Elekta, the big name in radiotherapy, radiosurgery, and oncology informatics that’s based in Sweden, recently partnered with IBM to offer the Watson for Oncology artificial intelligence (AI) platform along with its MOSAIQ Oncology Information System and other cancer care solutions. According to Richard Hausmann, the CEO of Elekta, it is “the first radiation therapy company to offer capabilities that combine conventional health information systems with artificial intelligence and cognitive cloud computing.” The hope is to introduce AI in a meaningful way to treatment planning and delivery so as to improve reliance on evidence-based care and lead to improved outcomes.
We spoke with Andrew Wilson, Elekta’s Vice President of Global Marketing for Software Solutions, to find out more about the new offering, what AI has in-store for oncology care overall, and how medicine will be affected more broadly by deep learning and related computing fields.
Medgadget: Artificial Intelligence is clearly a buzzword lately, and many companies claim to use the technology. It is not clear what the term means with respect to different applications. Can you give us an idea of what the AI is and its role in patient care?
Andrew Wilson, Elekta: There are a lot of companies that utilize AI or other tools that use machine enabled deep learning functionality in ways that can benefit the oncology treatment field. This is due to several factors: there’s an expansion in the amount of data that is collected and there are potential questions around the amount of data that can be related to disease that we may not understand at this stage. Today we understand how a number of items contribute to improved patient outcomes or improved diagnostics surrounding particular diseases, but we’re sure that by the time we aggregate these data sets we’ll start to see additional data sources providing information around patient outcomes and responses. Then we’ll have to think differently about which data are going to change the way patients are managed, and what those changes will look like.
From a general teaching perspective, there’s so much of this data being amassed so quickly that it’s going to be difficult for people to comprehend what all these sources are going to mean for improving care. Being able to identify which data sets are tags for particular aspects of a patient’s genetic information or impact treatment techniques will be important for determining the value outcome for any given analytic tool.
At Elekta, we’re looking at the AI landscape and working with our internal physics and AI teams to identify clinical need and opportunities to incorporate AI tools into our products. There are quite a few areas. The area of decision support is one where the American Society of Clinical Oncology (ASCO) has several definitions on where decision support engines need to be utilized and how they can qualify appropriately, and 15 criteria that should preempt the capability of a decision support engine. Partnering with a company like IBM, which has already fulfilled these requirements and has a global footprint, gives us high confidence that we will effectively incorporate AI into our clinical workflow solutions.
Medgadget: More specifically, can you give an idea of what variables Watson for Oncology is looking at, what kind of guidance it provides, and what its role is in the clinical workflow?
Wilson: Watson for Oncology provides an assessment of patient data using natural language processing. Different views and inputs are provided on the patient regarding patient diagnosis, laboratory results, patient assessment charts and letters, and clinical reports that talk about the disease. Watson for Oncology takes all this data and reviews it against the corpus of information that it reads. It’s a large corpus of over 300 journals and 250 textbooks, about 15 million pages of information, and Watson for Oncology will rapidly read all those pages and compare them with the patient data.
The system was trained by Memorial Sloan Kettering Cancer Center to analyze patient information in the context of this tremendous body of literature and guidelines and provide to the clinical team with ranked treatment recommendations. The ranking is based on factors such as outcome information, potential for improved response to therapy, availability of resources, and other variables. It’s a support tool, and the good thing about it is that all the evidence that contributes to the ranking is provided there for the physician to review. So it’s not just “here’s a top ten” and take your pick. If you go into the system it will give you pages and pages of literature, and articles, and response documents that have been published and peer reviewed that will allow the physician to understand why the system ranked one particular technique over another. Then it’s up to the physician to determine the best therapeutic approach based on that ranking, also taking into account his or her clinical experience, and the patient’s stated goals and objectives. If the physician has specialized protocols, he or she can evaluate which of those protocols best aligns with the rankings that Watson for Oncology has provided.
Clinicians can use it as a second opinion, and it may also be an important source of clinical insight in care centers that don’t have routine access to a vast number of experts. We believe that it will be an important tool for supporting clinical decisions that result in optimum care in a broad array of care settings.
Medgadget: Are there IT infrastructure requirements that the hospital has to fulfill?
Wilson: Watson for Oncology is cloud-hosted by IBM. There’s some configuration work that IBM will do with the customer site so that healthcare professionals at the hospital that want to use it have access to the system. It’s really just a case of internet logins. It’s very simple for departments to get cloud access. There’s some work that Elekta will do with our MOSAIQ customers as well.
Medgadget: Because there’s so much data that is coming into AI systems, do you fear that one day it will be nearly impossible for a physician to have a grasp of how a medical AI product came up with its conclusions?
Wilson: I don’t think that will be a problem the way Watson for Oncology is configured. Because the information in the Watson review is published data that’s peer reviewed. Because the way Watson works, it’s an advisory tool and it provides a ranking and all the information that went into that ranking. So it’s up to the physician and department to review that and determine whether that ranking is relevant to their current patient cohort or how they would normally treat a patient. Watson doesn’t force them to treat a patient in any specific way. It is even possible that there are five different ranked items that are acceptable to be delivered. So the department can review the data that exists or the evidence that appears with that data, and they can make the correct decision among the ranked options which is the most appropriate that the department can treat. Because the evidence is there, they actually have the ability to take that evidence up-front to the insurance provider in case there are any questions about insurance payment.
Medgadget: What is the future for AI in oncology care? Is Elekta working on its own technology in this field or do you intend to continue to partner with firms specializing in AI?
Wilson: I think the future of AI is an exciting topic for manufacturers, healthcare providers and patients. As a leader in RT for decades, we want to provide guidance on the data that we collect in the radiotherapy setting and the data that will provide influences, in terms of change, in terms of decisions made about how patients are treated, and what that can mean for the impact of treatment and the operational aspects of delivering that treatment.
We have a Memorandum of Understanding with IBM and expect to bring Watson into more multidisciplinary departments, which include radiotherapy and comprehensive oncology. The feedback that we’ll receive about workflows and impact on radiotherapy delivery techniques from these efforts will allow us to provide more advisory information to IBM to help them create a broader tool in the Watson for Oncology model. For example, combining Watson for Oncology with MOSAIQ will soon make hundreds of attributes from the patient’s electronic health record, including physicians’ notes and laboratory data, available for analysis using IBM’s Natural Language Processing technology. This will expand MOSAIQ’s capabilities while providing robust radiation oncology data sets that can be fed into IBM’s machine learning algorithms to advance the functionality of Watson for Oncology. These are the types of transformative advances that can occur when you have pioneers such as Elekta and IBM working together to innovative 21st century solutions for today’s patients and physicians.
Links: Watson for Oncology and Elekta MOSAIQ
Original announcement: Elekta taps IBM Watson Health to bring AI to comprehensive oncology field