must be addressed to facilitate their acceptance and adoption for clinical use.Copyright © 2021 The Royal Australian and New Zealand College of Radiologists

Artificial intelligence is a rapidly evolving field, with modern technological advances and the growth of electronic health data opening new possibilities in diagnostic radiology. In recent years, the performance of deep learning (DL) algorithms on various medical image tasks have continually improved. DL algorithms have been proposed as a tool to detect various forms of intracranial hemorrhage on non-contrast computed tomography (NCCT) of the head. In subtle, acute cases, the capacity for DL algorithm image interpretation support might improve the diagnostic yield of CT for detection of this time-critical condition, potentially expediting treatment where appropriate and improving patient outcomes. However, there are multiple challenges to DL algorithm implementation, such as the relative scarcity of labeled datasets, the difficulties in developing algorithms capable of volumetric medical image analysis, and the complex practicalities of deployment into clinical practice. This review examines the literature and the approaches taken in the development of DL algorithms for the detection of intracranial hemorrhage on NCCT head studies. Considerations in crafting such algorithms will be discussed, as well as challenges which must be overcome to ensure effective, dependable implementations as automated tools in a clinical setting.Copyright ©

BACKGROUND: While clinical guidelines recommend hepatocellular carcinoma (HCC) surveillance for at-risk individuals, reported surveillance rates in the United States and Europe remain disappointingly low. AIM: To quantify HCC surveillance in an Australian cohort, and assess for factors associated with surveillance underutilisation. METHODS: All patients undergoing HCC surveillance liver ultrasounds between January 1, 2018 to June 30, 2018 at a tertiary hospital in Melbourne, Australia, were followed until July 31, 2020, or when surveillance was no longer required. The primary outcome was the percentage of time up-to-date with HCC surveillance (PTUDS). Quantile regression was performed to determine the impact of factors associated with HCC surveillance underutilisation. RESULTS: Among 775 at-risk patients followed up for a median of 27.5 months, the median PTUDS was 84.2% (IQR: 66.3%-96.3%). 85.0% of patients were followed up by specialist gastroenterologists. Amongst those receiving specialist care, quantile regression demonstrated differential associations at various quantile levels of PTUDS for several factors. Older age at the 25th quantile (estimate 0.002 per percent, P = 0.03), and cirrhotic status at the 75th quantile (estimate 0.021, P = 0.017), were significantly associated with greater percentage of time up-to-date. African ethnicity (estimate -0.089, P = 0.048) and a culturally and linguistically diverse (CALD) background (estimate -0.063, P = 0.01) were significantly associated with lower PTUDS at the 50th quantile, and again for CALD at the 75th quantile (estimate -0.026, P = 0.045). CONCLUSION: While median PTUDS in this Australian cohort study was 84.2%, awareness of the impact of specific factors across PTUDS quantiles can aid targeted interventions towards improved HCC surveillance.

Artificial intelligence (AI) is making a profound impact in healthcare, with the number of AI applications in medicine increasing substantially over the past five years. In acute stroke, it is playing an increasingly important role in clinical decision-making. Contemporary advances have increased the amount of information - both clinical and radiological - which clinicians must consider when managing patients. In the time-critical setting of acute stroke, AI offers the tools to rapidly evaluate and consolidate available information, extracting specific predictions from rich, noisy data. It has been applied to the automatic detection of stroke lesions on imaging and can guide treatment decisions through the prediction of tissue outcomes and long-term functional outcomes. This review examines the current state of AI applications in stroke, exploring their potential to reform stroke care through clinical decision support, as well as the challenges and limitations which must be addressed to facilitate their acceptance and adoption for clinical use.

BACKGROUND: Management of single small hepatocellular carcinoma (HCC) is straightforward with curative outcomes achieved by locoregional therapy or resection. Liver transplantation is often considered for multiple small or single large HCC. Management of two small HCC whether presenting synchronously or sequentially is less clear. AIM: To define the outcomes of patients presenting with two small HCC. METHODS: Retrospective review of HCC databases from multiple institutions of patients with either two synchronous or sequential HCC ≤ 3 cm between January 2000 and March 2018. Primary outcomes were overall survival (OS) and transplant-free survival (TFS). RESULTS: 104 patients were identified (male n = 89). Median age was 63 years (interquartile range 58-67.75) and the most common aetiology of liver disease was hepatitis C (40.4%). 59 (56.7%) had synchronous HCC and 45 (43.3%) had sequential. 36 patients died (34.6%) and 25 were transplanted (24.0%). 1, 3 and 5-year OS was 93.0%, 66.1% and 62.3% and 5-year post-transplant survival was 95.8%. 1, 3 and 5-year TFS was 82.1%, 45.85% and 37.8%. When synchronous and sequential groups were compared, OS (1,3 and 5 year synchronous 91.3%, 63.8%, 61.1%, sequential 95.3%, 69.5%, 64.6%, P = 0.41) was similar but TFS was higher in the sequential group (1,3 and 5 year synchronous 68.5%, 37.3% and 29.7%, sequential 93.2%, 56.6%, 48.5%, P = 0.02) though this difference did not remain during multivariate analysis. CONCLUSION: TFS in patients presenting with two HCC ≤ 3 cm is poor regardless of the timing of the second tumor. All patients presenting with two small HCC should be considered for transplantation.

INTRODUCTION: Hepatic steatosis is a common incidental finding on computed tomography (CT) in patients presenting to the emergency department (ED). The aims of our study were to assess the prevalence of hepatic steatosis in ED patients with suspected renal colic and to assess documentation in radiology reports and medical charts correlated with alanine transaminase (ALT) levels. METHODS: Over 18 months from January 2016 to June 2017, all unenhanced CTs performed for suspected renal colic were reviewed. Quantitative assessment measuring hepatic and splenic attenuation in Hounsfield Units was performed. Hepatic steatosis was defined using multiple CT criteria including liver/spleen (L/S) ratio. Radiology reports, medical charts and ALT levels, if collected within 24 h of CT, were reviewed. RESULTS: A total of 1290 patients were included with a median age 52.5 years (range 16-98) and male predominance (835 [64.7%]). A total of 336 (26%) patients had hepatic steatosis measured by L/S ratio of ≤ 1.0. Ninety-four patients (28%) had radiology reports noting steatosis. Documentation in medical charts was noted in 18 of the 94 patients (19.1%) for whom steatosis was reported. Liver enzymes were available for 704 (54.6%) patients. There was a significantly higher mean ALT level in patients with hepatic steatosis (42.2 U/L; 95% CI 38.4-46.0) compared to patients without (28.8 U/L; 95% CI 25.7-31.9) (P < 0.0001). CONCLUSION: Our findings highlight multiple gaps in the reporting and evaluation of hepatic steatosis among radiologists and emergency clinicians alike. Recognising and reporting this incidental finding may impact health outcomes.

INTRODUCTION: Hepatic steatosis is a relatively common incidental finding on computed tomography (CT) studies performed for patients in the emergency department (ED). The aim of our survey was to explore the preferences and perspectives of emergency physicians regarding reporting of incidental findings with a focus on hepatic steatosis. METHODS: A prospective web-based questionnaire was conducted and distributed electronically to emergency clinicians with anonymous collection of responses. RESULTS: A total of 236 responses were received. The true response rate could not be determined due to different methods of electronic distribution. However, there was an estimated representation of 8.3% for ED physicians and 2.5% for trainees. The median time spent on the survey was less than 3 minutes. Seventy-seven per cent answered yes to giving an incidental finding more significance if mentioned in the conclusion section. More than half of respondents (60.2%) reported that they would like hepatic steatosis to be mentioned in a CT report while 30% reported that it was irrelevant in the emergency setting and 10% reported that they did not want it mentioned in the report. The majority (83.1%) reported that they would include this finding in the discharge summary for GP follow-up and less than half (44.1%) would mention it to patients. CONCLUSION: Our survey highlights the importance of clear communication between radiologists and ED physicians when incidental findings are encountered. Radiologists play an important role in alerting ED physicians and clinicians who have access to patients' radiology reports to the presence of incidental findings including hepatic steatosis.

BACKGROUND: The purpose of this study was to review the value of bone scans (BS) in the assessment of bone metastases from early-stage hepatocellular carcinoma (HCC) in patients assessed or waiting for liver transplant (LTx). METHODS: We reviewed BS studies performed at our center for patients with early-stage HCC either being assessed for LTx, or on the waiting list for LTx, from January 2010 to May 2017. The BS findings were classified as positive, equivocal, or negative. Correlation with final outcome based on clinical and radiological follow-up was performed. RESULTS: There were 360 BS performed in 186 patients during the study period with a mean age of 58.7 years (range, 34.9-70.4 years) and most were male patients (161/186 [86.6%]). None of the BSs resulted in delisting of patients from the LTx waiting list. Three BSs were reported as positive for metastases. All 3 were proven to be false positives on follow-up. Fourteen studies reported equivocal findings, none of which were confirmed to be metastases on follow-up. There was 1 false-negative BS: a bone metastasis was detected incidentally on magnetic resonance imaging and proven on biopsy. CONCLUSIONS: We have demonstrated that the diagnostic yield of BS in early HCC patients who are candidates for LTx is minimal, challenging the current inclusion of BS in guidelines for staging these HCC patients.

Our case report demonstrates the use of phase contrast magnetic resonance imaging (MRI) in monitoring the functional status of liver vasculature in a patient with hereditary hemorrhagic telangiectasia (HHT) who was treated with bevacizumab. Our report provides additional information that can be further utilized in clinical settings and research.