Artificial Intelligence in Radiology

Artificial intelligence in radiology has undergone something of a metamorphosis. From a few innovative solutions tentatively testing the edges of capability and potential to a virtual deluge of algorithms, platforms and solutions, it has grown as both a technology and a market. However, like most technology solutions, artificial intelligence (AI) is not perfect. It is also, most importantly, not a replacement for human beings. It is best defined as a collection of algorithms, machine learning tools, sophisticated neural networks, and systems that are changing how radiology services are delivered.   

Since it was first introduced as a concept in the medical profession, artificial intelligence has been eyed with suspicion. Many professionals have been concerned that it will be used to replace their expertise and potentially negatively impact on patient care and results. Others have been concerned that the technology isn’t advanced enough to even begin to engage with clinical practice. These are valid concerns. The truth is that artificial intelligence in radiology is not about taking over, it’s about reshaping the future. 

The real value of AI lies in how it can be used in collaboration with the radiologist or medical professional. In how it can be used to enhance and support the professional by streamlining the process, reducing the diagnosis burden, and improving workflow efficiency. Artificial intelligence in radiology is a tool, not a sentient being. It is an investment into technology that allows for ongoing improvements to diagnosis and patient care by supporting the radiologist as they battle increasingly weighty workloads. Back in 2015, a survey published in Academic Radiology found that radiologists need to review one image every 3-4 seconds to keep up with their workloads.  There’s little research highlighting how much heavier these workloads have become over the past five years, but it’s very likely to be even more demanding today. 

This is where the true value of AI lies. 

The Evolution of AI in Healthcare

In the Accenture report: ‘Artificial Intelligence: Healthcare’s New Nervous System’, the company points out that AI has the potential to ‘truly augment human activity’. This forms the crux of how technology has evolved, specifically within the radiology industry. It can only deliver on this premise if it is developed correctly, with the right datasets. The value of artificial intelligence in radiology lies within its ability to analyze structured and unstructured data that supports more efficient diagnose reduces errors and increases efficiencies. This is not a goal that can be achieved without the full co-operation of the radiology practice and medical profession. 

In a recent analysis of ‘The State of Radiology AI in 2019’, Diagnostic Imaging emphasizes the importance of the profession in the long-term success of AI solutions in radiology. Radiologists have to monitor and manage how AI is deployed in their environments, and they have to recognize that it is going to have an impact on their careers and their working environments.  While some radiologists and practices have embraced the potential of AI, many remain skeptical and aren’t paying attention to how it is changing and what this means for their futures. 

And it is changing. Over the past few years, artificial intelligence in radiology solutions have amassed some impressive accreditations and approvals from healthcare organizations. The Therapeutic Goods of Australia (TGA), the US Food and Drug Administration (FDA), and the European Union CE have all put their mark on AI solutions, certifying their ability to deliver the goods. Aidoc, one of the market leaders in the AI radiology space, has already obtained the CE mark for four algorithms and has FDA clearance in triage for cervical spine fractures, pulmonary embolism, and intracranial-hemorrhage and large-vessel occlusion. Considering that the number of FDA approvals for algorithms remain well below the 30 mark, this is an impressive achievement. 

It also shows how rapidly AI is evolving within the radiology space. 

At the 105th Scientific Assembly and Annual RSNA Meeting 2019, an entire area was dedicated to artificial intelligence solutions compared with a few measly tables only two years earlier. This trend will continue to echoe at the European Congress of Radiology (ECR) Vienna 2020 event which will launch its inaugural Artificial Intelligence Exhibition (AIX) and an entire AI-focused lecture itinerary. AI has become more than just a tentative stab at using technology to help save lives, it has become ingrained within the medical profession and radiology. 

How Artificial Intelligence is Becoming Ingrained in Radiology

Artificial intelligence in radiology is becoming an increasingly integral part of the profession and daily life. Often described as that extra team member, the one that never sleeps or gets tired and can sift through images without pause, AI adds immense value. It has become slowly embedded into numerous medical institutions globally, filtering into workflows and systems, and being customized to match radiologist requirements. 

What makes AI in radiology stand out is the fact that the algorithms have advanced to the point where they can absolutely support clinician decision making. Global Diagnostics Australia (GDA), was one of the first diagnostic imaging companies in Australia to include AI as part of its radiology workflow and it achieved notable results. The high-end algorithms were incorporated into the care management pathway to expedite patient diagnosis and treatment across the head, chest and neck. The AI solution was designed to prioritize patients according to their critical status, alerting the radiologist to urgent cases and reshaping their approach to workflow and diagnosis. 

At the University of Rochester, resident doctor Komal Chugtai, found that AI was of immense help when getting through immense workloads. In an interview with Aidoc, she explained that AI had already delivered some ‘Wow’ moments – flagging something an urgent case for prioritization. It was a view shared by Dr. Eric Grey at DCC San Diego, who uses the notifications from the AI platform to do a rapid assessment of patients at the start of the day. 

These notifications can provide the radiologist with a much-needed pillar of support in busy and overwhelming environments. The solutions themselves, however, can potentially reduce diagnostic errors and help mitigate the risk of physician burnout.

Benefits of Artificial Intelligence in Radiology

Physician burnout has become endemic. It’s a threat to a professional’s wellbeing and career and radiology is one of the high-risk areas within the medical profession. The Medscape Radiologist Lifestyle, Happiness and Burnout Report 2019 found that only 25% of radiologists were happy while 44% reported that they were experiencing some form of physician burnout. 

The long hours, the endless tasks, and the admin were just some of the reasons highlighted as the causes of burnout. There’s been a lot of research into how the profession can be more supportive of its people and reduce the stresses that cause burnout and there are solutions that include self-care, medication, and wellness programs. There’s also AI. It has the ability to step in and take away a significant portion of the load, providing the radiologist with a second pair of hands and eyes that give them a sense of support and security. 

AI has the ability to pick up enough of the image diagnosis weight so that the radiologist can focus on the complex cases that require their specialist attention. It can effectively flag urgent cases and streamline the process in the face of increased pressure from both the medical facilities and regulatory institutions.

In addition to helping to mitigate burnout, AI supports the teleradiologist. In fact, it is probably the most symbiotic of the AI and radiologist relationships. AI helps reduce waiting times for emergency patients that have to be transported from rural and remote areas as it can be used to help diagnosis and assessments over vast distances.  AI in teleradiology can be used to facilitate analysis and provide radiologist support. For GDA, the algorithms provided by the Aidoc system have helped to prioritize patients according to their critical status so they are treated and diagnosed first both onsite, and for their teleradiology services. This solution, developed in partnership with IDG, has so far managed to provide support to more than 400, 000 patients in Western Australia and has proven the value of AI in the teleradiology profession. 

Finally, another benefit that stands out is how artificial intelligence in radiology can help support report turnaround times (RTAT). AI solutions can potentially help speed up RTAT as the data is embedded within the workflows and can be easily extracted to facilitate report development and delivery. Aidoc is one of the market leaders in AI radiology solutions with a wide range of pathologies and modalities on offer. The company has a clearly defined roadmap that showcases its capabilities and outlines the pathologies that it’s planning to undertake over the next five to ten years, ensuring that it remains future-forward and future proof.

Explore the Latest AI Insights, Trends and Research