Effects of a comprehensive brain computed tomography deep learning model on radiologist detection accuracy.

OBJECTIVES: Non-contrast computed tomography of the brain (NCCTB) is commonly used to detect intracranial pathology but is subject to interpretation errors. Machine learning can augment clinical decision-making and improve NCCTB scan interpretation. This retrospective detection accuracy study assessed the performance of radiologists assisted by a deep learning model and compared the standalone performance of the model with that of unassisted radiologists. METHODS: A deep learning model was trained on 212,484 NCCTB scans drawn from a private radiology group in Australia. Scans from inpatient, outpatient, and emergency settings were included. Scan inclusion criteria were age ≥ 18 years and series slice thickness ≤ 1.5 mm. Thirty-two radiologists reviewed 2848 scans with and without the assistance of the deep learning system and rated their confidence in the presence of each finding using a 7-point scale. Differences in AUC and Matthews correlation coefficient (MCC) were calculated using a ground-truth gold standard. RESULTS: The model demonstrated an average area under the receiver operating characteristic curve (AUC) of 0.93 across 144 NCCTB findings and significantly improved radiologist interpretation performance. Assisted and unassisted radiologists demonstrated an average AUC of 0.79 and 0.73 across 22 grouped parent findings and 0.72 and 0.68 across 189 child findings, respectively. When assisted by the model, radiologist AUC was significantly improved for 91 findings (158 findings were non-inferior), and reading time was significantly reduced. CONCLUSIONS: The assistance of a comprehensive deep learning model significantly improved radiologist detection accuracy across a wide range of clinical findings and demonstrated the potential to improve NCCTB interpretation. CLINICAL RELEVANCE STATEMENT: This study evaluated a comprehensive CT brain deep learning model, which performed strongly, improved the performance of radiologists, and reduced interpretation time. The model may reduce errors, improve efficiency, facilitate triage, and better enable the delivery of timely patient care. KEY POINTS: • This study demonstrated that the use of a comprehensive deep learning system assisted radiologists in the detection of a wide range of abnormalities on non-contrast brain computed tomography scans. • The deep learning model demonstrated an average area under the receiver operating characteristic curve of 0.93 across 144 findings and significantly improved radiologist interpretation performance. • The assistance of the comprehensive deep learning model significantly reduced the time required for radiologists to interpret computed tomography scans of the brain.

Heterogeneous Strain Distribution in the Subchondral Bone of Human Osteoarthritic Femoral Heads, Measured with Digital Volume Correlation.

Osteoarthritis (OA) is a chronic disease, affecting approximately one third of people over the age of 45. Whilst the etiology and pathogenesis of the disease are still not well understood, mechanics play an important role in both the initiation and progression of osteoarthritis. In this study, we demonstrate the application of stepwise compression, combined with microCT imaging and digital volume correlation (DVC) to measure and evaluate full-field strain distributions within osteoarthritic femoral heads under uniaxial compression. A comprehensive analysis showed that the microstructural features inherent in OA bone did not affect the level of uncertainties associated with the applied methods. The results illustrate the localization of strains at the loading surface as well as in areas of low bone volume fraction and subchondral cysts. Trabecular thickness and connectivity density were identified as the only microstructural parameters with any association to the magnitude of local strain measured at apparent yield strain or the volume of bone exceeding yield strain. This work demonstrates a novel approach to evaluating the mechanical properties of the whole human femoral head in case of severe OA.

Abdominal compression increases upper airway collapsibility during sleep in obese male obstructive sleep apnea patients.

STUDY OBJECTIVES: Abdominal obesity, particularly common in centrally obese males, may have a negative impact on upper airway (UA) function during sleep. For example, cranial displacement of the diaphragm with raised intra-abdominal pressure may reduce axial tension exerted on the UA by intrathoracic structures and increase UA collapsibility during sleep. DESIGN: This study aimed to examine the effect of abdominal compression on UA function during sleep in obese male obstructive sleep apnea patients. SETTING: Participants slept in a sound-insulated room with physiologic measurements controlled from an adjacent room. PARTICIPANTS: Fifteen obese (body mass index: 34.5 +/- 1.1 kg/m2) male obstructive sleep apnea patients (apnea-hypopnea index: 58.1 +/- 6.8 events/h) aged 50 +/- 2.6 years participated. INTERVENTIONS: Gastric (PGA) and transdiaphragmatic pressures (P(DI)), UA closing pressure (UACP), UA airflow resistance (R(UA)), and changes in end-expiratory lung volume (EELV) were determined during stable stage 2 sleep with and without abdominal compression, achieved via inflation of a pneumatic cuff placed around the abdomen. UACP was assessed during brief mask occlusions. MEASUREMENTS AND RESULTS: Abdominal compression significantly decreased EELV by 0.53 +/- 0.24 L (P=0.045) and increased PGA (16.2 +/- 0.8 versus 10.8 +/- 0.7 cm H2O, P < 0.001), P(DI) (11.7 +/- 0.9 versus 7.6 +/- 1.2 cm H2O, P < 0.001) and UACP (1.4 +/- 0.8 versus 0.9 +/- 0.9 cm H2O, P = 0.039) but not R(UA)(6.5 +/- 1.4 versus 6.9 +/- 1.4 cm H2O x L/s, P=0.585). CONCLUSIONS: Abdominal compression negatively impacts on UA collapsibility during sleep and this effect may help explain strong associations between central obesity and obstructive sleep apnea.

Flow rate accuracy of ambulatory elastomeric and electronic infusion pumps when exposed to height and back pressures experienced during home infusion therapy.

Background: Elastomeric infusion pumps are widely used in the delivery of parenteral medications in the home, but real-life conditions may not match calibration or standardised testing conditions. This study investigated the impact of changes in infusion pump height and/or back pressure on infusion pump function. Methods: Volume delivered after one day, infusion duration, average and peak flow rates and time spent within stated accuracy were determined for four elastomeric and one electronic pump using gravimetric technique. Experiments were repeated after altering the height of the pump relative to the output (±40cm, ±20cm) and/or adding a back pressure (10-30mmHg) to the output of an attached catheter. Results: Under ideal operating conditions, the flow rate deviated from that specified by the manufacturer and between 88.5% and 99% of the total infusion volume was delivered. Varying the height or applying back pressure led to further changes in average flow rates and the volume of infusion solution delivered by the elastomeric pumps, but had little effect on the electronic pump. Conclusions: Clinicians should consider potential impact on drug delivery, safety and therapeutic effect for home infusion patients given variations in infusion pump performance observed in this study.

Upper airway surface tension but not upper airway collapsibility is elevated in primary Sjögren's syndrome.

STUDY OBJECTIVES: Primary Sjögren's syndrome is an autoimmune disease typified by xerostomia (dry mouth) that, in turn, could lead to increased saliva surface tension (gamma) and increased upper airway collapsibility. Fatigue, of unknown etiology, is also frequently reported by patients with primary Sjögren's syndrome. Recent preliminary data indicate a high prevalence of obstructive sleep apnea in healthy-weight women with primary Sjögren's syndrome. Concurrent research highlights a significant role of gamma in the maintenance of upper airway patency. The aim of this study was to compare oral mucosal wetness, saliva gamma, and upper airway collapsibility during wake and sleep between women with primary Sjögren's syndrome and matched control subjects. SETTING: Participants slept in a sound-insulated room with physiologic measurements controlled from an adjacent room. PARTICIPANTS: Eleven women with primary Sjögren's syndrome and 8 age- and body mass index-matched control women. INTERVENTIONS: Upper airway collapsibility index (minimum choanal-epiglottic pressure expressed as a percentage of delivered choanal pressure) was determined from brief negative-pressure pulses delivered to the upper airway during early inspiration in wakefulness and sleep. MEASUREMENTS AND RESULTS: Patients with primary Sjögren's syndrome had significantly higher saliva gamma ("pull-off" force method) compared with control subjects (67.2 +/- 1.1 mN/m versus 63.2 +/- 1.7 mN/m, P < 0.05). Upper airway collapsibility index significantly increased from wake to sleep (Stage 2 and slow wave sleep) but was not different between groups during wake (primary Sjögren's syndrome versus controls; 36.3% +/- 8.0% vs 46.0 +/- 13.8%), stage 2 sleep (53.1% +/- 11.9% vs 63.4% +/- 7.2%), or slow-wave sleep (60.8% +/- 12.2% vs 60.5% +/- 9.3%). CONCLUSIONS: Despite having a significantly "stickier" upper airway, patients with primary Sjögren's syndrome do not appear to have abnormal upper airway collapsibility, at least as determined from upper airway collapsibility index.

Automated Bone Screw Tightening to Adaptive Levels of Stripping Torque.

OBJECTIVE: To use relationships between tightening parameters, related to bone quality, to develop an automated system that determines and controls the level of screw tightening. METHODS: An algorithm relating current at head contact (IHC) to current at construct failure (Imax) was developed. The algorithm was used to trigger cessation of screw insertion at a predefined tightening level, in real time, between head contact and maximum current. The ability of the device to stop at the predefined level was assessed. RESULTS: The mean (±SD) current at which screw insertion ceased was calculated to be [51.47 ± 9.75% × (Imax - IHC)] + IHC, with no premature bone failures. CONCLUSIONS: A smart screwdriver was developed that uses the current from the motor driving the screw to predict the current at which the screw will strip the bone threads. The device was implemented and was able to achieve motor shut-off and cease tightening at a predefined threshold, with no premature bone failures.

Protocol for a randomised crossover trial to evaluate patient and nurse satisfaction with electronic and elastomeric portable infusion pumps for the continuous administration of antibiotic therapy in the home: the Comparing Home Infusion Devices (CHID) study.

INTRODUCTION: Previous studies comparing satisfaction with electronic and elastomeric infusion pumps are limited, and improvements in size and usability of electronic pumps have since occurred. The Comparing Home Infusion Devices (CHID) study plans to assess patient and nurse satisfaction with an elastomeric and electronic pump for delivering intravenous antibiotic treatment in the home. Secondary objectives are to determine pump-related complications and actual antibiotic dose administered, evaluate temperature variation and compare pump operating costs. METHODS AND ANALYSIS: The CHID study will be a randomised, crossover trial. A trained research nurse will recruit patients with infectious disease aged ≥18 years and prescribed ≥8 days of continuous intravenous antibiotic therapy from the Royal Adelaide Hospital (RAH) (Adelaide, Australia). Patients will be randomised to receive treatment at home via an elastomeric (Baxter Infusor) or an electronic (ambIT Continuous) infusion pump for 4-7 days, followed by the other for a further 4-7 days. Patient satisfaction will be assessed by a 10-item survey to be completed at the end of each arm. Nurse satisfaction will be assessed by a single 24-item survey. Patient logbooks and case notes from clinic visits will be screened to identify complications. Pumps/infusion bags will be weighed to estimate the volume of solution delivered. Temperature sensors will record skin and ambient temperatures during storage and use of the pumps throughout the infusion period. Costs relating to pumps, consumables, antibiotics and servicing will be determined. Descriptive statistics will summarise study data. ETHICS AND DISSEMINATION: This study has been approved by the RAH Human Research Ethics Committee (HREC/16/RAH/133 R20160420, version 6.0, 5 September 2016). Study results will be disseminated through peer-reviewed publications and conference presentations. The CHID study will provide key insights into patient and provider satisfaction with elastomeric and electronic infusion pumps and inform future device selection. TRIAL REGISTRATION NUMBER: ACTRN12617000251325; Pre-results.

"Turn-of-the-Nut" Method Is Not Appropriate for Use in Cancellous Bone.

OBJECTIVE: The level to which bone screws are tightened is determined subjectively by the operating surgeon. It is likely that the tactile feedback that surgeons rely on is based on localized tissue yielding, which may predispose the screw-bone interface to failure. A limited number of studies have investigated the ratio between clinical tightening torque and stripping torque. The purpose of this study was to measure, for the first time, the ratio between yield torque (T yield) and stripping torque (T max) during screw insertion into the cancellous bone and to compare these torques with clinical levels of tightening reported in the literature. Additionally, a rotational limit was investigated as a potential end point for screw insertion in cancellous bone. METHODS: A 6.5-mm outer diameter commercial cancellous bone screw was inserted into human femoral head specimens (n = 89). Screws were inserted to failure, while recording insertion torque, compression under the screw head, and rotation angle. RESULTS: The median, interquartile ranges, and coefficient of variation were calculated for each of the following parameters: T yield, T max, T yield/T max, slope, T plateau, and rotation angle. The median ratio of T yield/T max and rotation angle was 85.45% and 96.5 degrees, respectively. The coefficient of variation was greatest for the rotation angle compared with the ratio of T yield/T max (0.37 vs. 0.12). CONCLUSIONS: The detection of yield may be a more precise method than the rotation angle in cancellous bone; however, bone-screw constructs that exhibit a T yield close to T max may be more susceptible to stripping during insertion. Future work can identify factors that influence the ratio of T yield/T max may help to reduce the incidence of screw stripping.

Pullout strength of cancellous screws in human femoral heads depends on applied insertion torque, trabecular bone microarchitecture and areal bone mineral density.

For cancellous bone screws, the respective roles of the applied insertion torque (TInsert) and of the quality of the host bone (microarchitecture, areal bone mineral density (aBMD)), in contributing to the mechanical holding strength of the bone-screw construct (FPullout), are still unclear. During orthopaedic surgery screws are tightened, typically manually, until adequate compression is attained, depending on surgeons' manual feel. This corresponds to a subjective insertion torque control, and can lead to variable levels of tightening, including screw stripping. The aim of this study, performed on cancellous screws inserted in human femoral heads, was to investigate which, among the measurements of aBMD, bone microarchitecture, and the applied TInsert, has the strongest correlation with FPullout. Forty six femoral heads were obtained, over which microarchitecture and aBMD were evaluated using micro-computed tomography and dual X-ray absorptiometry. Using an automated micro-mechanical test device, a cancellous screw was inserted in the femoral heads at TInsert set to 55% to 99% of the predicted stripping torque beyond screw head contact, after which FPullout was measured. FPullout exhibited strongest correlations with TInsert (R=0.88, p<0.001), followed by structure model index (SMI, R=-0.81, p<0.001), bone volume fraction (BV/TV, R=0.73, p<0.001) and aBMD (R=0.66, p<0.01). Combinations of TInsert with microarchitectural parameters and/or aBMD did not improve the prediction of FPullout. These results indicate that, for cancellous screws, FPullout depends most strongly on the applied TInsert, followed by microarchitecture and aBMD of the host bone. In trabecular bone, screw tightening increases the holding strength of the screw-bone construct.

Does cancellous screw insertion torque depend on bone mineral density and/or microarchitecture?

During insertion of a cancellous bone screw, the torque level reaches a plateau, at the engagement of all the screw threads prior to the screw head contact. This plateau torque (T(Plateau)) was found to be a good predictor of the insertion failure torque (stripping) and also exhibited strong positive correlations with areal bone mineral density (aBMD) in ovine bone. However, correlations between T(Plateau) and aBMD, as well as correlations between T(Plateau) and bone microarchitecture, have never been explored in human bone. The aim of this study was to determine whether T(Plateau), a predictor of insertion failure torque, depends on aBMD and/or bone microarchitecture in human femoral heads. Fifty-two excised human femoral heads were obtained. The aBMD and microarchitecture of each specimen were evaluated using dual X-ray Absorptiometry and micro-computed tomography. A cancellous screw was inserted into specimens using an automated micro-mechanical test device, and T(Plateau) was calculated from the insertion profile. T(Plateau) exhibited the strongest correlation with the structure model index (SMI, R=-0.82, p<0.001), followed by bone volume fraction (BV/TV, R=0.80, p<0.01) and aBMD (R=0.76, p<0.01). Stepwise forward regression analysis showed an increase for the prediction of T(Plateau) when aBMD was combined with microarchitectural parameters, i.e., aBMD combined with SMI (R(2) increased from 0.58 to 0.72) and aBMD combined with BV/TV and BS/TV (R(2) increased from 0.58 to 0.74). In conclusion, T(Plateau), a strong predictor for insertion failure torque, is significantly dependent on bone microarchitecture (particularly SMI and BV/TV) and aBMD.