Application of AI in Medical Imaging – Where to begin
Artificial Intelligence has captured the imagination of healthcare academia, clinicians, students, product developers, investors and industry analysts, thereby generating debate around the topic of man vs machines, and if AI is going to replace doctors. Although Artificial Intelligence (AI) has been discussed since the ’50s, its application in medical practice and imaging diagnostics had been very limited. It is only in the past 4-5 years or so that the hype of machine-based learning became so audible, and questions surrounding its application, benefits and impact on desired outcomes became hot topics for discussions and white paper publications.
At international conferences, the physician community, academia, young graduates and industry thought leaders interested in the intersection of AI and medicine have been keenly observing latest developments when it comes to the application of AI in clinical practice, more so in medical imaging where we are witnessing a profound impact of these machine learning-based algorithms.
In general, the discussions, publications and analysis have focused on the following topics:
- Big Data: leveraging clinical data for machine learning and decision making
- AI Validation: publications documenting the accuracy of machine learning models
- Clinical Application: Use cases and application of AI in medical imaging
- Ethical Aspects: use of patient data, anonymization, cloud and regulatory aspects
- Market Size: industry reports predicting the growth of the market exceeding billion dollars by 2023
So Many Al Algorithms, Where to Begin?
While the discussions have been mostly focused on the use of big data for research analysis and development of AI models, the need for fixing the basics first before a healthcare organization embarks upon introducing AI algorithms into clinical practice needs to be addressed first.
Step 1: Understanding your data, lessons learned from medical records and informatics
What did we learn from EMRs (Electronic Medical Records) – With EMRs, care organizations were successful in consolidating patient centred medical record data, however, a consolidated imaging health record gap still remained unaddressed as images remained in their own silos. Keeping in mind the recent advancements in medical imaging, and as radiology moves out of the traditional basement to operating rooms, to the point of care and even patient’s home, thanks to smart applications and devices; the traditional departmental PACS approach has created IT infrastructure, integration and interoperability challenges.
Thought leading care organizations are leading from the front when it comes to enabling Enterprise Imaging strategy and strong governance to bring meaningful and clinically relevant medical data and imaging informatics, all together.
Establishing an Enterprise Imaging strategy with a solution provider that can help enable a multidisciplinary imaging consolidation is key to removing the siloed approach that exists in healthcare organizations today.
Step 2: Governance, what clinical programs are you trying to improve?
Yes, there are hundreds of AI algorithm developers out there, and every other algorithm may demonstrate a unique feature that the other algorithm may not, how do you decide then? Successful organizations that have enabled an Enterprise Imaging strategy will talk about the role of a strong governance team, that brings together multispecialty stakeholders helping break technology barriers and enabling clinically relevant and collaborative patient-centred care.
What does this mean for AI? Understand first the clinical programs you want to improve; are you an academic centre? Do you run cancer screening programs and diagnostics? Are you a multi-hospital network? Are you a trauma centre? Do you have staff capacity challenge? Or are you acquiring and adding more hospitals or diagnostic centres to your network? You may be a teleradiology provider seeking to improve your diagnostic capabilities, or you may be a public health organization seeking to fund and provide guidance to your care delivery network on the use and application of AI.
A solution provider who has successfully deployed large scale Enterprise Imaging solutions would answer the above questions and provide you guidance when it comes to implementing AI instead of discussing the feature level benefits of dozens of AI algorithms for a lung nodule for example.
Al algorithm developers are doing due diligence when it comes to a standards-based approach, the governance teams’ role is to find outcomes-driven meaningful application of AI algorithms into clinical practice. And, that’s where governance team needs an understanding of the regulatory clearance and validation status of AI algorithms of interest.
Step 3: Power up your AI algorithms with Enterprise Imaging platform
As we breakdown silos of imaging workflows and enable multidisciplinary consolidation and collaboration, the power of a consolidated platform results in the creation of a vast data lake, ready for analysis by radiologists, diagnosticians, researchers and academics to help improve quality of care by better understanding disease and population health data. This helps care organizations progress from descriptive to predictive analytics models as you will have access to clinically relevant data.
The Enterprise Imaging platform strategy unleashes advanced analytics that can be extracted from either single or multiple algorithms, to enrich diagnostic intelligence and automate your workflows. The platform also helps establish a secure ecosystem of standard based APIs and eases the deployment of AI algorithms in clinical practice.
Step 4: Workflow Optimization
AI will not replace Radiologists or other physicians, but in fact enhance their workflow even further, empowering them in their ecosystem and help them make collaborative and intelligent decisions. AI, if it is not embedded into clinical workflows in the logical sense, will not offer many benefits. As technology advances further and diagnostic capabilities improve, so does the impact on how caregivers provide care, demanding more access to critical information in a fast and efficient manner. It is not the Radiologists who are demanding more visibility on the frontlines, it is bidirectional, and there has never been more demand by caregivers and referring physicians, and even patients, seeking consultative dialogue with radiologists more speedily.
Radiologists will be at the forefront when it comes to helping deliver quality care, with diagnostic reports and peer collaboration; AI should help augment radiology tasks even further with critical alerts and notifications. Physicians will benefit from fast access to critical results that require immediate attention, helping reduce wait times and improve referral services for cases that require necessary and urgent patient care coordination.
Step 5: Precision Health, evidence-based KPIs:
Care organizations and health authorities across the globe are faced with pressing population health challenges. Whether it comes to detecting cancers, or chronic disease surveillance, AI and advanced analytics should help improve radiologists and diagnosticians deliver intelligent care with less focus on manual repetitive tasks, and more focused clinical input.
Missed diagnosis, incidental findings, chronic disease surveillance, cancers and related pathologies are creating challenges not only from a screening perspective but also demanding delivery of efficient and cost-effective quality care. The increasing number of imaging modalities, growing exam volumes, limited radiologist capacity and availability, is prompting radiologists to explore applications of AI embedded into their departmental workflows and beyond the boundaries of the hospital to deliver on time-critical intelligence.
Aggregation of patients’ genomic data, imaging biomarkers, pixel intelligence and diagnostic data will be the next steps where Radiology will play a significant role in enabling the path towards precision health with evidence-based application of AI in healthcare.
The industry is coming to a realization that AI will not replace physicians or radiologists, but in fact, enhance their workflows. Algorithm developers will benefit from a collaborative framework that enables a consolidated ecosystem, is standards-based, and improves interoperability and delivers on KPIs. Don’t get pulled into the hype of how many AI algorithms you need, build a strategy around your Enterprise Imaging platform first, and consider the clinical programs you want to improve, AI algorithms will then begin to make sense, and so will the workflows.