AI is revolutionizing every industry and in particular healthcare. Even though with many AI inventions and a multitude of startups aiming to improve healthcare, we are still not there yet.
What are the different AI use cases in Healthcare?
Artificial Intelligence has found its ground in healthcare with use cases for preliminary diagnosis, assisted surgery, hospital admin activities, nursing assistants, automated image diagnosis to name a few. As per the research by PwC, there are more than 300 use cases for AI to enhance healthcare.
1. Predictive Diagnosis
Deep Learning algorithms have gained traction where it can detect diseases like cancer, thyroid, diabetes, eye disease, etc., But we haven’t reached a point where we can fully trust the AI diagnosis. AI might be predicting diseases with an accuracy rate of 87% when compared to 86% by healthcare professionals, but there is still skepticism in AI diagnosis.
2. Assisted Surgery
Assistive surgery with the help of robots started as early as 1985 when industrial mechanics were used in surgical procedures as part of an experiment. The experimentation proved to attain a strong foothold in the robotic AI handling delicate surgical procedures with much ease and precision.
Robots like Da Vinci, Transenterix, Titan Sport use their robotic hands for surgeries while CyberKnife robot destroys tumors. With the help of the precise movements of surgical robots along with their laparoscopic cameras, the success rates and patient care have improved in recent years.
3. Hospital Admin Activities
Hospital admin activities involve allocating budgets, managing health records, formulating emergency plans, communicating regulations to healthcare professionals and caregivers. A nurse spends 25% of their time in administrative and regulatory activities. There is substantial growth in automating admin activities with machine learning models learning from the experience of healthcare professionals. Robotic Process Automation (RPA) gathers both digital and physical data and route them to the appropriate activities like scheduling appointments, streamlining claims, automating discharge instructions and generating analytics for further treatments.
4. Nursing Assistants
With the shortage of nurses, we are reluctantly looking forward to the robotic nurses reducing the overtime of nurses and thereby preventing any human error due to lethargy.
Robotic nurses are already on the horizon with RoNA, Robotic Nursing Assistants assisting nurses to lift patients, TUG robots transporting medical supplies, AI-enabled chatbots providing less critical medical advice and so on.
5. Automated Image Diagnosis
Deep Learning along with Big Data and readily-available computational power is enriching the healthcare industry with intelligence to detect patterns and diagnose diseases from image processing. Even though the excitement around image-based diagnosis is bustling, its adoption is slow in progress.
What are the hurdles in adopting AI in Healthcare?
1. Data Readiness
The data used for predictive diagnosis might not be fully integrated with all the data sources that healthcare professionals are accustomed to while treating their patients. The lack of unified data sources integrating clinical data, life science, employee benefit along with research in drugs and pharmaceutical advancements prove to be a continuous struggle in embracing healthcare technology.
2. Data Privacy
Data in healthcare are sensitive in nature whether it might be the patient’s medical data, employee’s healthcare provider information, all the wired and IoT devices collecting patient health statistics. It is crucial to securely store and manage all Electronic Health Records (EHR) complying with the Health Insurance Portability and Accountability Act(HIPAA) compliance. As per HIPAA compliance, it is mandatory to authorize medical access, encrypt data, streamline ePHI(Electronic Protected Healthcare Information), audit reports and activities logged in different medical devices. The data privacy rule is trickier to handle and the wide range of regulations ranging from GDPR(General Data Protection Regulation) in EU to CCPA(California Consumer Privacy Act) convolutes technological advancement in healthcare.
3. Lack of Data Pipeline
It is a journey to automation that is steadily improving healthcare but at the same time, it involves proper data management as data is the key. Data enormity along with the lack of streamlined data pipeline complicates the automation procedure. Batch mode operation is no longer sufficient as more and more real-time data are growing with the increase in medical devices collecting our physicals every minute. Big Data technologies like Splunk turning data into doing, Spark managing computation in cluster mode, Kafka transferring data are encouraging technologists to streamline medical data from different data sources.
4. AI being black-box in nature
The lack of AI knowledge among domain experts makes them believe AI to be a black-box miraculously solving all the problems. Spreading awareness and educating the need for AI in healthcare will pave the way for robust healthcare for all.
5. Data Scientists Vs Domain Experts
The fact that training data used in Machine Learning models to predict diseases are prepared by professionals not having knowledge of the medical data used for diagnosis adds to the skepticism of trusting AI. Data Scientists should work very closely with domain experts of the respective field for automating repetitive procedures. Since the data scientists are not the domain experts, any small assumption in building ML models can result in failure lessening the trust in AI.
6. Lack of End to End AI Automation
Building predictive models to diagnose diseases is not a one-time activity, rather it involves continuous improvement with human-in-the-loop intervention and retraining models whenever there is new medical data. There is a need for automating the entire end to end lifecycle of AI projects. We are in dire need of a unified platform for experimenting with different ML libraries and algorithms, deploying to any cloud environment and monitoring the model performance for continuous improvement.
Healthcare for All
Healthcare for all should no longer be a dream but rather a reality with the same standards of healthcare reaching all nooks and corners of the world. Automating admin activities not only helps healthcare professionals to provide personal care but also reduces human-errors and manages the growing cost of medical treatments. By automating the tedious manual jobs, it frees healthcare professionals from the mundane activity and rallies their time and energy to the much-needed patient care.
We are already witnessing the future of robotic surgeries with surgeries performed remotely where the Corindus Vascular Robotics with its CorPath technology was used to place a stent inside the heart blood vessels.
The revolution of Artificial Intelligence along with Virtual Reality is preparing the world to be more interconnected with health resources making the impossible possible.