Performance of an artificial intelligence tool with real-time clinical workflow integration – Detection of intracranial hemorrhage and pulmonary embolism

Buls, Nico; Watté, Nina; Nieboer, Koenraad; Ilsen, Bart; Mey, Johan De

doi:10.1016/j.ejmp.2021.03.015

Cited by 20 publications

(12 citation statements)

References 29 publications

(43 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our best model also had better specificity with almost similar sensitivity than a multimodal fusion model combining information from CT images and electronic health records, which had a specificity of 90.2% and a sensitivity of 87.3% [ 37 ]. We achieved only slightly worse specificity with our best model than Aidoc’s commercial model (93.5% vs. 95–95.5%, respectively), despite their model having been trained with a considerably larger dataset (600 vs. ~ 28,000 CTPAs) [ 24 , 38 ]. Models presented by Yang et al and Tajbaksh et al aimed to localize each distinct embolus accurately, and their performance was evaluated differently than our model, which is why the models cannot be directly compared [ 21 , 39 ].…”

Section: Discussionmentioning

confidence: 99%

“…The model achieved sensitivity and specificity of 92.7% and 95.5%, respectively [ 23 ]. Buls et al also tested the model (version 1.3) and achieved a similar specificity of 95% but a lower sensitivity of 73% [ 24 ]. However, creating an annotated training set of this magnitude for own model development may not be feasible for single hospitals or research teams.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automated detection of pulmonary embolism from CT-angiograms using deep learning

Huhtanen

Nyman

Mohsen³

et al. 2022

BMC Med Imaging

View full text Add to dashboard Cite

Background The aim of this study was to develop and evaluate a deep neural network model in the automated detection of pulmonary embolism (PE) from computed tomography pulmonary angiograms (CTPAs) using only weakly labelled training data. Methods We developed a deep neural network model consisting of two parts: a convolutional neural network architecture called InceptionResNet V2 and a long-short term memory network to process whole CTPA stacks as sequences of slices. Two versions of the model were created using either chest X-rays (Model A) or natural images (Model B) as pre-training data. We retrospectively collected 600 CTPAs to use in training and validation and 200 CTPAs to use in testing. CTPAs were annotated only with binary labels on both stack- and slice-based levels. Performance of the models was evaluated with ROC and precision–recall curves, specificity, sensitivity, accuracy, as well as positive and negative predictive values. Results Both models performed well on both stack- and slice-based levels. On the stack-based level, Model A reached specificity and sensitivity of 93.5% and 86.6%, respectively, outperforming Model B slightly (specificity 90.7% and sensitivity 83.5%). However, the difference between their ROC AUC scores was not statistically significant (0.94 vs 0.91, p = 0.07). Conclusions We show that a deep learning model trained with a relatively small, weakly annotated dataset can achieve excellent performance results in detecting PE from CTPAs.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Automated detection of pulmonary embolism from CT-angiograms using deep learning

Huhtanen

Nyman

Mohsen³

et al. 2022

BMC Med Imaging

View full text Add to dashboard Cite

show abstract

“…Machine learning can effectively detect intracranial hemorrhage on CT and can reduce the false negative rate for the detection of intracranial hemorrhage [6][7][8][9][10][11]. Consequently, artificial intelligence software has begun to be incorporated into the clinical workflow.…”

Section: Discussionmentioning

confidence: 99%

Implementation of Machine Learning Software on the Radiology Worklist Decreases Scan View Delay for the Detection of Intracranial Hemorrhage on CT

Ginat

2021

Brain Sciences

View full text Add to dashboard Cite

Background and Purpose: Prompt identification of acute intracranial hemorrhage on CT is important. The goal of this study was to assess the impact of artificial intelligence software for prioritizing positive cases. Materials and Methods: Cases analyzed by Aidoc (Tel Aviv, Israel) software for triaging acute intracranial hemorrhage cases on non-contrast head CT were retrospectively reviewed. The scan view delay time was calculated as the difference between the time the study was completed on PACS and the time the study was first opened by a radiologist. The scan view delay was stratified by scan location, including emergency, inpatient, and outpatient. The scan view delay times for cases flagged as positive by the software were compared to those that were not flagged. Results: A total of 8723 scans were assessed by the software, including 6894 cases that were not flagged and 1829 cases that were flagged as positive. Although there was no statistically significant difference in the scan view time for emergency cases, there was a significantly lower scan view time for positive outpatient and inpatient cases flagged by the software versus negative cases, with a reduction of 604 min on average, 90% in the scan view delay (p-value < 0.0001) for outpatients, and a reduction of 38 min on average, and 10% in the scan view delay (p-value <= 0.01) for inpatients. Conclusion: The use of artificial intelligence triage software for acute intracranial hemorrhage on head CT scans is associated with a significantly shorter scan view delay for cases flagged as positive than cases not flagged among outpatients and inpatients at an academic medical center.

show abstract

“…In addition, no blood draw is required, and ionizing radiation is not used as it is in x-ray or CT. 18,24,27,32,33 Computer-aided diagnosis has become a component of clinical work at many hospitals for a variety of applications, such as breast cancer screening and identification of pulmonary embolism or stroke. 34,35 DL has shown human-level performances in many imaging-based tasks, such as classification, segmentation, and recognition. 36,37 These studies also show that DL has the potential for various automated ultrasound (US) image analysis tasks.…”

Section: Related Workmentioning

confidence: 99%

Classification of clinically relevant intravascular volume status using point of care ultrasound and machine learning

Wshah

Xu²,

Steinharter

et al. 2022

J. Med. Imag.

View full text Add to dashboard Cite

Purpose: This is a foundational study in which multiorgan system point of care ultrasound (POCUS) and machine learning (ML) are used to mimic physician management decisions regarding the functional intravascular volume status (IVS) and need for diuretic therapy. We present this as an impactful use case of an application of ML in aided decision making for clinical practice. IVS represents complex physiologic interactions of the cardiac, renal, pulmonary, and other organ systems. In particular, we focus on vascular congestion and overload as an evolving concept in POCUS diagnosis and clinical relevance. It is critical for physicians to be able to evaluate IVS without disrupting workflow or exposing patients to unnecessary testing, radiation, or cost. This work utilized a small retrospective dataset as a feasibility test for ML binary classification of diuretic administration validated with clinical decision data. Future work will be directed toward artificial intelligence (AI) delivery at the bedside and assessment of the impact on patient-centered outcomes and physician workflow improvement.Approach: We retrospectively reviewed and processed 1039 POCUS video clips, including cardiac, thoracic, and inferior vena cava (IVC) views. Multiorgan POCUS clips were correlated with clinical data extracted from the electronic health record and deidentified for algorithm training and validation. We implemented a two-stream three-dimensional (3D) deep learning approach that fuses heart and IVC data to perform binary classification of the need for diuretic use.Results: Our proposed approach achieves high classification accuracy (84%) for the determination of diuretic use with 0.84 area under the receiver operating characteristic curve.Conclusions: Our two-stream 3D deep neural network is able to classify POCUS video clips that match physicians' classification for or against diuretic use with high accuracy. This serves as a foundational step in the progress toward AI-aided diagnosis and AI implementation in the field of IVS evaluation by POCUS.

show abstract

Performance of an artificial intelligence tool with real-time clinical workflow integration – Detection of intracranial hemorrhage and pulmonary embolism

Cited by 20 publications

References 29 publications

Automated detection of pulmonary embolism from CT-angiograms using deep learning

Automated detection of pulmonary embolism from CT-angiograms using deep learning

Implementation of Machine Learning Software on the Radiology Worklist Decreases Scan View Delay for the Detection of Intracranial Hemorrhage on CT

Classification of clinically relevant intravascular volume status using point of care ultrasound and machine learning

Contact Info

Product

Resources

About