top of page
  • Writer's pictureEdited by: Atif Zafar, MD

This Is Why Artificial Intelligence Will Transform The Healthcare World.

The new cool thing in town, Artificial Intelligence – #AI, is creating some ripples. Most physicians and nurses I have spoken with are clueless about AI and its implications. Prior work on AI's role in other imaging modalities such as in cardiac function imaging and neuro-imaging has been showing promising results. However, the most recent publication in Nature has created waves in the media.




This collaborative initiative between #Google Health, DeepMind, and various academic programs reports how an AI system is capable of surpassing human radiologists in screening for breast cancer among the population. This was not a randomized control trial but a preliminary assessment model. No doubt it worked. We will talk more about this project in a little bit.


Let's first understand what AI is. The term AI was first conceptionally proposed in the 1950s. The idea initially was to reproduce human intelligence using computers. However, with rapid tech advancement, AI as a term and entity has evolved, encompassing many techniques aiming to train computers into intelligent beings (potentially even more intelligent than humans!). Now, let's familiarize ourselves with some other terms being used alongside AI: #MachineLearning & #DeepLearning. Machine learning is a form of AI, comprising of computational algorithms focusing on data patterns that learn from experience. For example, the recent breast screening imaging scenario is that of machine learning. Review of hundreds of thousands of images interpreted in the past many years was a requirement to build this program, that eventually can interpret really well, better or just like it was trained to be. Deep learning, on the other hand, is the most sophisticated type of AI known to man, a more advanced form of machine learning. Here the predictive strength of computational algorithms, by definition, should not really require experiential learning. It is this Deep learning which, in my humble opinion, be revolutionary in the sense that it will replace humans in many basic-intellect-level jobs. Machine learning, on the other hand, is more likely a high-functioning and efficient partner of physicians in the evolving healthcare.


Now let's go back to the popular manuscript published in #Nature on the topic of breast cancer imaging interpretation. Traditionally, breast cancer screening is recommended for females. So, a large sample is available for machine learning to be able to build algorithms for the computer. Besides, there is a need for improvement based on how the screening system is currently run. False-positive and false-negative results are common in the current healthcare system. In women with dense breasts, the chances of missing a potentially cancerous lesion (false-negative results) can be up to 1 in 5 to 1 in 10 patients. False-positive reads (over-calling a non-cancerous lesion as cancerous), especially in the first mammogram, are even higher. With the availability of enormous size sophisticated and dependable data for machine learning, and since this was an area of need, where the two reasons why this specific project was selected for this initiative. What this study showed was a reduction in false positive and false negative results consistently. In another analysis in this same study, the AI system outperformed radiologists with regards to the accuracy of detecting target lesions.



Other areas where exciting developments are emerging include predictive modeling for prognosis in critical diseases, and as a supplemental tool for diagnostic and therapeutic decision-making. For example, the IBM Watson Health Cognitive Computing System is utilizing machine learning to assist cancer doctors as a decision support system for cancer therapy protocols. #IBM in partnership with #ClevelandClinic is also mining data to help physicians provide personalized medicine. Machine learning is evolving as a cutting-edge supplemental tool for physicians to improve efficiency, accessibility and enhance decisional accuracy within the healthcare system. In the surgical and procedural arena, robotics and 5G technology are making rapid inroads, and the automation assistance is part and parcel of most surgical procedures. Although not purely AI, approximately 5,000 surgical robots were utilized in more than 1 million procedures worldwide in the past year alone. Da Vinci system, Waldo, Cyberknife, motorized laparoscopy, Heartland are all robotic advancements with basic computational precision modeling, however, the true essence of AI in the surgical arena is yet to come to the mainstream. Nonetheless, companies like Google, J&J, IBM, Medtronics along with various other companies in the US and Europe are doing phenomenal work in the AI in the healthcare arena. I'll be surprised if this current decade does not unravel the mind-boggling advancements in healthcare, disrupting how things are done.



In this blog, we have not even discussed the influence of AI on customer care in clinics and hospitals, appointment coordination, the role of AI in nursing, the virtual AI nursing assistance, the role of #Alexa like devices in clinical encounters, EMR (an electronic medical record) as a tool for AI development. Will touch base on other areas as new advancements roll out. Yes, there is enough excitement of disruption, but at the same time, there are many obstacles that exist in healthcare. Applying Machine Learning requires accurate, authentic data that is easily accessible. If you paid attention to the last line, and you are a physician, nurse or medical researcher, you'd know we have a long way to go. As of right now, the EMR data is disappointingly inaccurate, not appropriately updated and due to #HIPAA regulations, not accessible as easily. Different #hospitals use different EMRs, and there is no inter-communication or overlaps, making the compilation of data for Machine Learning a very challenging task. Epic and Cerner and the two largest EMR systems in the US and with a more organized approach, probable collaboration with insurance databanks like CMS and other billing/coding organizations, there is a strong chance that the innovation industry will be successful in creating disruption in the healthcare world, within our lifespan.


In the end, we need to touch base on the unique perspectives and the unknown anticipation, so that as a healthcare community we are aware of what we are getting into. AI has made remarkable advancements in creating genuine & intuitive feelings in response to situations – it still stands in the artificial zone compared to humans. Some artificially created feelings do very well in superficial (day-to-day) circumstances but when it comes to the true sense of reality and responding to that. It may very well let many patients & families down. I would also add my personal perspective in here. If you have ever been tempted by offers such as that from Wall Street Journal where you get 1$ or 2$ for 1-2 months access. Tempting you to subscribe to that monthly deal. And later you realize, it's not worth it and you would rather cancel the subscription? Let's call it you tried to "get out" of that subscription. If you have experienced that you'd know, that the unsubscribing process is made 10x difficult compared to the subscription process on purpose. It requires time and energy to get out of the 'trap'. The intention of the creator of the AI model will define the repercussions, with regards to Machine Learning. Deep Learning will be a totally different ball game. The tech world will bring convenience in our #healthcare, but it comes with its issues.


37 views0 comments
bottom of page