Can Artificial Intelligence Replace Medical Surgeons?
How close are we towards an Advanced Futuristic World?
Artificial Intelligence (AI) is now expanding beyond its primary field of application in major industrial sectors and other high-tech financial and cybersecurity companies. Their current target is the healthcare industry.
According to data from statista.com, the revenue of the US healthcare industry in 2019 was $2487 billion. Startups and Venture Capitalist (VC) are realizing the potential of AI and Machine Learning (ML) solutions in the healthcare industry as it offers the solution of increasing efficiency and reducing the cost to the patients. According to the American Society of Plastic Surgeons (ASPS), Americans spent over $16.5 billion in cosmetic plastic surgery in 2018. In general, the cost of surgeon and the procedures increased a little over 4 % since the previous year.
Corrective surgeries also called refractive eye surgery is used for vision correction. AI is used in changing traditional approaches to cataract surgery. Previously, surgeons had to rely on static images of patient eyes and other general measurements for removing cataract. This provides scope of including AI technology to improve the efficiency and complications associated with surgery. Now, AI can be used to compare against previous data to determine the patient corneal strength and the appropriate lens. This will simultaneously reduce cost associated with revision surgery and improve the patient well being.
Optiwave Refractive Analysis (ORA) can use AI and real time measurements can enable surgeons to accurately predict the position of the intraocular lens (IOL). ORA performs instantaneous differential calculation to measure the refractive power of the eye that is corrected by IOL.
Around 1950, Alan Turing proposed a robot-assisted clinical procedure. He suggested that human decisions can be copied by machines.
Artificial Intelligence in the medical field was previously limited to data analysis and diagnostic purposes. They were usually used by physicians to compliment the decision making process for patient’s disease diagnosis and treatment procedures. This maximized reliability and efficiency encouraged surgeons to opt for the technology for robot-assisted surgery that have cognitive AI or ML capabilities.
By incorporating Graphical Processing Unit (GPU) for image analysis, AI is becoming popular in the field of ophthalmology for image analysis. Data analysis from previous records or practices is compared for prediction of the surgery.AI can improve image quality, categorization of relevant data required for surgery and can effectively compliment the work of the surgeon.
Natural Language Processing (NLP) is another AI technology that has the applications in the field of surgery. Since chart review is time consuming and laborious, NLP can extract critical information from electronic health records (EHR). Cody et al. designed a proof of concept using NLP to identify operative notes required for total hip arthroplasty (THA). The NLP algorithm was able to extract EHR data with an accuracy of almost 99% . It confirms the validity and potential application of this technology in identifying data elements from operative notes.
Implemening NLP into the field of surgery requires a series of steps. The first approach would be to translate the raw data to machine readable language. EPIC, AthenaHealth, GE Centricity are electronic health vendors (EHR) that focuses on developing software and cloud based capabilities for hospitals and other academic medical centers. Unstructured “raw” medical data produce challenge in both storage and performing analytics. NLP has the ability to produce a result-orientated structured data that can allow analysis and complement the work of surgeons. At the same time, it can eliminate potential human error, thereby reducing processing time and improving efficiency.
NLP uses AI technology to process unstructured useful data and allow speech-recognition capabilities. It has three tasks: (i) classify information (ii) extract relevant data, and (iii) summarize it into a machine- readable format.
In natural language processing, critical patient information can include: (i) social and demographic history (ii) existing medical condition (iii) blood type, etc. The primary interface is divided into sections associated with a keyword for instance: COVID-19 : Fever, Chills, Trouble breathing. This interfacing can allow easier diagnosis and help the role of physicians.
Amy Jin, worked with mentors of Stanford University to develop a software program that can measure a surgeon’s technical skills. The purpose of the software is to observe the video operations performed by the surgeons and track the movement and timing of the instruments.
Surgical robots can accurately determine the depth and speed of movements during any procedure. Surgeons often have a challenging repetitive tasks and AI integration ensures precision and avoid any unintentional accidents due to prolonged hours of motor muscle fatigue. Deep learning (DL) can be used to determine surgical patterns collected from data on observing surgeons in real-time and offer advice on improving accuracy.
Verb Surgical is formed by a partnership between Verily (from Google) and Ethicon Endo Surgery ( Johnson and Johnson). It aims to integrate data analysis and ML to make surgeons more proficient and improve clinical outcomes with cost-effective solution. Dave Herrmann, the Senior Vice President of Verb Surgical, described the startup as a tool used by surgeons to prepare, perform and follow-up on cases using a single platform. It will also allow to share procedural data and learn from the best practices globally. The digital surgery platform aims to transform technology into actionable information.
The University of California at Berkley is working on developing an algorithm that allows robots to perform suturing ( a process which involves sewing of an open wound). The initial model were developed for laparoscopy surgery but the success of the model will allow adaptation across various other surgical robotic application. Initial study showed a success of 87 % but the rate dropped drastically with increase in surgical complication. The wide-scope adoption of AI technology in this process will hep reduce both surgical complications and mortality rate.
The Advanced Robotics and Control Lab at the University of California, San Diego is using Machine Learning to design biomedical robots for context-aware robots. They are investigating control strategies for surgical robots using AI and motion planning algorithm. This will increase precision, control, dexterity and allow robots to behave appropriatey in a realistic surgical environment.
Similar research related to suturing is performed at John Hopkins University. The project, STAR ( Smart Tissue Autonomous Robot) had shown comparable success to standard surgical performance. However, extended research is required before the platform in adopted in surgical hospitals.
Conclusion
AI technology is healthcare is constantly being developed to integrate some of the repetitive tasks of the surgeons. It will eventually allow surgeons to handle the complex challenges of surgery and not focus much on the repetitive task. However, the role of AI is to complement the work of the surgeon, not to eliminate its role, because there are challenges in limiting its procedures to algorithm.
The role of NLP (Natural Language Processing) will convert raw medical data into machine readable language. It will allow scope for AI to perform data analytics to get a wider perspective of underlying issues. There has been some progress (both in industries and academia) in developing complex algorithm to replicate the movement and timing of the instruments used by surgeons. Though there has been notable success for repetitive task but it still possess unique challenges for complex novel medical cases.
References
[1]https://www.healthcatalyst.com/insights/how-healthcare-nlp-taps-unstructured-datas-potential
[4]https://www.verbsurgical.com/physicians/
[5]https://emerj.com/ai-sector-overviews/machine-learning-in-surgical-robotics-4-applications/
