PENet - A Scalable Deep-Learning Model for Automated Diagnosis of Pulmonary Embolism Using Volumetric CT Imaging

Huang, Shih-Cheng; Kothari, Tanay; Banerjee, Imon; Chute, Chris; Ball, Robyn L.; Borus, Norah; Huang, Aihua; Patel, Bhavik N.; Rajpurkar, Pranav; Irvin, Jeremy; Dunnmon, Jared; Ng, Andrew Y.; Lungren, Matthew P.

doi:10.2139/ssrn.3384889

Cited by 20 publications

(31 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The sample size in this manuscript is 167.145 annotated ultrasound imaging frames for model training from 255 healthy volunteers, 15.523 annotated frames from 26 participants for internal model validation and 17.855 frames in video recordings from 51 patients with suspected DVT for external, prospective model evaluation. This is in line with other studies evaluating algorithmic diagnostic decision support, who recently reported external validation sample sizes of, e.g., 50 32,33 , 91 34 , and 198 35 subjects during retrospective testing and 80 36 to 97 37 during prospective testing.…”

Section: Methodssupporting

confidence: 91%

See 1 more Smart Citation

Non-invasive Diagnosis of Deep Vein Thrombosis from Ultrasound with Machine Learning

Kainz

Makropoulos

Oppenheimer³

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep Vein Thrombosis (DVT) is a blood clot most found in the leg, which can lead to fatal pulmonary embolism (PE). Compression ultrasound of the legs is the diagnostic gold standard, leading to a definitive diagnosis. However, many patients with possible symptoms are not found to have a DVT, resulting in long referral waiting times for patients and a large clinical burden for specialists. Thus, diagnosis at the point of care by non-specialists is desired.We collect images in a pre-clinical study and investigate a deep learning approach for the automatic interpretation of compression ultrasound images. Our method provides guidance for free-hand ultrasound and aids non-specialists in detecting DVT.We train a deep learning algorithm on ultrasound videos from 246 healthy volunteers and evaluate on a sample size of 51 prospectively enrolled patients from an NHS DVT diagnostic clinic. 32 DVT-positive patients and 19 DVT-negative patients were included. Algorithmic DVT diagnosis results in a sensitivity of 93.8% and a specificity of 84.2%, a positive predictive value of 90.9%, and a negative predictive value of 88.9% compared to the clinical gold standard.To assess the potential benefits of this technology in healthcare we evaluate the entire clinical DVT decision algorithm and provide cost analysis when integrating our approach into a diagnostic pathway for DVT. Our approach is estimated to be cost effective at up to $150 per software examination, assuming a willingness to pay $26 000/QALY.

show abstract

Section: Methodssupporting

confidence: 91%

“…Two studies reported sensitivity of 84.4 -90.0% and specificity of 97.0 -97.1% when intensely trained nurses and GPs were the ultrasound operators 9,10 . 34 , and 198 35 subjects during retrospective testing and 80 36 to 97 37 during prospective testing.…”

Section: Discussionmentioning

confidence: 99%

Non-invasive Diagnosis of Deep Vein Thrombosis from Ultrasound with Machine Learning

Kainz

Makropoulos

Oppenheimer³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To optimize the utility of Emboleye for clinical purposes, we established an extremely high decision confidence threshold, which sought to maximize specificity. To this aim, we believe that we have achieved start-of-art performance at the 0.99 specificity point compared to previous recent work on PE detection, which previously achieved a maximum sensitivity of 0.32 at this point (27,28) . For a pathology with a relatively low incidence (0.07 on angiography exams and 0.01 on non-angiography), even a modestly high specificity would result in a large number of false positives.…”

Section: Discussionmentioning

confidence: 71%

“…Interest in artificial intelligence (AI) and deep learning has proliferated in medicine and been shown to equal or exceed physician performance on several tasks (21)(22)(23) . AI algorithms have been developed for a variety of specific tasks, including detecting intracranial hemorrhage on head CT exams (24) , acute abdominal findings on CT exams (25) , critical and urgent findings on chest radiographs (26) ), and pulmonary embolism (27,28) . However, the training and validation process has typically been limited to data sourced from a single or small group of hospitals.…”

Section: Introductionmentioning

confidence: 99%

Development and Validation of a Highly Generalizable Deep Learning Pulmonary Embolism Detection Algorithm

Schmid

Johnson

Ngo

et al. 2020

Preprint

View full text Add to dashboard Cite

Several algorithms have been developed for the detection of pulmonary embolism, though generalizability and bias remain potential weaknesses due to small sample size and sample homogeneity. We developed and validated a highly generalizable deep-learning algorithm, Emboleye, for the detection of PE by using a large and diverse dataset, which included 30,574 computed tomography (CT) exams sourced from over 2,000 hospital sites. On angiography exams, Emboleye demonstrates an AUROC of 0.79 with a specificity of 0.99 while maintaining a sensitivity of 0.37 and PPV of 0.77. On non-angiography CT exams, Emboleye demonstrates an AUROC of 0.77 with a specificity of 0.99 while maintaining a sensitivity of 0.18 and PPV of 0.35.

show abstract

“…These improvements support us getting better extracted features with our transfer learning method. In addition, we also test our transfer learning method on 3D-ResNet18 and PENet ( 54 ). Table 2 shows that our transfer learning method can improve ACC by 1.8 and 0.7%.…”

Section: Experiments and Resultsmentioning

confidence: 99%

Diagnosis of COVID-19 Pneumonia Based on Graph Convolutional Network

et al. 2021

View full text Add to dashboard Cite

A three-dimensional (3D) deep learning method is proposed, which enables the rapid diagnosis of coronavirus disease 2019 (COVID-19) and thus significantly reduces the burden on radiologists and physicians. Inspired by the fact that the current chest computed tomography (CT) datasets are diversified in equipment types, we propose a COVID-19 graph in a graph convolutional network (GCN) to incorporate multiple datasets that differentiate the COVID-19 infected cases from normal controls. Specifically, we first apply a 3D convolutional neural network (3D-CNN) to extract image features from the initial 3D-CT images. In this part, a transfer learning method is proposed to improve the performance, which uses the task of predicting equipment type to initialize the parameters of the 3D-CNN structure. Second, we design a COVID-19 graph in GCN based on the extracted features. The graph divides all samples into several clusters, and samples with the same equipment type compose a cluster. Then we establish edge connections between samples in the same cluster. To compute accurate edge weights, we propose to combine the correlation distance of the extracted features and the score differences of subjects from the 3D-CNN structure. Lastly, by inputting the COVID-19 graph into GCN, we obtain the final diagnosis results. In experiments, the dataset contains 399 COVID-19 infected cases, and 400 normal controls from six equipment types. Experimental results show that the accuracy, sensitivity, and specificity of our method reach 98.5%, 99.9%, and 97%, respectively.

show abstract

PENet - A Scalable Deep-Learning Model for Automated Diagnosis of Pulmonary Embolism Using Volumetric CT Imaging

Cited by 20 publications

References 0 publications

Non-invasive Diagnosis of Deep Vein Thrombosis from Ultrasound with Machine Learning

Non-invasive Diagnosis of Deep Vein Thrombosis from Ultrasound with Machine Learning

Development and Validation of a Highly Generalizable Deep Learning Pulmonary Embolism Detection Algorithm

Diagnosis of COVID-19 Pneumonia Based on Graph Convolutional Network

Contact Info

Product

Resources

About