Search

OBJECTIVES: Around 30% of patients undergoing surgical resection for drug-resistant mesial temporal lobe epilepsy (MTLE) do not obtain seizure freedom. Success of anterior temporal lobe resection (ATLR) critically depends on the careful selection of surgical candidates, aiming at optimizing seizure freedom while minimizing postoperative morbidity. Structural MRI and FDG-PET neuroimaging are routinely used in presurgical assessment and guide the decision to proceed to surgery. In this study, we evaluate the potential of machine learning techniques applied to standard presurgical MRI and PET imaging features to provide enhanced prognostic value relative to current practice. METHODS: Eighty two patients with drug resistant MTLE were scanned with FDG-PET pre-surgery and T1-weighted MRI pre- and postsurgery. From these images the following features of interest were derived: volume of temporal lobe (TL) hypometabolism, % of extratemporal hypometabolism, presence of contralateral TL hypometabolism, presence of hippocampal sclerosis, laterality of seizure onset volume of tissue resected and % of temporal lobe hypometabolism resected. These measures were used as predictor variables in logistic regression, support vector machines, random forests and artificial neural networks. RESULTS: In the study cohort, 24 of 82 (28.3%) who underwent an ATLR for drug-resistant MTLE did not achieve Engel Class I (i.e., free of disabling seizures) outcome at a minimum of 2 years of postoperative follow-up. We found that machine learning approaches were able to predict up to 73% of the 24 ATLR surgical patients who did not achieve a Class I outcome, at the expense of incorrect prediction for up to 31% of patients who did achieve a Class I outcome. Overall accuracies ranged from 70% to 80%, with an area under the receiver operating characteristic curve (AUC) of .75-.81. We additionally found that information regarding overall extent of both total and significantly hypometabolic tissue resected

Brain computed tomography (CTB) scans are widely used to evaluate intracranial pathology. The implementation and adoption of CTB has led to clinical improvements. However, interpretation errors occur and may have substantial morbidity and mortality implications for patients. Deep learning has shown promise for facilitating improved diagnostic accuracy and triage. This research charts the potential of deep learning applied to the analysis of CTB scans. It draws on the experience of practicing clinicians and technologists involved in development and implementation of deep learning-based clinical decision support systems. We consider the past, present and future of the CTB, along with limitations of existing systems as well as untapped beneficial use cases. Implementing deep learning CTB interpretation systems and effectively navigating development and implementation risks can deliver many benefits to clinicians and patients, ultimately improving efficiency and safety in healthcare.

Brain computed tomography (CTB) scans are widely used to evaluate intracranial pathology. The implementation and adoption of CTB has led to clinical improvements. However, interpretation errors occur and may have substantial morbidity and mortality implications for patients. Deep learning has shown promise for facilitating improved diagnostic accuracy and triage. This research charts the potential of deep learning applied to the analysis of CTB scans. It draws on the experience of practicing clinicians and technologists involved in development and implementation of deep learning-based clinical decision support systems. We consider the past, present and future of the CTB, along with limitations of existing systems as well as untapped beneficial use cases. Implementing deep learning CTB interpretation systems and effectively navigating development and implementation risks can deliver many benefits to clinicians and patients, ultimately improving efficiency and safety in healthcare.

OBJECTIVES: To evaluate the ability of a commercially available comprehensive chest radiography deep convolutional neural network (DCNN) to detect simple and tension pneumothorax, as stratified by the following subgroups: the presence of an intercostal drain; rib, clavicular, scapular or humeral fractures or rib resections; subcutaneous emphysema and erect versus non-erect positioning. The hypothesis was that performance would not differ significantly in each of these subgroups when compared with the overall test dataset. DESIGN: A retrospective case-control study was undertaken. SETTING: Community radiology clinics and hospitals in Australia and the USA. PARTICIPANTS: A test dataset of 2557 chest radiography studies was ground-truthed by three subspecialty thoracic radiologists for the presence of simple or tension pneumothorax as well as each subgroup other than positioning. Radiograph positioning was derived from radiographer annotations on the images. OUTCOME MEASURES: DCNN performance for detecting simple and tension pneumothorax was evaluated over the entire test set, as well as within each subgroup, using the area under the receiver operating characteristic curve (AUC). A difference in AUC of more than 0.05 was considered clinically significant. RESULTS: When compared with the overall test set, performance of the DCNN for detecting simple and tension pneumothorax was statistically non-inferior in all subgroups. The DCNN had an AUC of 0.981 (0.976-0.986) for detecting simple pneumothorax and 0.997 (0.995-0.999) for detecting tension pneumothorax. CONCLUSIONS: Hidden stratification has significant implications for potential failures of deep learning when applied in clinical practice. This study demonstrated that a comprehensively trained DCNN can be resilient to hidden stratification in several clinically meaningful subgroups in detecting pneumothorax.

Renal echo planar diffusion tensor imaging (DTI) has clinical potential but suffers from geometric distortion. We evaluated feasibility of reversed gradient distortion correction in 10 diabetic patients and 6 volunteers. Renal area, apparent diffusion coefficient, fractional anisotropy, and tensor eigenvalues were measured on uncorrected and distortion-corrected DTI. Corrected DTI correlated better than uncorrected DTI (r = 0.904 vs 0.840, P = 0.002) with reference anatomic T2-weighted imaging, with no significant difference in DTI metrics.

BACKGROUND: Understanding the risk factors and pregnancy outcomes in women affected by Type 1 and Type 2 diabetes is important for pre-pregnancy counselling. AIM: To explore differences in pregnancy outcomes between women with Type 1 and Type 2 diabetes, and healthy controls, and to examine the relationships between potential adverse risk factors and pregnancy outcomes in this cohort of women. METHODS: This is a 10-year retrospective study of women with Type 1 diabetes (n = 92), Type 2 diabetes (n = 106) and healthy women without diabetes (controls) (n = 119) from a tertiary obstetric centre. Clinical and biochemical characteristics of women with Type 1 and Type 2 diabetes were determined and related to major obstetric outcomes using univariate analysis. RESULTS: Women with pre-existing diabetes had higher adverse pregnancy outcomes (preeclampsia, emergency caesarean section, preterm birth <32 and 37 weeks, large for gestational age, neonatal jaundice, Apgar score < 7 at 5 min, neonatal intensive care admission and neonatal hypoglycaemia) compared to controls. A higher birth weight gestational centile (97.4% vs 72.4%, P = 0.001) and large for gestational age rate (63.4% vs 35.8%, P = 0.001) were observed in Type 1 diabetes compared to Type 2 diabetes. There were no differences in other outcomes between women with Type 1 and 2 diabetes. CONCLUSION: In this exploratory study, risk factors for maternal adverse outcomes differ between Type 1 and Type 2 diabetes. Maternal and foetal adverse outcomes were higher in pregnancies affected by diabetes compared to healthy women but occurred with similar frequency in women with Type 1 and Type 2 diabetes.

Despite its simple acquisition technique, the chest X-ray remains the most common first-line imaging tool for chest assessment globally. Recent evidence for image analysis using modern machine learning points to possible improvements in both the efficiency and the accuracy of chest X-ray interpretation. While promising, these machine learning algorithms have not provided comprehensive assessment of findings in an image and do not account for clinical history or other relevant clinical information. However, the rapid evolution in technology and evidence base for its use suggests that the next generation of comprehensive, well-tested machine learning algorithms will be a revolution akin to early advances in X-ray technology. Current use cases, strengths, limitations and applications of chest X-ray machine learning systems are discussed.

ObjectivesTo evaluate the ability of a commercially available comprehensive chest radiography deep convolutional neural network (DCNN) to detect simple and tension pneumothorax, as stratified by the following subgroups: the presence of an intercostal drain; rib, clavicular, scapular or humeral fractures or rib resections; subcutaneous emphysema and erect versus non-erect positioning. The hypothesis was that performance would not differ significantly in each of these subgroups when compared with the overall test dataset.DesignA retrospective case–control study was undertaken.SettingCommunity radiology clinics and hospitals in Australia and the USA.ParticipantsA test dataset of 2557 chest radiography studies was ground-truthed by three subspecialty thoracic radiologists for the presence of simple or tension pneumothorax as well as each subgroup other than positioning. Radiograph positioning was derived from radiographer annotations on the images.Outcome measuresDCNN performance for detecting simple and tension pneumothorax was evaluated over the entire test set, as well as within each subgroup, using the area under the receiver operating characteristic curve (AUC). A difference in AUC of more than 0.05 was considered clinically significant.ResultsWhen compared with the overall test set, performance of the DCNN for detecting simple and tension pneumothorax was statistically non-inferior in all subgroups. The DCNN had an AUC of 0.981 (0.976–0.986) for detecting simple pneumothorax and 0.997 (0.995–0.999) for detecting tension pneumothorax.ConclusionsHidden

OBJECTIVES: Artificial intelligence (AI) algorithms have been developed to detect imaging features on chest X-ray (CXR) with a comprehensive AI model capable of detecting 124 CXR findings being recently developed. The aim of this study was to evaluate the real-world usefulness of the model as a diagnostic assistance device for radiologists. DESIGN: This prospective real-world multicentre study involved a group of radiologists using the model in their daily reporting workflow to report consecutive CXRs and recording their feedback on level of agreement with the model findings and whether this significantly affected their reporting. SETTING: The study took place at radiology clinics and hospitals within a large radiology network in Australia between November and December 2020. PARTICIPANTS: Eleven consultant diagnostic radiologists of varying levels of experience participated in this study. PRIMARY AND SECONDARY OUTCOME MEASURES: Proportion of CXR cases where use of the AI model led to significant material changes to the radiologist report, to patient management, or to imaging recommendations. Additionally, level of agreement between radiologists and the model findings, and radiologist attitudes towards the model were assessed. RESULTS: Of 2972 cases reviewed with the model, 92 cases (3.1%) had significant report changes, 43 cases (1.4%) had changed patient management and 29 cases (1.0%) had further imaging recommendations. In terms of agreement with the model, 2569 cases showed complete agreement (86.5%). 390 (13%) cases had one or more findings rejected by the radiologist. There were 16 findings across 13 cases (0.5%) deemed to be missed by the model. Nine out of 10 radiologists felt their accuracy was improved with the model and were more positive towards AI poststudy. CONCLUSIONS: Use of an AI model in a real-world reporting environment significantly improved radiologist reporting and showed good agreement with radiologists, highlighting the potential for AI diagn

OBJECTIVE: To determine variables that could facilitate safe discharge from the ED following a single high-sensitivity troponin I (HsTnI) result to exclude acute myocardial infarction (AMI). METHODS: A retrospective cohort study was performed at a tertiary hospital of all patients that had serial HsTnI performed within 12 h of arrival to the ED over a 3 year period. The primary exposure variable of interest was a very low troponin initial result (HsTnI <5 ng/L). Medical record review and risk stratification score calculations were undertaken for all patients with the exposure variable of interest and an abnormal second troponin measurement (HsTnI ≥16 ng/L in women and HsTnI ≥26 ng/L in men). RESULTS: There were 11 970 patients who presented between 1 July 2013 and 30 June 2016 that had serial HsTnI measurements performed. Of these, 4172 (34.9%) patients had an initial HsTnI measurement <5 ng/L. Of the patients with an initial HsTnI <5 ng/L that met inclusion criteria, 56 (1.3%) had a second troponin result above the 99th percentile and 32 (0.8%) cases of non-ST elevation myocardial infarction were diagnosed as well as 15 (0.4%) cases of ST elevation myocardial infarction. There were 44 (93.6%) of all AMI cases that met criteria for high-risk presentations under the National Heart Foundation of Australia guidelines. The negative predictive value of an initial HsTnI <5 ng/L to exclude AMI was 98.9% (95% confidence interval 98.5-99.1). CONCLUSIONS: This supports the utilisation of a rapid rule out strategy to exclude AMI for patients that have an initial HsTnI measurement <5 ng/L in conjunction with a robust risk assessment.

Large birthweight, or macrosomia, is one of the commonest complications for pregnancies affected by diabetes. As macrosomia is associated with an increased risk of a number of adverse outcomes for both the mother and offspring, accurate antenatal prediction of fetal macrosomia could be beneficial in guiding appropriate models of care and interventions that may avoid or reduce these associated risks. However, current prediction strategies which include physical examination and ultrasound assessment, are imprecise. Biomarkers are proving useful in various specialties and may offer a new avenue for improved prediction of macrosomia. Prime biomarker candidates in pregnancies with diabetes include maternal glycaemic markers (glucose, 1,5-anhydroglucitol, glycosylated haemoglobin) and hormones proposed implicated in placental nutrient transfer (adiponectin and insulin-like growth factor-1). There is some support for an association of these biomarkers with birthweight and/or macrosomia, although current evidence in this emerging field is area is still limited. Thus, although biomarkers hold promise, further investigation is needed to elucidate the potential clinical utility of biomarkers for macrosomia prediction for pregnancies affected by diabetes.

Diabetic kidney disease (DKD) represents a major component of the health burden associated with type 1 and type 2 diabetes. Recent advances have produced an explosion of 'novel' assay-based risk markers for DKD, though clinical use remains restricted. Although many patients with progressive DKD follow a classical albuminuria-based pathway, non-albuminuric DKD progression is now well recognized. In general, the following clinical and biochemical characteristics have been associated with progressive DKD in both type 1 and type 2 diabetes: increased hemoglobin A1c, systolic blood pressure, albuminuria grade, early glomerular filtration rate decline, duration of diabetes, age (including pubertal onset) and serum uric acid; the presence of concomitant microvascular complications; and positive family history. The same is true in type 2 diabetes for male sex category, in patients following an albuminuric pathway to DKD, and also true for the presence of increased pulse wave velocity. The following baseline clinical characteristics have been proposed as risk factors for DKD progression, but with further research required to assess the nature of any relationship: dyslipidemia (including low-density lipoprotein, total and high-density lipoprotein cholesterol); elevated body mass index; smoking status; hyperfiltration; decreases in vitamin D, hemoglobin and uric acid excretion (all known consequences of advanced DKD); and patient test result visit-to-visit variability (hemoglobin A1c, blood pressure and high-density lipoprotein cholesterol). The development of multifactorial 'renal risk equations' for type 2 diabetes has the potential to simplify the task of DKD prognostication; however, there are currently none for type 1 diabetes-specific populations. Significant progress has been made in the prediction of DKD progression using readily available clinical data, though further work is required to elicit the role of several variables, and to consolidate data to facilitate clini

Objective: A substantial number of children exposed to gestational opioids have neurodevelopmental, behavioral and cognitive problems. Opioids are not neuroteratogens but whether they affect the developing brain in more subtle ways (for example, volume loss) is unclear. We aimed to determine the feasibility of using magnetic resonance imaging (MRI) to assess volumetric changes in healthy opioid-exposed infants. Study design: Observational pilot cohort study conducted in two maternity hospitals in New South Wales, Australia. Maternal history and neonatal urine and meconium screens were obtained to confirm drug exposure. Volumetric analysis of MRI scans was performed with the ITK-snap program. Result: Scans for 16 infants (mean (s.d.) gestational age: 40.9 (1.5) weeks, birth weight: 3022.5 (476.6) g, head circumference (HC): 33.7 (1.5 cm)) were analyzed. Six (37.5%) infants had HC methadone, four used buprenorphine and 11 used more than one opioid (including heroin, seven). All scans were structurally normal whole brain volumes (357.4 (63.8)) and basal ganglia (14.5 (3.5)) ml were significantly smaller than population means (425.4 (4.8), 17.1 (4.4) ml, respectively) but lateral ventricular volumes (3.5 (1.8) ml) were larger than population values (2.1(1.5)) ml. Conclusion: Our pilot study suggests that brain volumes of opioid-exposed babies may be smaller than population means and that specific regions, for example, basal ganglia, that are involved in neurotransmission, may be particularly affected. Larger studies including correlation with neurodevelopmental outcomes are warranted to substantiate this finding.

Objective: A substantial number of children exposed to gestational opioids have neurodevelopmental, behavioral and cognitive problems. Opioids are not neuroteratogens but whether they affect the developing brain in more subtle ways (for example, volume loss) is unclear. We aimed to determine the feasibility of using magnetic resonance imaging (MRI) to assess volumetric changes in healthy opioid-exposed infants. Study design: Observational pilot cohort study conducted in two maternity hospitals in New South Wales, Australia. Maternal history and neonatal urine and meconium screens were obtained to confirm drug exposure. Volumetric analysis of MRI scans was performed with the ITK-snap program. Result: Scans for 16 infants (mean (s.d.) gestational age: 40.9 (1.5) weeks, birth weight: 3022.5 (476.6) g, head circumference (HC): 33.7 (1.5 cm)) were analyzed. Six (37.5%) infants had HC methadone, four used buprenorphine and 11 used more than one opioid (including heroin, seven). All scans were structurally normal whole brain volumes (357.4 (63.8)) and basal ganglia (14.5 (3.5)) ml were significantly smaller than population means (425.4 (4.8), 17.1 (4.4) ml, respectively) but lateral ventricular volumes (3.5 (1.8) ml) were larger than population values (2.1(1.5)) ml. Conclusion: Our pilot study suggests that brain volumes of opioid-exposed babies may be smaller than population means and that specific regions, for example, basal ganglia, that are involved in neurotransmission, may be particularly affected. Larger studies including correlation with neurodevelopmental outcomes are warranted to substantiate this finding.

As technology continues to influence many aspects of our social and work lives, it is important that school experiences equip students the skills and knowledge that will enable them to develop into effective independent, creative, and lifelong learners to cope with the influx of changes. Given that teachers play a key role in the effective use of technology in education, there is a need to ensure that teacher education programs prepare teachers for the effective integration of ICT in the classrooms. We believe that there is a need to adopt a student-centered learning framework to design our ICT based Student-Centred Learning (SCL) curriculum for all pre-service teachers. In this paper, we presents parts of the findings from a curriculum review which evaluated 483 pre-service teachers' overall satisfaction level towards an ICT based SCL course. We also provide some recommendations to the ICT curriculum based on the results found. © 2007 Chwee Beng Lee, Timothy Teo, Ching Sing Chai, Doris Choy, Ashley Tan and Jimmy Seah.