Machine Learning, AI and Medical Imaging: 4 SIIM Takeaways

The impact of artificial intelligence (AI) and machine learning (ML) on the future of healthcare is one of the hottest topics in the industry. The technology has the potential to transform multiple departments in the health system, from pathology to radiology to patient care and beyond.

Recently, SIIM held their annual Conference for Machine Intelligence in Medical Imaging (CMIMI). Change Healthcare was there to participate in the ongoing conversation, provide guidance, and stay up-to-date on the latest developments.

The Future of Machine Learning and AI in Medical Imaging

Here are four major topics of discussion at CMIMI that every medical imaging professional should be aware of.

1. Natural Language Processing

One of the hottest topics at CMIMI was about how AI might change the way radiologists go about their daily tasks. Natural language processing is the backbone of most dictation and translation systems today. It could make generating and searching reports or messaging providers as easy as checking the weather with Alexa.

With the advent of new NLP approaches, one can now mine millions of physicians’ and radiologists’ notes and reports with precious clinical data to automatically generate reports, conduct smart contextual search, and feed with reduced resources the training of image-based AI with a large volume of clinical data. Like image-based AI, natural language processing is a powerful enabler of more efficient workflows and allows radiologists to focus on high-value tasks.

2. AI Augmenting Radiology

AI is continuing to show its value in situations where it can overcome the limitations of human perception. Routine chest X-rays is one example where there are some difficult to detect and diagnose abnormalities such as pneumonia. A recent study showed that a deep learning algorithm was better or equal to radiologists with different levels of experience. However, when it came to assessing changes in pulmonary opacities over time, the algorithm had significant limitations. Over time, this type of algorithm could help a radiology department expedite image interpretation and help improve read accuracy.

3. Regulatory Clearance

At C-MIMI, it was reported that the regulatory process is evolving with the recent introductions of a software pilot precertification program and the de novo approval process for low to moderate risk devices. Also, the FDA may be able to judge an application on a “clinical need” basis rather than performance, which would help extend the regulatory process to new types of AI-based solutions. However, clearance and approval remain to be done for each application, which can be costly and is hardly scalable. There are also new AI-driven workflows that yet need to be addressed by the FDA such as “reinforcement learning,” where the algorithm is updated constantly with new data.

4. Expectations Vs. Reality

Every day, new use cases for the application of AI in healthcare are being explored. If the opportunities to dramatically impact radiology productivity, safety, accuracy, and early disease detection exist, then there is still a mismatch between the expectations of AI and today’s reality in medical imaging.

This is something that was well recognized by the panelists at the C-MIMI conference in San Francisco. Addressing aspects such as the clinical assessment value of AI and the algorithm market, data collection costs, workflow integration, regulatory paths, and deployment challenges is critical to ensuring a successful commercialization. An increasing number of academic and industrial players are now paying closer attention to these challenges, which announces more infusion of AI in PACS and vendor systems soon.

For now, the dream of a full-fledged AI radiology assistant is still a few years off. However, AI is highly useful as an enabler of more efficient workflows, improved diagnostic capacity, and the potential for earlier and more efficient patient treatment. Health leaders should take a clear assessment of AI for medical imaging by looking at what is available commercially and how it can make a difference in their organizations.

Enterprise Medical Imaging solutions from Change Healthcare can help improve clinical, financial and operational outcomes.