Suki
May 1, 2022
Article of the Month – May 2022
May 1, 2022Aidoc
Overview
Aidoc is a company that provides artificial intelligence (AI) tools to assist radiologists in identifying diseases. Aidoc’s technology assists radiologists by increasing quality and efficiency, reducing turnaround time, and marking alterations in real-time. Radiologists could benefit from a machine learning technology continuously “learning” to decrease workload, allowing the channeling of efforts to diagnose disease.
Aidoc’s AI Aspect
Aidoc is an investment in technology that allows for ongoing improvements to diagnosis and patient care by supporting radiologists as they battle increasingly weighty workloads.
Aidoc’s products offer a broad set of options, allowing the diagnosis of an extensive range of diseases. The companies’ development team has a combined experience working with AI that allows them to contribute to the healthcare industry uniquely.
Coming from a background as officers in the Israeli Defense Forces intelligence unit, Aidoc’s founders offer their unique points of view, including computation, algorithmic, research, and medical capabilities, applying their knowledge for the development of AI systems with improving response and efficiency.
Combined with Aidoc’s three U.S-certified on-staff physicians, their technological analysis creates a team that leads the industry from a medical and technological perspective.
The prioritization tool allows radiologists to detect urgent cases faster and reduce overall report turnaround time directly in the workflow. This integrated workflow consists of a system that flags anomalies in real-time to assist radiologists in triage.(1)
Aidoc is an investment in technology that allows for ongoing improvements to diagnosis and patient care by supporting radiologists as they battle increasingly weighty workloads.
Aidoc’s products offer a broad set of options, allowing the diagnosis of an extensive range of diseases. The companies’ development team has a combined experience working with AI that allows them to contribute to the healthcare industry uniquely.
Coming from a background as officers in the Israeli Defense Forces intelligence unit, Aidoc’s founders offer their unique points of view, including computation, algorithmic, research, and medical capabilities, applying their knowledge for the development of AI systems with improving response and efficiency.
Combined with Aidoc’s three U.S-certified on-staff physicians, their technological analysis creates a team that leads the industry from a medical and technological perspective.
The prioritization tool allows radiologists to detect urgent cases faster and reduce overall report turnaround time directly in the workflow. This integrated workflow consists of a system that flags anomalies in real-time to assist radiologists in triage.(1)
Aidoc’s IT Aspects
Aidoc develops advanced healthcare-grade AI-based management support software. The system records and analyzes the images, identifies the anomalies, and recommends possible interventions. It supports healthcare providers in prioritizing life-threatening cases and improves patient care.
Aidoc’s solution analyzes medical images directly after the patient is scanned and notifies the radiologist when suspicious findings are detected. As a result, Aidoc magnifies the impact of the radiologists’ diagnostic capacity helping them accelerate patient management and improve outcomes. Aidoc’s healthcare-grade deep learning algorithms benefit from large quantities of data, making their solutions very comprehensive and enabling them to provide diagnostic aid to a broad set of pathologies.
FDA Cleared and CE marked:
- Intracranial hemorrhage: Non-enhanced head CT images are analyzed, then the algorithm flags and suggests possible findings.
- C-spine Fractures: Performs analysis of cervical spine CT images and communicates suspected positive findings of linear lucencies in the cervical spine bone in patterns compatible with fractures.
Aidoc develops advanced healthcare-grade AI-based management support software. The system records and analyzes the images, identifies the anomalies, and recommends possible interventions. It supports healthcare providers in prioritizing life-threatening cases and improves patient care.
Aidoc’s solution analyzes medical images directly after the patient is scanned and notifies the radiologist when suspicious findings are detected. As a result, Aidoc magnifies the impact of the radiologists’ diagnostic capacity helping them accelerate patient management and improve outcomes. Aidoc’s healthcare-grade deep learning algorithms benefit from large quantities of data, making their solutions very comprehensive and enabling them to provide diagnostic aid to a broad set of pathologies.
FDA Cleared and CE marked:
- Intracranial hemorrhage: Non-enhanced head CT images are analyzed, then the algorithm flags and suggests possible findings.
- C-spine Fractures: Performs analysis of cervical spine CT images and communicates suspected positive findings of linear lucencies in the cervical spine bone in patterns compatible with fractures.
- Intra-abdominal free gas: the software provides the analysis of abdomen CT images, flags, and communicates suspected cases of intra-abdominal free gas.
- Large Vessel Occlusions: Analyzes head CT angiography images and communicates suspected positive findings of large vessel occlusions.
- Pulmonary Embolism: Analysis of CTPA images.
- Incidental PE: Analysis of CT images (not the dedicated CTPA protocol), flags, and communication of incidental pulmonary embolism on GE and Siemens scanners.
By the numbers, Aidoc’s software has analyzed 3,955,688 scans and has saved 5,532,885 minutes in turnaround time. These numbers are possible through its prioritization tool and workflow integration that helps the radiologist in workflow triage. (2)
Aidoc’s Financial Aspects
Since 2019, Aidoc has secured more than $20 million for its Series B funding led by Square Peg Capital. The new funds raise the Series B funding to $47 million and give the company a total of $60 million for research and new projects.
The new funds will increase the adoption of technologies by physicians and the addition of new tools.
With Aidoc supporting more than 400 healthcare centers worldwide, its CEO expects to increase that number to more than 500 by 2020. Also, the company announced it expects to have 10 FDA clearances by next year.(3)
Utility of AI as a Prospective Radiology Peer Reviewer – Detection of Unreported Intracranial Hemorrhage
A missed intracranial hemorrhage (ICH) could result in decreased patient survival, worse outcomes, and possible medicolegal cases. Also, the higher workload of radiologists can increase the chances of error and affect the quality of care provided. Researchers hypothesized that using an AI algorithm in association with the conventional radiologist interpretation can lower the prevalence of false-negative interpretation. A total of 6565 non-contrast CT scans were included in the study. Of those, 5585 scans were reported to be negative for ICH. The AI solution was used on these images, of which the system suggested 28 were ICH.
After evaluation by three neuroradiologists, 16 of the scans were found to have ICH. The false rate of radiologists for ICH detection was 1.6%. The parafalcine regions and the cerebral convexity were the most missed areas for ICH.
The study suggests that an AI solution could be applied as an adjunct to current peer review tools of non-contrast CT scans determined to be negative for ICH. (4)
A missed intracranial hemorrhage (ICH) could result in decreased patient survival, worse outcomes, and possible medicolegal cases. Also, the higher workload of radiologists can increase the chances of error and affect the quality of care provided. Researchers hypothesized that using an AI algorithm in association with the conventional radiologist interpretation can lower the prevalence of false-negative interpretation. A total of 6565 non-contrast CT scans were included in the study. Of those, 5585 scans were reported to be negative for ICH. The AI solution was used on these images, of which the system suggested 28 were ICH.
After evaluation by three neuroradiologists, 16 of the scans were found to have ICH. The false rate of radiologists for ICH detection was 1.6%. The parafalcine regions and the cerebral convexity were the most missed areas for ICH.
The study suggests that an AI solution could be applied as an adjunct to current peer review tools of non-contrast CT scans determined to be negative for ICH. (4)
AI Solutions for Prompt Diagnosis of Proximal Large Vessel Occlusion
An intracranial proximal vessel occlusion automatic detection convolutional neural network (CNN) model was used to diagnose middle cerebral arteries (MCA-M1) and internal carotid artery (ICA) occlusions.
A total of 243 patients, including 105 known proximal large vessel occlusion (LVO) – ICA and Proximal MCA (M1) occlusions were observed.
Results showed a sensitivity of 92.3% and specificity of 94.9%, with an accuracy of 94.3% for identifying the occlusion site. These values suggest the possible use of this system as a screening tool in an emergency setting to accelerate diagnosis and increase productivity.(5)
An intracranial proximal vessel occlusion automatic detection convolutional neural network (CNN) model was used to diagnose middle cerebral arteries (MCA-M1) and internal carotid artery (ICA) occlusions.
A total of 243 patients, including 105 known proximal large vessel occlusion (LVO) – ICA and Proximal MCA (M1) occlusions were observed.
Results showed a sensitivity of 92.3% and specificity of 94.9%, with an accuracy of 94.3% for identifying the occlusion site. These values suggest the possible use of this system as a screening tool in an emergency setting to accelerate diagnosis and increase productivity.(5)
Automated Detection of Pulmonary Embolism in CT Pulmonary Angiograms using an AI
Fast and accurate detection of pulmonary embolism (PE), a potentially life-threatening condition, is an essential step for early anticoagulation and improved outcomes. The use of AI to diagnose this condition could assist physicians by highlighting positive exams, hence speeding diagnosis and communications.
In this study, researchers tried to evaluate the performance of an AI system for the detection of PE on chest computer tomography from a large dataset. They identify all CT pulmonary angiograms (CTPA) from their institution during 2017 (n = 1499); then excluded all images with other clinical questions (n = 34) and classified the remaining into positive (n = 232) and negative (n = 1233) for PE. The system was trained and validated using 28,000 CTPAs from other institutions.
The algorithm had a sensitivity of 92.7%, a specificity of 95.5%, and a positive predictive value of 86.6%. The results showed that the system could be used as a decision support tool to speed up diagnosis and complement traditional worklist prioritization.(6)
AI Contribution in Detection of Intracranial Hemorrhage in Emergent Care Head CT
AI Prospective Randomized Observer Blinding Evaluation (AI-PROBE) was used to assess radiology AI systems and a radiology Information Technology (IT) infrastructure.
In medical imaging studies, timely detection of Intra-Cranial Hemorrhage (ICH) is critical because rapid therapeutic interventions in emergency settings are needed to avoid life-threatening outcomes. To evaluate the functionality of the AI algorithm, researchers presented a prospective randomized clinical trial on an investigation of the effect of automatic identification of ICH in 620 emergent care head CT scans on radiology study turnaround time (TAT) in a clinical environment. The analyzed cases’ total sensitivity, specificity, and accuracy were 95.0%, 96.7%, and 96.4%, respectively. Results showed a reduced TAT for diagnosed ICH in emergency setting head CT scans, promising better ICH treatment in a more timely manner. (7)
AI Prospective Randomized Observer Blinding Evaluation (AI-PROBE) was used to assess radiology AI systems and a radiology Information Technology (IT) infrastructure.
In medical imaging studies, timely detection of Intra-Cranial Hemorrhage (ICH) is critical because rapid therapeutic interventions in emergency settings are needed to avoid life-threatening outcomes. To evaluate the functionality of the AI algorithm, researchers presented a prospective randomized clinical trial on an investigation of the effect of automatic identification of ICH in 620 emergent care head CT scans on radiology study turnaround time (TAT) in a clinical environment. The analyzed cases’ total sensitivity, specificity, and accuracy were 95.0%, 96.7%, and 96.4%, respectively. Results showed a reduced TAT for diagnosed ICH in emergency setting head CT scans, promising better ICH treatment in a more timely manner. (7)
References