The Coming of Advanced Visualisation Inc.
Radiology is no longer a separate and isolated unit where an imaging order is submitted, and a report is sent out. RSNA this year illustrated how radiology is becoming more and more incorporated into the wider healthcare provider organisation. With IT having spread throughout wider healthcare and artificial intelligence slowly being accepted to increase efficiency and improve diagnostic accuracy, the stage is now set for strengthening the interactions between radiology and the rest of the healthcare organisation. Although RSNA showed us the latest developments in core advanced visualisation technologies, many vendors were focused on how imaging becomes part of the wider network through workflow tools, clinical context, and 3D printing connecting the imaging team with surgical planning and patients.
Yes, we do need AI
The time where artificial intelligence was surrounded by myths and fear of losing jobs has passed, and a new era of acceptance and focus on the great benefits of the technology has begun. Recent years’ successful incorporation of machine-augmented diagnostics in advanced visualisation tools to perform automatic tissue segmentation and measurements, has paved the ground for a general acceptance of the technology. This is enabling AI in all forms to spread throughout the entire workflow; backwards to patient registration, scheduling and dose optimisation, and forwards through assisted diagnostic interpretation & second read, to providing and organising clinical context and longitudinal records to be used in the final decision making.
The rising demand for efficiency and speed in diagnosis combined with an increasing data volume to sift through, in a healthcare system increasingly focused on value, is encouraging the radiology community to work together with the imaging IT industry to develop and invent the tools they need to perform successfully and efficiently. AI is not going to replace the radiologist in near future, but it is hopefully going to make radiology more efficient, safe and exact. The initial focus of AI has been on image analysis and diagnosis, but many companies are now going for the low hanging fruit and applying AI in workflow and supporting tools instead. As AI based workflow and supporting tools gain acceptance, this will pave the way for image analysis and interpretation, where the requirements with regards to accuracy and regulations are making it more challenging. With the help from AI, radiology is now being connected closely with upstream processes, clinical context and the whole clinical workflow. The benefit of this will be a less siloed, more fluid structure, where knowledge and data is easily transferred between teams in more intuitive ways.
Core Radiology Tools and Advanced Visualisation Still at the Centre
Although much of the AI focus this year has moved to workflow and supporting tools, assisted image analysis and automated second read continues to get some attention. With 26.5% lung cancer cases missed by traditional chest x-ray according to National Lung Screening Trial (NLST), several AI companies are taking on the challenge with deep learning technologies, attempting to improve the stats of the radiologist by increasing speed and accuracy as interpretation or second read technologies. Examples of this in the x-ray, CT and MRI space include Lunit, RADlogics, Riverain Technologies and Infervision.
In addition to detection applications, AI is also being applied in visualisation, segmentation and quantification. The purpose of this is to generate reproducible, accurate and fast diagnosis by providing the radiologist with automated tools and measurements. Arterys showcased its 4D Flow medical imaging software for 2D and 4D phase contrast workflows and cardiac function measurements, enabling physicians to measure and visualise blood patterns within the acquired 3D volume. Quantib is applying machine learning for quantification and segmentation of MRI brain scans and have just received additional funding, and Quibim is exploring quantitative biomarker analysis in radiology to detect changes produced by disease or drugs.
Vital Images presented their newest Vitrea Advanced Visualization software with semi-automated cardiac measurements, modality updates, increased rendering and faster time to first image. Vital Images still have a strong position for their advanced visualisation, multimodality, multiuser enterprise solution, but it will be exciting to follow the company after the acquisition of Toshiba’s medical division by Canon, to see if the synergies they expect with Canons image processing technologies in the safety and security sector as well as healthcare will pay off over the coming years.
Enterprising Advanced Visualisation
Disparate and incompatible systems are inhibiting the development of a truly fluid, intuitive, AI assisted healthcare system, where image analysis goes hand in hand with historical patient data, family disease history, genotyping data, and public health monitoring & comparison datasets. Enterprise solutions enabling imaging IT to merge with the remainder of the organisation in a wider clinical content management system was still a hot topic at this years RSNA. However, until enterprise solutions can truly satisfy the needs of the individual radiologist, they will not be able to rightfully replace PACS and diagnostic viewers with specialised clinical applications.
The divide between diagnostic and enterprise viewers may be replaced by enterprise systems with advanced visualisation features embedded and enabled for radiologists, side-by-side with simpler clinical review viewers. Visage Imaging has long been pushing the single viewer concept, and showcased its Visage 7 enterprise platform with AI assisted diagnosis, advanced visualisation and clinical modules for mammography, neurology, oncology, cardiology and vessel analysis. Likewise, Philips announced the introduction of IntelliSpace Enterprise Edition for Radiology giving access to workflow and AI guidance tools as well as IntelliSpace Portal for advanced visual analytics and clinical applications. With more and more advanced features and clinical applications being incorporated into enterprise viewers, the next model of clinical content management may be winning support. However, some questions remain: What will the preferred architecture be for dealing with the high processing needs for AV? Will cloud speed ever be able to catch up with the increase in data volume, or will we continue with enterprise user adaptable interface with workstation and server storage? And will the advanced visualisation features and capabilities on enterprise soon be able to satisfy the requirements of the radiology community so a transition would be feasible?
3D Printing and Virtual Reality Links Imaging to Downstream
When it comes to connecting diagnosis in the radiology department with treatment & surgical planning, 3D modelling and printing is becoming increasingly popular. Philips announced the debut of IntelliSpace Portal 10 which included a new 3D modelling application for creating and exporting 3D models into the clinical workflow. Siemens Healthineers syngo.via platform now gives access to Mimics inPrint software for 3D printing of anatomical models intended for surgical planning, training and patient communication, and GE Healthcare previously announced STL file creation functionality for direct 3D printing without the need for further processing in their Advantage Workstation. More partnerships with 3D print companies, such as Stratasys enabling print on demand, opens the market for smaller healthcare providers not interested in having their own printing facilities. If 3D printing for illustrative purposes is just a new feature easily explored in healthcare by using existing technologies, or if it is going to bring any actual value in a world increasingly digitalised, is the question. However, when we have the option easily at hand, why not exploit it? Once the capabilities are fully established and incorporated into the existing systems, it may speed up the development of 3D printed prosthetics and 3D bioprinting in the longer term, and aid in the surgical planning and anatomical modelling for complicated anatomy in the shorter term.
EchoPixel presented their 3D modelling solution somewhat different from traditional 3D printing. With its True 3D interactive mixed reality software using a pen, 3D glasses and a screen, the user can get a more intuitive experience than a 2D screen without the use of large 3D goggles, and be able to identify for example flat lesions of colon cancer not easily detected otherwise. For illustrative, research and training purposes this may have potential, particularly if tissue density and movements could be animated in later versions for surgical simulation and training for complex surgical interventions.
The wider healthcare system and healthcare IT is developing quickly around the radiology department, but it is important to remember the strong position imaging and advanced visualisation still has when it comes to influencing the progress, as it remains at the centre of the diagnostic decision-making. Radiology is being integrated into the wider organisation through workflow tools, clinical context, 3D printing and enterprise solutions, but unless the analysis speed and core advanced features necessary for correct and timely diagnosis will be sustained, the system will not hold water. One thing is for sure though: Imaging will no longer live in isolation.