Ultrasound image analysis using deep neural networks for discriminating between benign and malignant ovarian tumors: comparison with expert subjective assessment

OBJECTIVES

To develop and test the performance of computerized ultrasound image analysis using deep neural networks (DNNs) in discriminating between benign and malignant ovarian tumors and to compare its diagnostic accuracy with that of subjective assessment (SA) by an ultrasound expert.

METHODS

We included 3077 (grayscale, n = 1927; power Doppler, n = 1150) ultrasound images from 758 women with ovarian tumors, who were classified prospectively by expert ultrasound examiners according to IOTA (International Ovarian Tumor Analysis) terms and definitions. Histological outcome from surgery (n = 634) or long-term (≥ 3 years) follow-up (n = 124) served as the gold standard. The dataset was split into a training set (n = 508; 314 benign and 194 malignant), a validation set (n = 100; 60 benign and 40 malignant) and a test set (n = 150; 75 benign and 75 malignant). We used transfer learning on three pre-trained DNNs: VGG16, ResNet50 and MobileNet. Each model was trained, and the outputs calibrated, using temperature scaling. An ensemble of the three models was then used to estimate the probability of malignancy based on all images from a given case. The DNN ensemble classified the tumors as benign or malignant (Ovry-Dx1 model); or as benign, inconclusive or malignant (Ovry-Dx2 model). The diagnostic performance of the DNN models, in terms of sensitivity and specificity, was compared to that of SA for classifying ovarian tumors in the test set.

RESULTS

At a sensitivity of 96.0%, Ovry-Dx1 had a specificity similar to that of SA (86.7% vs 88.0%; P = 1.0). Ovry-Dx2 had a sensitivity of 97.1% and a specificity of 93.7%, when designating 12.7% of the lesions as inconclusive. By complimenting Ovry-Dx2 with SA in inconclusive cases, the overall sensitivity (96.0%) and specificity (89.3%) were not significantly different from using SA in all cases (P = 1.0).

CONCLUSION

Ultrasound image analysis using DNNs can predict ovarian malignancy with a diagnostic accuracy comparable to that of human expert examiners, indicating that these models may have a role in the triage of women with an ovarian tumor. © 2020 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Sleep-disordered breathing in severe mental illness: clinical evaluation of oximetry diagnosis and management limitations

BACKGROUND

To describe the diagnosis and management pathway of sleep-disordered breathing (SDB) in a sample of patients with severe mental illness (SMI), and to assess the feasibility and patient acceptability of overnight oximetry as a first-step screening method for detecting severe SDB in this population.

METHODS

The study was a retrospective audit of patients with SMI seen at a Collaborative Centre for Cardiometabolic Health in Psychosis service who were invited for overnight oximetry between November 2015 and May 2018. The adjusted oxygen desaturation index (ODI) was calculated using 4% desaturation criteria. Results were discussed with a sleep specialist and categorized into a 4-level risk probability tool for SDB.

RESULTS

Of 91 adults consenting for overnight oximetry, 90 collected some oximetry data, though 11 of these 90 patients collected technically unsatisfactory oximetry. Thus 79/90 patients (88%) collected adequate oximetry data for at least one night. The oximetry traces suggested likely minimal obstructive sleep apnea (OSA) in 41 cases, moderate to severe OSA in 25 patients, severe OSA in 9 patients and possible obesity hypoventilation syndrome (OHS) in 4 cases. Full polysomnography was recommended for 39 patients but only one-third underwent testing. Nineteen patients were reviewed by a sleep specialist. Of the 10 patients who initiated CPAP, four were considered adherent to treatment.

CONCLUSION

Home oximetry may be a pragmatic option for SDB screening in patients with SMI but reliable full diagnostic and management pathways need to be developed.

Resection and Surgically Targeted Radiation Therapy for the Treatment of Larger Recurrent or Newly Diagnosed Brain Metastasis: Results From a Prospective Trial

Introduction Achieving durable local control (LC) for larger (e.g., >2-3 cm) brain metastasis whether newly diagnosed or recurrent remains problematic. Resection (R) alone is typically insufficient and adding radiation therapy (RT) still results in a 12-month recurrence rate of 20% or more in many series. Hypothesizing that R plus immediate radiation utilizing brachytherapy may improve outcomes for this cohort of patients, we designed and prospectively evaluated a permanently implanted surgically targeted radiation therapy (STaRT) device consisting of cesium-131 (Cs-131) seeds positioned within a collagen carrier (GammaTile, GT Medical Technologies, Tempe, AZ). The device was designed to prevent direct source-to-brain contact and maintain inter-source spacing after closure. Methods This was a subgroup analysis of a cohort of patients with either recurrent or previously untreated brain metastases enrolled in a prospective, multi-histology single-arm trial (ClinicalTrials.gov, NCT#03088579), conducted between February 2013 and February 2018, of resection and tumor bed brachytherapy with Cs-131 containing permanently implanted collagen tiles to deliver 60 Gray (Gy) at .5 cm depth. No additional local therapy was given without progression. Results A total of 16 metastases in 11 patients were treated; 12 tumors were recurrent and four were previously untreated. The median preoperative maximum diameter was 3.2 cm (range: 1.9-5.1 cm). Histology was seven breasts, six lungs, and three sarcomas. The median age was 60 years (range: 41-80 years); the Karnofsky Performance Status (KPS) was 70 (range: 70-90). The cohort consisted of seven females and four males. The mean time for implantation completion was five minutes. The median overall survival (OS) was 9.3 months. At a median radiographic follow-up of 9.5 months' treatment, site progression was found in 1/16 (6%) at 10.9 months, and the median treatment site time-to-progression (TTP) has not been reached [95% confidence interval (CI): >10.9 months]. At 12 months, the Kaplan-Meier (K-M) estimates for LC after R+STaRT for all tumors was 83%; for previously untreated tumors, LC at 12 months was 100% and for recurrent tumors, it was 80%. Two tumor beds (12.5%) experienced radiation brain changes: one had grade two and the other grade three. No surgical adverse events occurred. Conclusion In this single-arm precommercial study, R+STaRT demonstrated excellent safety and LC in this cohort. The device has recently received FDA clearance for use in newly diagnosed and recurrent brain metastasis, and randomized clinical trials vs. standard of care treatments in both settings are scheduled to open in 2020.

CTNI-58. A PROSPECTIVE TRIAL OF RESECTION AND SURGICALLY TARGETED RADIATION THERAPY FOR INITIAL OR SALVAGE TREATMENT OF AGGRESSIVE MENINGIOMA

INTRODUCTION

Achieving durable local control (LC) for aggressive or recurrent meningiomas remains problematic. Resection (R) alone is insufficient and even with the addition of radiation therapy (RT), outcomes are suboptimal in many series.

OBJECTIVE

Hypothesizing R plus Surgically Targeted Radiation Therapy (STaRT) may improve LC, we evaluated a permanently implanted brachytherapy device consisting of Cs-131 seeds positioned within modular resorbable collagen carrier/spacer tiles (GammaTile, GT Medical Technologies, Tempe AZ).

METHODS

From 2/2013-2/2018 recurrent and newly diagnosed aggressive meningiomas were treated on a prospective single-arm study (NCT#03088579). At completion of maximum safe resection the tumor bed was lined with collagen tiles imbedded with Cs-131, delivering 60–80 Gy at 5 mm depth. No additional local therapy was given absent progression.

RESULTS

35 meningiomas in 28 patients were treated; 29 recurrent (22 prior R+RT, 6 R only, and 1 RT only) and 6 without prior therapy. WHO grade was I in 2 patients, II in 28, and III in 5. Median age was 66 years (range 37–82), KPS 80 (70–100), female: male ratio 15:13. Mean time for implantation was 5 minutes. At a median radiographic follow-up of 25.5 months (range 0.1–71) LC was 80% (28/35) and median time to progression had not been reached (95% CI > 35.6 months). LC at 12/24/36/48 months was 100/89/72/72% for all tumors, 100/93/79/79% for Grade II, and 100/50/0/0% for Grade III, respectively. No patient receiving STaRT as their initial RT failed. Median overall survival was 50 months. Four symptomatic adverse events occurred, 2 wound breakdowns requiring surgery and 2 radiation-related brain changes, medically treated.

CONCLUSION

R+STaRT demonstrates favorable safety and LC outcomes in this single-arm prospective trial that includes heavily pre-treated patients. A commercial version of the device recently received FDA clearance for use in newly diagnosed malignant or recurrent intracranial neoplasms including meningiomas.

Economic impact of sudden cardiac arrest

Background

Each year, there are approximately 5,000 out-of-hospital cardiac arrests (OHCAs) in the state of Victoria, Australia (population 6.4 million, state healthcare budget AUD$2.9 billion / €1.8billion). Mortality from OHCA approaches ninety percent. High mortality rates and survivors not returning to work is likely to have an adverse effect on the Victorian economy but this has not been previously investigated.

Purpose

To model the economic impact of OHCA mortality and survivors not returning to work.

Methods

Data on all OHCAs transported by Ambulance Victoria from July 2017- June 2018 in Victoria, Australia was collected, including age, gender, survival to hospital, survival to discharge, and survival to 12 months. Cases were excluded if arrest was precipitated by trauma, exsanguination, overdose, terminal illness, hanging, SIDS, electrocution, sepsis, respiratory causes, drowning, or neurological causes. Pre-arrest employment status of patients was modelled using the Australian Bureau of Statistics Economic Security dataset, which provides contemporary employment rates for gender-matched five-year cohorts for Australians aged 15–79 years. For survivors to 12 months, pre-arrest and post-arrest work status were confirmed. Economic impact was then calculated to a five year horizon utilizing a Markov model with probabilistic sensitivity analysis.

Results

4,934 arrests meeting the inclusion criteria were transported by Ambulance Victoria in twelve months, of whom 4,639 were determined to be cardiac arrests without any exclusion criteria as a precipitant. 695 patients survived to hospital (15.0%), and 325 to discharge (7.0%). At 12 months, 303 patients were alive (6.5% of overall cases, 93.2% of those discharged from hospital). Economic modelling of age and gender-matched data indicated that 1516 patients (35%) would have been employed pre-cardiac arrest, but only 216 survivors (4.7%) would be employed at five years post-arrest. Using Markov modelling incorporating estimated earnings and the pre-determined value of a statistical life, the annual economic burden of cardiac arrest approximated AUD$4 billion (€2.5 billion) at a five-year horizon.

Conclusion

The annual economic impact of cardiac arrest in Victoria, Australia is approximately AUD$4 billion (€2.5 billion) in a five-year horizon. As the annual Victorian state budget for all healthcare is AUD$2.93 billion (€1.8 billion), our data suggests that the economic impact of cardiac arrest is under-appreciated. Therefore, research in this area and providing state-of-the-art care for all cardiac arrest patients should be a healthcare priority.

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): NHMRC/NHF Postgraduate Scholarship, RACP JJ Billings Scholarship

Metabolomics and a breath sensor identify acetone as a biomarker for heart failure

Introduction

Metabolomics delivers more biological and clinical insight than targeted investigations. We applied metabolomics to patients with heart failure (HF) with reduced ejection fraction (HFrEF).

Methods

46 patients with HFrEF and 20 controls underwent metabolomic profiling, including liquid/gas chromatography mass spectrometry (LC-MS/GC-MS), nuclear magnetic resonance (NMR) metabolomics and solid-phase microextraction (SPME) volatilomics in plasma and urine. HFrEF was defined using left ventricular global longitudal strain (GLS) <18%, ejection fraction <50% and NTproBNP ≥35 pmol/L. A low cost consumer breath acetone (BrACE) sensor validated SPME results in 69 patients.

Results

34 metabolites were identified by GCMS, 33 by LCMS and 2 volatiles by SPME (acetone, 2-pentanone in plasma and urine). Alanine, aspartate and glutamate, citric acid cycle, arginine biosynthesis, glyoxylate and dicarboxylate metabolism were altered in HFrEF. Plasma acetone correlated with NT-proBNP (r=0.59, 95% CI 0.4 to 0.7), triacylglycerol (55:9), 2-oxovaleric and cis-aconitic acid, involved with ketone metabolism and mitochondrial energetics. BrACE >1.5 ppm discriminated HF from other cardiac pathology (AUC 0.88, 95% CI 0.77 to 0.99, P<0.0001).

Conclusion

Breath acetone was detectable in HFrEF patients using a consumer sensor ($1/test) and although not cardiac specific, discriminated HF from other cardiac pathology.

Funding Acknowledgement

Type of funding source: Public grant(s) – National budget only. Main funding source(s): Health Research Council of New Zealand

Diagnosis and treatment of a caecal mucocoele in a dog

An 11-year-old male intact Staffordshire Bull terrier was referred for diabetic ketoacidosis. Abdominal ultrasonographic examination revealed a 5 cm × 2 cm intraluminal caecal mass-like structure. Exploratory laparotomy and typhlectomy were subsequently performed. Histopathology of the caecal mass-like structure was consistent with a caecal mucocoele, defined as a cystic dilation of the caecal lumen with stasis of mucus. This lesion has been previously described in humans, where it is termed an appendiceal mucocoele. The patient was euthanased 58 days post-operatively due to unrelated diabetic complications.

Collateral damage: Hidden impact of the COVID-19 pandemic on the out-of-hospital cardiac arrest system-of-care

AIM

Out-of-hospital cardiac arrest (OHCA) during COVID-19 has been reported by countries with high case numbers and overwhelmed healthcare services. Imposed restrictions and treatment precautions may have also influenced OHCA processes-of-care. We investigated the impact of the COVID-19 pandemic period on incidence, characteristics, and survival from OHCA in Victoria, Australia.

METHODS

Using data from the Victorian Ambulance Cardiac Arrest Registry, we compared 380 adult OHCA patients who received resuscitation between 16th March 2020 and 12th May 2020, with 1218 cases occurring during the same dates in 2017-2019. No OHCA patients were COVID-19 positive. Arrest incidence, characteristics and survival rates were compared. Regression analysis was performed to understand the independent effect of the pandemic period on survival.

RESULTS

Incidence of OHCA did not differ during the pandemic period. However, initiation of resuscitation by Emergency Medical Services (EMS) significantly decreased (46.9% versus 40.6%, p = 0.001). Arrests in public locations decreased in the pandemic period (20.8% versus 10.0%; p < 0.001), as did initial shocks by public access defibrillation/first-responders (p = 0.037). EMS caseload decreased during the pandemic period, however, delays to key interventions (time-to-first defibrillation, time-to-first epinephrine) significantly increased. Survival-to-discharge decreased by 50% during the pandemic period (11.7% versus 6.1%; p = 0.002). Survivors per million person-years dropped in 2020, resulting in 35 excess deaths per million person-years. On adjusted analysis, the pandemic period remained associated with a 50% reduction in survival-to-discharge.

CONCLUSION

The COVID-19 pandemic period did not influence OHCA incidence but appears to have disrupted the system-of-care in Australia. However, this could not completely explain reductions in survival.

New ^{13}C(α,n)^{16}O Cross Section with Implications for Neutrino Mixing and Geoneutrino Measurements

Precise antineutrino measurements are very sensitive to proper background characterization. We present an improved measurement of the ^{13}C(α,n)^{16}O reaction cross section which constitutes significant background for large ν[over ¯] detectors. We greatly improve the precision and accuracy by utilizing a setup that is sensitive to the neutron energies while making measurements of the excited state transitions via secondary γ-ray detection. Our results shows a 54% reduction in the background contributions from the ^{16}O(3^{-},6.13  MeV) state used in the KamLAND analysis.

36. A PROSPECTIVE TRIAL OF RESECTION PLUS SURGICALLY TARGETED RADIATION THERAPY FOR BRAIN METASTASIS

INTRODUCTION

Achieving durable local control for larger brain metastases remains problematic. Resection (R) alone is typically insufficient. Even with the addition of stereotactic radiation the 12-month recurrence rate for larger lesions (i.e., &gt;2.5–3 cm) is 20% or more in many series. To improve outcomes we designed and prospectively evaluated a permanently implanted radiation device consisting of Cs-131 seeds positioned within a collagen tile (GammaTile, GT Medical Technologies, Tempe AZ). We combined maximum safe resection and collagen tile brachytherapy (CTBT) with the hypothesis that immediate radiation initiation and/or dose intensification could improve outcomes.

MATERIALS/METHODS

From 2013–2018 patients undergoing resection with either previously untreated or recurrent brain metastasis were enrolled on a single arm, multi-histology study (ClinicalTrials.gov, NCT#03088579). At resection completion the tumor bed was lined with collagen tiles imbedded with Cs-131, delivering 60–80 Gy at 5 mm depth. The device was designed to prevent direct source-to-brain contact and to maintain inter-source spacing after closure. No additional local therapy was given unless progression occurred.

RESULTS

16 metastases (12 recurrent/4 previously untreated) in 11 patients were treated. Median diameter 3.1 cm, range 1.9–5.1. Histology was 7 breast, 6 lung, and 3 sarcoma. Median age 60 years; 7 females/4 males. Average time for implantation was 5 minutes. At median radiographic follow-up of 9.5 months (range 0.1–25.2) treatment site progression occurred 1/16 (6%) at 10.9 months. Median treatment site time-to-progression (TTP) has not been reached (95% CI, &gt;10.9 months). Median overall survival (OS) 9.3 months. No surgical adverse events occurred. One patient (6.2%) experienced radiation brain changes and was treated medically.

CONCLUSION

R+CTBT demonstrated excellent safety and local control outcomes in this single-arm pre-commercial study. The device recently received FDA clearance for use in newly diagnosed and recurrent brain metastasis. Randomized clinical trials vs standard of care treatments are expected to open in 2020.

High-Brightness Continuous-Wave Electron Beams from Superconducting Radio-Frequency Photoemission Gun

Continuous-wave photoinjectors operating at high accelerating gradients promise to revolutionize many areas of science and applications. They can establish the basis for a new generation of monochromatic x-ray free electron lasers, high-brightness hadron beams, or a new generation of microchip production. In this Letter we report on the record-performing superconducting rf electron gun with CsK_{2}Sb photocathode. The gun is generating high charge electron bunches (up to 10  nC/bunch) and low transverse emittances, while operating for months with a single photocathode. This achievement opens a new era in generating high-power beams with a very high average brightness.

AB1174 TRACKING THE EFFECTS ON A CLINICAL SERVICE OF INTRODUCING ULTRASOUND FOR DIAGNOSIS OF GIANT CELL ARTERITIS: DESIGN OF A SERVICE EVALUATION USING LEAN METHODOLOGY

Background:

In our large, multi-site hospital, patients with suspected GCA are started promptly on high-dose prednisolone but until 2019, patients waited for temporal artery biopsy (TAB) until the GCA diagnosis could be confirmed (“GCA”) or refuted (“not-GCA”). Reports of the impact of introducing temporal and axillary artery ultrasound (TAUS) have mainly come from smaller hospitals. Agreement between TAUS and TAB has been reported by others with a Cohen’s kappa of 0.35 [1] and 0.40 [2]. We used Lean methodology to identify metrics across 5 key domains: delivery, quality, service, morale and cost.

Objectives:

To design metrics for a service evaluation to measure impact of introducing TAUS, and to test their feasibility of measurement within routine care.

Methods:

Our primary driver was time from presenting to our service to diagnostic confirmation (lead time). Pathway mapping, value stream mapping and a driver diagram identified key ideas for improvement.

We chose to measure: Delivery (mean lead time for each month), Quality (proportion of patients with GCA and positive TAB/TAUS; total (cumulative) prednisolone dose in patients with not-GCA, Service (patient feedback), Morale (staff feedback) and Cost (number of patients; cost of tests per patient; overall costs). We plotted these by month on run charts and defined a significant shift as 6 consecutive monthly values below baseline median. Cohen’s kappa was calculated using GraphPad QuickCalcs.

Results:

Routine TAUS for suspected GCA was introduced from January 2019, alongside a multidisciplinary team monthly meeting. TAUS was done a median of 2.5 days from referral. Agreement between TAB and TAUS results was good (Table 1). The run chart showed a significant shift in our Delivery (median lead time fell from 28.7 days to 21 days after introduction of ultrasound) and both Quality metrics (proportion of GCA with positive TAB/TAUS increased from 29% to 69%; total prednisolone dose for not-GCA fell from 1.335g to 0.846g).

Table 1.

Concordance between temporal and axillary artery ultrasound (TAUS) and temporal artery biopsy (TAB) in scans performed through 2019. Cohen’s weighted kappa 0.59 (including equivocal results as separate category).

TAUS positiveTAUS negativeTAUS equivocalTAB positive1411TAB negative5275TAB equivocal030

Within Costs, average per-patient costs of TAB/TAUS declined from £1004/patient to £792/patient, but total referrals for TAB/TAUS increased from 6/month to 10/month, increasing overall costs. Staff and patient feedback (Service, Morale) revealed that further improvements to the care pathway were needed to manage the additional complexity.

Conclusion:

Lean methodology identified multiple metrics for evaluating the impact of TAUS on our service. Introducing TAUS improved Delivery and Quality, but measuring Costs, Morale and Service helped identify unintended consequences. Concordance between TAUS and TAB was good. We plan to continue to improve and monitor the care pathway based on our multi-stakeholder feedback.

References:

[1]Luqmani et al., HTA 2016

[2]Mukhtyar et al., Clin Rheum 2019

Disclosure of Interests:

Sarah Mackie Grant/research support from: Roche (attendance of EULAR 2019; co-applicant on research grant), Consultant of: Sanofi, Roche/Chugai (monies paid to my institution not to me), Andrew Barr: None declared, Alison Cracknell: None declared, Shannon Farrell: None declared, Jimmy Parvin: None declared, Ajay Patil: None declared, Ian Simmons: None declared, Kate Smith Grant/research support from: Medical education grants from Sanofi and Biogen, Speakers bureau: Novartis, Andrea Sweeting: None declared, Max Troxler: None declared, Tara Webster: None declared, Richard Wakefield Speakers bureau: Novartis, Janssen, GE

Impacts of rat hindlimb Fndc5/irisin overexpression on muscle and adipose tissue metabolism

Myokines, such as irisin, have been purported to exert physiological effects on skeletal muscle in an autocrine/paracrine fashion. In this study, we aimed to investigate the mechanistic role of in vivo fibronectin type III domain-containing 5 (Fndc5)/irisin upregulation in muscle. Overexpression (OE) of Fndc5 in rat hindlimb muscle was achieved by in vivo electrotransfer, i.e., bilateral injections of Fndc5 harboring vectors for OE rats (n = 8) and empty vector for control rats (n = 8). Seven days later, a bolus of D₂O (7.2 mL/kg) was administered via oral gavage to quantify muscle protein synthesis. After an overnight fast, on day 9, 2-deoxy-d-glucose-6-phosphate (2-DG6P; 6 mg/kg) was provided during an intraperitoneal glucose tolerance test (2 g/kg) to assess glucose handling. Animals were euthanized, musculus tibialis cranialis muscles and subcutaneous fat (inguinal) were harvested, and metabolic and molecular effects were evaluated. Muscle Fndc5 mRNA increased with OE (~2-fold; P = 0.014), leading to increased circulating irisin (1.5 ± 0.9 to 3.5 ± 1.2 ng/mL; P = 0.049). OE had no effect on protein anabolism or mitochondrial biogenesis; however, muscle glycogen was increased, along with glycogen synthase 1 gene expression (P = 0.04 and 0.02, respectively). In addition to an increase in glycogen synthase activation in OE (P = 0.03), there was a tendency toward increased glucose transporter 4 protein (P = 0.09). However, glucose uptake (accumulation of 2-DG6P) was identical. Irisin elicited no endocrine effect on mitochondrial biogenesis or uncoupling proteins in white adipose tissue. Hindlimb overexpression led to physiological increases in Fndc5/irisin. However, our data indicate limited short-term impacts of irisin in relation to muscle anabolism, mitochondrial biogenesis, glucose uptake, or adipose remodeling.