Healthcare sector is welcoming machine learning as it ushers accuracy and better predictive decisions. Mansha Gagneja catches up with Dr Vidur Mahajan, Associate Director, Mahajan Imaging to learn about AI’s impact on the radiology sector and understand its adoption at his centre
In the world of scientific revolution, simulation of human intelligence processes which includes learning, reasoning and self correction by machines took flight. Artificial intelligence (AI) was soon spreading its wings across healthcare spectrum. Bringing machine learning to healthcare proved to be beneficial for identifying critical medical conditions, which would potentially allow for earlier intervention and better outcomes. The biggest benefits of AI encompasses improving patient care while reducing costs. The radiology sector fails to remain untouched by the AI wave and is betting on new technology to boost the sector.
To understand the application of AI in Indian radiology, we took insights from Dr Vidur Mahajan, Associate Director, Mahajan Imaging, who follows his passion of improving access and affordability of high-end medical services with the help of AI. He gives an elaborate view on why AI is the next big thing that will revolutionise the industry. Having education and experience in both the clinical practice and business management he explains how healthcare providers want to provide faster, cheaper and more effective care to their patients and build a sustainable business.
Genesis
Advent of AI was inevitable, but it came in much later, even though most of the mathematical models that are used by AI scientists today were developed during the 1950s-1970s. Earlier, computer systems were not dynamic, which in turn staggered the development of such technologies.
Radiologists have started using advanced post-processing tools and acquisition consoles ever since the development of CT, MRI and nuclear medicine techniques. The accuracy and efficiency of these post-processing tools has been improving drastically over the last few years for example, post-processing a coronary CT angiogram (a CT scan of the heart to see its blood vessels) used to take in excess of an hour till a decade ago – today it can be done in less than 5-10 minutes with the right tools. Dr Mahajan highlights that many of us use some form of Artificial Specific Intelligence (ASI) every day without even realising it. For instance, when we ask Google Maps to show us the way from one place to another, there is an AI system that does millions of computations to provide the shortest and quickest path. The next frontier is the development of Artificial General Intelligence (AGI) which will be designed to act more like a human and will be able to accomplish many tasks which were earlier only our forte.
Analysing Indian demographics, we can draw conclusions that physician to population ratio is quite inadequate. This issue may not be resolved by merely allotting more medical seats. Telemedicine service providers, along with the government, are taking great strides in improving the penetration of healthcare in India. However, it is imperative to incorporate some kind of AI system to streamline the process. Dr Mahajan comments, “While a lot of venture capital funding has gone into radiology, true integration of AI seems to be at least a few years away from truly transforming the industry, given the complexity of the human body, huge data requirements, ethical considerations, and legal connotations.”
Synthesis
AI in Indian radiology is yet to pick up pace and to facilitate this, Mahajan Imaging is already experimenting with the new technology. He acknowledges,“This revolution is fundamentally transforming Mahajan Imaging’s long-term strategy causing a shift in our mindset of being a typical healthcare provider, to more of a technology company and have started using Mammography-Computer Assisted Diagnostics(CAD) around four years ago”. He highlights that the doctors work on advanced radiology workstations – GE Healthcare’s Advantage Windows and Philips’ Intellispace Portal workstations. To aid the transformation of radiology, they have collaborated with multiple AI companies. These companies range from start-ups like Predible Health, Bangalore; Labsadvisor, Delhi; Semantic MD, US, to GE Healthcare. He adds, “One of our prized partnerships is with the Department of Brain and Cognitive Sciences at MIT, where we are helping develop computational models for human vision which we have been doing for eight years now.” Giving the companies access to their data and providing a platform to test their innovations in the real world may drive growth. And if this partnership turns out to be fruitful, chances are that Mahajan Imaging will attain customised products backed by latest machine learning. Dr Mahajan believes that AI will definitely transform radiology one day and they want to drive that transformation, versus just being an observer.
Metamorphosis
Mahajan Imaging Centre acquired CAD system along with Fujifilm’s full field digital mammography system and it has enabled them to pick up lesions that might have otherwise been missed. The available applications improve efficiency which speeds up the process of post-processing of images, most commonly CT Angiograms, Liver Segmentation, Tumour Segmentation etc. In regards to accuracy, Mammography CAD has been around for a long time and is assisting radiologists read digital mammography scans more thoroughly. He identifies,“Our ‘ah-ha!’ moment came when we were able to pick up a lesion that one of the world’s leading hospitals in the US missed, on the same images, simply because we were able to use CAD.” Another emerging field is Radiomics, where quantitative analysis of images is performed to reveal deeper insights, is another aspect that will redefine the field of radiology as humans can only see a limited approach of the data acquired by imaging machines. These are only the tip of an iceberg. Most of the applications, that will eventually have a widespread impact are currently either under development, or under validation.
The challenge that AI companies face today is the issue of accountability. Although AI systems would be suggesting diagnoses and assisting radiologists through analysing reports, soon AI systems would constitute a major role in diagnosis and prescription. Dr Mahajan expresses his concern as who would be held responsible for errors made? Would it be the software developer, the company that procured the software or the doctor who is signing off? This is the same issue that self-driving cars are facing today –who is to be blamed for an accident –the software, the manufacturer or the passenger?
Indeed, if we are devising these innovations, the burden of identifying and rectifying loopholes rest with all stakeholders. How will the radiology sector take up these innovations and use it to thrive will unfold only with time.