Comparison of pulmonary congestion severity using artificial intelligence-assisted scoring versus clinical experts: A secondary analysis of BLUSHED-AHF

AIM

Acute decompensated heart failure (ADHF) is the leading cause of cardiovascular hospitalizations in the United States. Detecting B-lines through lung ultrasound (LUS) can enhance clinicians' prognostic and diagnostic capabilities. Artificial intelligence/machine learning (AI/ML)-based automated guidance systems may allow novice users to apply LUS to clinical care. We investigated whether an AI/ML automated LUS congestion score correlates with expert's interpretations of B-line quantification from an external patient dataset.

METHODS AND RESULTS

This was a secondary analysis from the BLUSHED-AHF study which investigated the effect of LUS-guided therapy on patients with ADHF. In BLUSHED-AHF, LUS was performed and B-lines were quantified by ultrasound operators. Two experts then separately quantified the number of B-lines per ultrasound video clip recorded. Here, an AI/ML-based lung congestion score (LCS) was calculated for all LUS clips from BLUSHED-AHF. Spearman correlation was computed between LCS and counts from each of the original three raters. A total of 3858 LUS clips were analysed on 130 patients. The LCS demonstrated good agreement with the two experts' B-line quantification score (r = 0.894, 0.882). Both experts' B-line quantification scores had significantly better agreement with the LCS than they did with the ultrasound operator's score (p < 0.005, p < 0.001).

CONCLUSION

Artificial intelligence/machine learning-based LCS correlated with expert-level B-line quantification. Future studies are needed to determine whether automated tools may assist novice users in LUS interpretation.

External validation of the ultrasound competency assessment tool

Objective

Point-of-care ultrasound (POCUS) is a core component of emergency medicine (EM) residency training. No standardized competency-based tool has gained widespread acceptance. The ultrasound competency assessment tool (UCAT) was recently derived and validated. We sought to externally validate the UCAT in a 3-year EM residency program.

Methods

This was a convenience sample of PGY-1 to -3 residents. Utilizing the UCAT and an entrustment scale, as described in the original study, six different evaluators split into two groups graded residents in a simulated scenario involving a patient with blunt trauma and hypotension. Residents were asked to perform and interpret a focused assessment with sonography in trauma (FAST) examination and apply the findings to the simulated scenario. Demographics, prior POCUS experience, and self-assessed competency were collected. Each resident was evaluated simultaneously by three different evaluators with advanced ultrasound training utilizing the UCAT and entrustment scales. Intraclass correlation coefficient (ICC) between evaluators was calculated for each assessment domain; analysis of variance was used to compare UCAT performance and PGY level and prior POCUS experience.

Results

Thirty-two residents (14 PGY-1, nine PGY-2, and nine PGY-3) completed the study. Overall, ICC was 0.9 for preparation, 0.57 for image acquisition, 0.3 for image optimization, and 0.46 for clinical integration. There was moderate correlation between number of FAST examinations performed and entrustment and UCAT composite scores. There was poor correlation between self-reported confidence and entrustment and UCAT composite scores.

Conclusions

We had mixed results in our attempt to externally validate the UCAT with poor correlation between faculty and moderate to good correlation with faculty to diagnostic sonographer. More work is needed to validate the UCAT before adoption.

Evaluation of Point-of-Care Ultrasound Training for Family Physicians Using Teleultrasound

BACKGROUND AND OBJECTIVES

The goal of this study was to assess family physicians' change in knowledge and ability to perform abdominal aorta ultrasound after implementation of a novel teleultrasound curriculum.

METHODS

This was a prospective, observational study conducted at a single academic institution. Family physicians completed a preassessment, test, and objective structured clinical evaluation (OSCE). Physicians then individually completed a standard curriculum consisting of online content and an hour-long, hands-on training session on abdominal aorta ultrasound using teleultrasound technology. Physicians then performed a minimum of 10 independent examinations over a period of 8 weeks. After physicians completed the training curriculum and 10 independent scans, we administered a postassessment, test, and OSCE. We analyzed differences between pre- and postcurriculum responses using Fisher exact and Wilcoxon signed rank tests.

RESULTS

Thirteen family physicians completed the curriculum. Comparing pre- to postcurriculum responses, we found significant reductions in barriers to using aorta POCUS and improved confidence in using, obtaining, and interpreting aorta POCUS (P<0.01). Knowledge improved from a median score of 70% to 90% (P<0.01), and OSCE scores improved from a median of 80% to 100% (P=0.012). Overall, 211 aorta ultrasound examinations were independently acquired with a median image quality of 4 (scale 1 to 4).

CONCLUSIONS

After an 8-week teleultrasound curriculum, family physicians with minimal experience with POCUS showed improved knowledge and psychomotor skill in abdominal aorta POCUS.

What are the minimum requirements to establish proficiency in lung ultrasound training for quantifying B-lines?

AIMS

The goal of this study was to determine the number of scans needed for novice learners to attain proficiency in B-line quantification compared with expert interpretation.

METHODS AND RESULTS

This was a prospective, multicentre observational study of novice learners, physicians and non-physicians from three academic institutions. Learners received a 2 h lung ultrasound (LUS) training session on B-line assessment, including lecture, video review to practice counting and hands-on patient scanning. Learners quantified B-lines using an eight-zone scanning protocol in patients with suspected acute heart failure. Ultrasound (US) machine settings were standardized to a depth of 18 cm and clip length of 6 s, and tissue harmonics and multibeam former were deactivated. For quantification, the intercostal space with the greatest number of B-lines within each zone was used for scoring. Each zone was given a score of 0-20 based on the maximum number of B-lines counted during one respiratory cycle. The B-line score was determined by multiplying the percentage of the intercostal space filled with B-lines by 20. We compared learner B-line counts with a blinded expert reviewer (five US fellowship-trained faculty with > 5 years of clinical experience) for each lung zone scanned; proficiency was defined as an intraclass correlation of > 0.7. Learning curves for each learner were constructed using cumulative sum method for statistical analysis. The Wilcoxon rank-sum test was used to compare the number of scans required to reach proficiency between different learner types. Twenty-nine learners (21 research associates, 5 residents and 3 non-US-trained emergency medicine faculty) scanned 2629 lung zones with acute pulmonary oedema. After a mean of 10.8 (standard deviation 14.0) LUS zones scanned, learners reached the predefined proficiency standard. The number of scanned zones required to reach proficiency was not significantly different between physicians and non-physicians (P = 0.26), learners with no prior US experience vs. > 25 prior patient scans (P = 0.64) and no prior vs. some prior LUS experience (P = 0.59). The overall intraclass correlation for agreement between learners and experts was 0.74 and 0.80 between experts.

CONCLUSIONS

Our results show that after a short, structured training, novice learners are able to achieve proficiency for quantifying B-lines on LUS after scanning 11 zones. These findings support the use of LUS for B-line quantification by non-physicians in clinical and research applications.

Design and rationale of the B-lines lung ultrasound guided emergency department management of acute heart failure (BLUSHED-AHF) pilot trial

BACKGROUND

Medical treatment for acute heart failure (AHF) has not changed substantially over the last four decades. Emergency department (ED)-based evidence for treatment is limited. Outcomes remain poor, with a 25% mortality or re-admission rate within 30days post discharge. Targeting pulmonary congestion, which can be objectively assessed using lung ultrasound (LUS), may be associated with improved outcomes.

METHODS

BLUSHED-AHF is a multicenter, randomized, pilot trial designed to test whether a strategy of care that utilizes a LUS-driven treatment protocol outperforms usual care for reducing pulmonary congestion in the ED. We will randomize 130 ED patients with AHF across five sites to, a) a structured treatment strategy guided by LUS vs. b) a structured treatment strategy guided by usual care. LUS-guided care will continue until there are ≤15 B-lines on LUS or 6h post enrollment. The primary outcome is the proportion of patients with B-lines ≤ 15 at the conclusion of 6 h of management. Patients will continue to undergo serial LUS exams during hospitalization, to better understand the time course of pulmonary congestion. Follow up will occur through 90days, exploring days-alive-and-out-of-hospital between the two arms. The study is registered on ClinicalTrials.gov (NCT03136198).

CONCLUSION

If successful, this pilot study will inform future, larger trial design on LUS driven therapy aimed at guiding treatment and improving outcomes in patients with AHF.

Systematic Sonography for Detection of Occult Wounds in Trauma

Treating victims of penetrating trauma is challenging, especially in a mass-casualty or resource-limited setting. The traditional focused assessment with sonography for trauma examination is a well-established clinical tool in the initial evaluation of trauma victims. This article describes a novel technique of systematic sonography of the abdomen to look for occult wounds, which is used in conjunction with the focused assessment with sonography for trauma examination to evaluate occult intraperitoneal injury and improve patient treatment and use of resources. The technique is highlighted in a case of a child injured in a conflict zone in northern Iraq.

Stereoscopic digital mammogram: Usefulness in daily practice

OBJECTIVES

To assess the benefit of stereoscopic digital mammography in daily practice.

METHODS

Stereoscopic digital (SD) mammography allows a fused 3D view of mammogram. A 4-degree cranio-caudal (CC) angle view matched with a regular CC view allows generating a stereoscopic view. Three breast radiologists reviewed retrospectively 1110 stereoscopic digital mammograms (1075 women) performed between November 2011 and February 2013 with the following sequence: each evaluated firstly conventional mammogram alone, and then integrated SD mammograms. The benefit was quoted in 3-grade scale: 0 for no benefit, 1 moderate and 2 excellent. The concordance between radiologists was evaluated by the W Randall coefficient. Subgroup analysis according to the BI-RADS classification, breast density and type of abnormalities were performed by calculating odds-ratio.

RESULTS

The readers had the same opinion regarding the value of stereoscopic digital mammograms in 87% of cases (962/1110). Benefit was null, moderate and excellent in 8, 52 and 26% respectively. The concordance of radiologists was excellent with a W coefficient above 0.89.

CONCLUSIONS

SD mammogram improved interpretation of abnormal mammograms. This potentially interesting and promising complementary tool might be beneficial in daily breast imaging practice.

Primary peritonitis due to Streptococcus A: laparoscopic treatment

Primary peritonitis is defined as peritoneal infection without an evident intraperitoneal septic focus. This is a rare condition and few cases are reported in the literature. We report a case of primary peritonitis in a 23-year-old female that was diagnosed and treated laparoscopically. The challenge for the surgeon is to consider the possibility of this diagnosis, and to avoid conversion to laparotomy in search of a hypothetical septic focus when none is apparent on laparoscopy.

Protein structure-based design, synthesis, and biological evaluation of 5-thia-2,6-diamino-4(3H)-oxopyrimidines: potent inhibitors of glycinamide ribonucleotide transformylase with potent cell growth inhibition

The design, synthesis, biochemical, and biological evaluation of a novel series of 5-thia-2,6-diamino-4(3H)-oxopyrimidine inhibitors of glycinamide ribonucleotide transformylase (GART) are described. The compounds were designed using the X-ray crystal structure of human GART. The monocyclic 5-thiapyrimidinones were synthesized by coupling an alkyl thiol with 5-bromo-2, 6-diamino-4(3H)-pyrimidinone, 20. The bicyclic compounds were prepared in both racemic and diastereomerically pure forms using two distinct synthetic routes. The compounds were found to have human GART KiS ranging from 30 microM to 2 nM. The compounds inhibited the growth of both L1210 and CCRF-CEM cells in culture with potencies down to the low nanomolar range and were found to be selective for the de novo purine biosynthesis pathway. The most potent inhibitors had 2,5-disubstituted thiophene rings attached to the glutamate moiety. Placement of a methyl substituent at the 4-position of the thiophene ring to give compounds 10, 18, and 19 resulted in inhibitors with significantly decreased mFBP affinity.