Enhancing the Medical Imaging Industry with AI

Medical imaging is an important component of modern healthcare delivery. It ensures appropriate screening and diagnosis, enables preventive care and appropriate management of diseases. It also provides important prognostic data and risk stratification of health-related outcomes.

AI has been widely positioned as being the most important and potentially disruptive recent advancement in radiology. Singapore radiologists have been quick to embrace this technology as part of the natural progression of the discipline toward a vision of how clinical medicine, empowered by technology, can deliver value-based and patient-centric care.

Addressing Industry Needs
Current Medical Imaging AI solutions use proprietary platforms which can only deploy and run the platform provider’s own limited number of AI models and applications. Current Medical Imaging AI models are narrow in scope and focus on a specific modality, such as X-rays and 3D MRIs, or specific body parts such as the chest and brain.

However, medical imaging workflows such as diagnostic radiology require clinicians to work with multiple modalities and body parts, and across multiple clinical purposes such as triage, diagnosis and treatment. Also, current proprietary platform solutions generally do not support the development and fine tuning of AI models and applications by users.

All of the above makes widespread deployment and scaling of AI models and applications difficult. To enable healthcare solutions to provide the best care for patients, they cannot be isolated to specific platforms. Applications must be able to talk to and integrate with other systems. Additionally, data is the essential fuel for AI modelling development, so access must also be made simpler and also kept secure.

Developing an “App Store” for Medical Imaging Professionals

Today, smartphone users are spoilt for choice. A quick glance at Apple’s App Store or the Google Play Store reveals an open, thriving ecosystem that provides universal applications from both first and third-party developers.

Imagine if Apple’s App Store was a fraction of its current size and iPhone users were limited to using only Apple-developed applications with no access to solutions from third-party vendors. It could be challenging to find a solution for your needs and users might be forced to carry multiple non-Apple devices to gain access to applications from other third-party developers.

To avoid such a scenario in healthcare, IHiS and its partners aim to create possibly the world’s first common AI Medical Imaging Platform that is open and vendor neutral. Essentially, this will be the “app store” for medical imaging professionals. This complex solution will address the challenges they face in delivering quality healthcare today

What We Hope to Achieve
The proposed platform has four key objectives:

Provide an open and vendor-neutral platform which can deploy, connect and run AI models/applications from different sources for various imaging modalities to assist clinicians’ work. These sources include third-party AI providers, user self-developed, and/or the platform provider, whether hosted on the platform or externally.
Provide access to a marketplace of AI models / applications.
Deploy and operationalise AI models / applications for local validation and routine clinical practice.
Support the development and trial of AI models / applications on an explorational platform that can be deployed seamlessly onto an operational platform.
Depending on the implantation, an AI-augmented medical imaging workflow can lead to improvements in outcomes and productivity.
Promote Better Clinical Outcomes

• AI model can be trained to detect early sign of diseases

• AI acts as 2nd peer AI-reader to improve accuracy by flagging out discrepancies between human-reader

• Different AI models can review the same images and co-relate with medical records to detect multiple conditions

• AI acts as 1st read for workload prioritization to improve the time to treat

Improve Productivity

• AI can work 24x7

• AI can highlight abnormal cases to allow clinicians to focus on more complex tasks and decisions

• AI can be used as an audit tool to sample imaging read by junior radiologists

• AI can be used as a training tool for junior radiologists

Examples of Use Cases
Here are two key examples of current projects which would benefit from being deployed on the AI-enabled Medical Imaging Platform.

An AI-enabled predictive model for Community-Acquired Pneumonia

The Singapore Eye LEsioN Analyzer (SELENA+), as part of Singapore’s Integrated Diabetic Retinopathy Programme (SiDRP), performs automated analysis of retinal images to detect diabetic retinopathy.

With the ongoing COVID-19 pandemic, there are demands to leverage AI technology to support the healthcare system in its fight against the disease. This includes COVID-19 models from Alibaba and BioMind, and a pneumonia model by NCID-A*Star.

Supporting Partners