There is so much work going on in the field of Artificial Intelligence and that makes a layperson like me worry about ‘bewildered’ machines. Weird as that may sound to you cool technologists, I am the eternal skeptic, though always in awe of science, constantly looking to know where it takes us, but when it comes to literally stockpiling efficiency like we are training machines to do, my awe for the subject turns to a perpetual state of wonder.
Amongst everything that AI can do and is learning to master, its role in Healthcare truly amazes me with its broad spectrum of application. There’s a lot of potential to improve patient care through AI; studies have proven that algorithms have been seen to outsmart or perform better than pathologists with the significant super efficiency they have shown in performance or accuracy, which is good for patient care. For humans, some of those tasks are very tedious and time-consuming, and it would save time for the specialists, and allow them to focus on more high-level intellectual tasks like, synthesizing diagnostic information rather than, say, looking for that one thing in a petri dish or a glass slide.
Today, AI is used across a broad spectrum, right from diagnosis, aspects of surgery, planning treatment protocols, medication, aftercare, medical signal and image processing, and so on.
The Data – With Healthcare systems having animperative to do something with the data to improve the quality and the value of the care that they provide and the availability of cheap enough computers to actually use those data within specific fields have an enormous opportunity in AI.
AI systems need to be trained with the data that is generated from clinical processes like screening, diagnosis, treatment decisions and protocols for the system to learn about the subject and the outcomes that it would be responsible to generate. The data comes from electronic recordings, medical notes, prescriptions, physical examinations, images, and laboratory notes.
Devices – AI Devices in healthcare largely fall intotwo broad categories – Machine Learning techniques to analyze structured like data from EP, imaging, etc., ML work to provide probabilities of diseases and conditions, outcomes by attempting to cluster patient traits. Natural Language Processing (NLP) is part of the second category that extracts details from unstructured data like medical journals, clinical notes and so on which supplements the structured data. NLP also works to convert texts to structured data that can be understood by machines.
Man vs Machine – Analysing highly complex medical data through ML algorithms, creating logic and arriving at conclusions that must emulate human cognition has given AI its super status in modern medicine. AI can bring screenings and precise diagnostics to less-developed/rural areas where medical professionals are not available. It is fascinating to read, how in Radiology, a deep-learning-based algorithm was developed using more than 50,000 normal chest images and almost 7,000 scans with active TB. The algorithm is reputed to have become so good that in performance tests it easily beat radiologists.
In Dermatology, a deep learning Convolutional Neural Network (CNN) has proven to be more efficient than dermatologists at diagnosing skin cancer. Researchers worked with the algorithm exposing it to over 100,000 images of malignant melanomas and non-malignant moles. The study reports that CNN was better at diagnosing right when compared to the diagnosis by 58 dermatologists from 17 countries. “The CNN missed fewer melanomas, meaning it had a higher sensitivity than the dermatologists, and it misdiagnosed fewer benign moles as malignant melanoma, which means it had a higher specificity; this would result in less unnecessary surgery,” – Professor Haenssle.
In Oncology, AI is expected to crack the code of personalized treatment through ML. They expect to be able to establish an intuitive method of sorting through all the data. They are also using ML to help study how cancers develop, tumours progress, and even working on programming cells to fight cancer and other fatal diseases. So, the study of biological processes as well as diagnosis, treatment, and prevention are being augmented by AI.
Cardiovascular disease is one of the main causes of disability and death, globally. Early detection and treatment are critical for management as well as treatment, and AI can be the game-changer here. AI-based predictions and deep learning can help identify risk factors through retinal images that can be done cheaply and most important non-invasively. Google’s health-tech arm ‘Verily’ has evolved a method using ML, to assess one’s risk of heart disease. Scans of patients’ eyes are analyzed, and the software is reportedly very accurate in deducing data like if the patient is a smoker, the patient’s age, blood pressure, etc., which is used to predict their risk for cardiac events.
In the US, waste in healthcare is estimated to be an astounding $765 billion annually. Routine tests and unneeded tests make up a large part of this, and these tests can be significantly expensive for individuals with no treatment outcomes. AI can help such situations by reducing the number of tests that a patient needs to go through as it runs with entire data of patient information across various healthcare systems to predict based on the individual’s medical history and symptoms. Armed with information like that physicians can reduce/minimize the number of tests, save cost and time. With mobile devices integrated into the hospital workflow, AI-driven decision support provides physicians even minute data that they might have missed. All this helps physicians make a much more informed decision and save their patients’ money and discomfort.
AI, of course, has its limitations too. Studies done in clinical labs prove that algorithms can be precise with very specific tasks while actual healthcare facilities are so much more.
If we really look at what tasks they are asking AI models to perform versus the comprehensive things that an average specialist in medicine, using these models require an additional level of technology and infrastructure and it takes time to learn how to do completely digital diagnosis with an AI model. So, until we can seamlessly incorporate that level of technology into current workflows, it could pose a barrier to the widespread adoption of AI in several areas of Medicine.
And if you think about it, AI does have the potential to increase inequality in healthcare – societies that have access to medical AI versus those that don’t. We already have so many healthcare disparities globally, AI can increase that gap. Or, what if someone comes up with a general adversarial network? All the artifacts and the image acquisitions can be used to cause the model to evolve confident about a wrong diagnosis.
Having said that, AI has revolutionized primary care, in-home care, and we probably would be able to do away with the long-term acute care facilities if we can come up with the right ways to care for people in their homes.
We should also be thinking in terms of training the next generation of physicians who can utilize advanced AI in ways that enable us to deliver better care to our patients. It might call for cross-training, like training physicians in data literacy to be part of medicine’s core curriculum rather than an optional subject.
With everything that AI can do for Medicine today, I still do believe that Medicine is inherently a human enterprise and empathy and caring for another person is not something that an algorithm can reproduce…
- https://www.researchgate.net/ publication/336264780_The_impact_of_artificial_ intelligence_in_medicine_on_the_future_role_of_ the_physician
- https://www.h2o.ai/healthcare/?gclid=EAIaIQobChMI3pnj5Ku65gIVBZWPCh1rOw0XEAAYASAAE gL-YvD_BwE
- https://www.forbes.com/sites/ razvancreanga/2019/03/04/data-governance-ai-and-healthcare-an-exciting-new-world-of-health-provision/#5c3f14b3a44b
Bindu Vijayan is a Sr. Manager at GAVS, a true-blue GAVSian, having spent 8 years with the company. She is an avid reader, loves music, poetry, traveling, yoga and meditation, and admits she is entirely influenced by Kafka’s perspective, “Don’t bend; don’t water it down; don’t try to make it logical; don’t edit your own soul according to the fashion. Rather, follow your most intense obsessions mercilessly.”
Nowadays, the number of software systems used by organizations has increased drastically. Since IT infrastructures comprise of complex software systems, the number of incidents has also increased. An incident response process requires collaboration between various IT teams in an organization.
In the past few years, AI has proved transformational for numerous industries, including the manufacturing and logistics industry. AIOps (Artificial Intelligence for IT Operations) is being used by manufacturing and logistics firms to improve their productivity.
API security is the protection of network-exposed APIs that an organization, both owns and uses. APIs are becoming the preferred method to develop new-age applications. They are one of most common ways to interact between microservices and containers like systems and apps. API are developed using REST or SOAP methods. However, the true strength of API security depends on how there are implemented.