ARTERYS

Published by Daniel at November 1, 2022

Arterys, an AI cloud imaging platform, promises to unlock radiology's full potential and become a fundamental pillar in healthcare...

Arising from the need to improve diagnosis for pediatric patients with heart defects, Arterys was founded in 2011 by Albert Hsiao, MD, Ph.D. Fabien Beckers, Ph.D., John Axerio-Cilies, Ph.D., and Shreyas Vasanawala at Stanford University. The founders noticed a high rate of error in pediatric imaging and therefore developed a platform to help physicians achieve a more accurate diagnosis. The platform uses a 4D Flow technology with cloud computer processing that allows a full view of the newborn’s heart and accurately detects congenital conditions in only six minutes. With headquarters in San Francisco and offices in Paris and Calgary, Arterys was the first company to receive FDA clearance for a cloud-based product with AI. It is currently cleared to operate in over 100 countries.(1,2,3) Arterys’ mission is to offer practical and faster radiology image examination while improving the detection rate by up to 70%, translating into significantly improved physician experiences and patient outcomes.(2,4)

Cleared. Seven letters, seven approvals

Arterys is transforming healthcare by decreasing subjectivity in clinical diagnosis, decreasing radiologists’ workload, and responding to the need for increased consistency between practices. The platform, optimized for speed and power, allows users to stay connected 24/7, access patient data worldwide, and protect private information. Nowadays, the company has seven FDA clearances and is considered at the forefront of transforming healthcare. (1,4)

An efficient solution for every need

The world’s first medical imaging cloud platform was initially focused entirely on the pediatric field. Nowadays, its potential can be seen in several areas of healthcare, enabling greater collaboration and data sharing between physicians and patients. Arterys has developed four main solutions for imaging processing. The first is Cardio AI, which assists cardiac MRI imaging analysis and uses parameters including LV/RV function and vessel flow without compromising precision. Another section is the Lung AI division, created for chest CT analysis for screening exams, which can recognize solid, part solid, and ground glass nodules. One of the newest sections is Chest and Musculoskeletal AI XR, planned to aid physicians in using X-rays in the emergency department.

This tool detects seven of the most common abnormalities in that area, including fracture, dislocation, pleural effusion, and pneumothorax, and categorizes cases according to priority. Neuro AI, the final division, offers solutions for neurovascular emergencies and detects acute intracranial hemorrhage in a few seconds. (5,6,7,8)

Another section is the Lung AI division, created for chest CT analysis for screening exams, which can recognize solid, part solid, and ground glass nodules. One of the newest sections is Chest and Musculoskeletal AI XR, planned to aid physicians in using X-rays in the emergency department. This tool detects seven of the most common abnormalities in that area, including fracture, dislocation, pleural effusion, and pneumothorax, and categorizes cases according to priority. Neuro AI, the final division, offers solutions for neurovascular emergencies and detects acute intracranial hemorrhage in a few seconds. (5,6,7,8)

Where AI, research, and workflow meet

These products have several advantages, such as their web-based nature and integration with multiple software. It can integrate with picture archives and communication systems, radiology information systems, electronic medical records, and dictation software. These cloud-based solutions decrease on-premise infrastructure and accelerate deployment, scalability, and research. This is done by including advanced, formerly time-consuming, image processing in radiologists’ workflow with minimal workload repercussions. Measurements and calculations in Cardio AI have demonstrated a significant improvement in measurement compared with manual measurement on heart MRI analysis. The impact of its aid in rendering and segmentation is estimated at 25 minutes saved on average per study.

Similar benefits for pulmonary nodules on low-dose CT include misdetection reduction of up to 70%, automated Lung-RADS calculation, and a standardized reporting system. Along with these features, reductions in low-dose CT reading time by 45% and increased peer-to-peer standardization and agreement have been demonstrated. (9,10,11,12,13,14)

Measurements and calculations in Cardio AI have demonstrated a significant improvement in measurement compared with manual measurement on heart MRI analysis. The impact of its aid in rendering and segmentation is estimated at 25 minutes saved on average per study. Similar benefits for pulmonary nodules on low-dose CT include misdetection reduction of up to 70%, automated Lung-RADS calculation, and a standardized reporting system. Along with these features, reductions in low-dose CT reading time by 45% and increased peer-to-peer standardization and agreement have been demonstrated. (9,10,11,12,13,14)

Unified yet diverse tools

Arterys also allows the development and deployment of new third-party models. These use Arterys Software Development Kit to generate an algorithm and then upload it, allowing widespread implementation and acceleration of external validation by data and imaging input. The data can be anonymized and, if desired, shared with the algorithm developers maintaining this anonymity to support the endeavor with training data. Sharing the output with colleagues for medical discussion is also a useful feature. Some users of this marketplace include UCSD’s AiDA laboratory, Imaging Biometrics, and Vuno. NVIDIA Clara was recently incorporated into the Arterys platform, making it easier for developers to upload their models on this image analysis platform. Many other cleared algorithms or in-training models are available or under development to meet radiologist needs in several fields. (15,16)

Sharing the output with colleagues for medical discussion is also a useful feature. Some users of this marketplace include UCSD’s AiDA laboratory, Imaging Biometrics, and Vuno. NVIDIA Clara was recently incorporated into the Arterys platform, making it easier for developers to upload their models on this image analysis platform. Many other cleared algorithms or in-training models are available or under development to meet radiologist needs in several fields. (15,16)

Global data-driven efforts

NVIDIA Clara tools for image analysis on CT scans of patients with suspected COVID-19 are currently under development in Arterys’ marketplace. Based on CT findings, this tool’s complete development can estimate the odds of a patient having COVID-19. The training model used more than 3000 CTs, including 2000 CTs from confirmed COVID cases in the US, Italy, Japan, and China. Some other models for the assessment of COVID-19, including AiTrox’s and PingAn’s, are also available on Arterys’ page. (17,18)

A picture leader

This novel opportunity for innovation has earned many awards and distinctions. Some worthy of notice are the Top 100 Digital Health Companies by The Medical Futurist and the winner of Cardiology2.0’s Top 25 cardiac imaging leaders 2020. (10,19)