Artificial Intelligence in Medicine
Build-up to AI
The first use of the word robot comes from the Czech word robota, a term used in Karel Capek’s 1921 play Rossum’s Universal Robots, to signify a factory wherein biosynthetic machines were used as slave labour. The famous science fiction pioneer, Isaac Asimov, popularized the term robot during the 1930s and 40s. His Robot series features 38 short stories and 5 novels which feature conscious, aka positronic, robots. Contemporary to his time, a particle called a positron was newly discovered, and inspired future works of sci-fi such as Doctor Who, I, Robot, and Star Trek to reflect on the possibilities of new materials in the construction of intelligent machines. If we leave the Western world behind, and attend to the feats of earlier engineers and scientists, we can find the first trace of humanoid automatons in the third century CE in China. The first automaton, constructed by Yan Shi for Emperor Mu, was crafted from leather, wood, and artificial metal organs. Another instance comes from the 12th century, when mechanical engineer al-Jazari manufactured an android device capable of striking cymbals. While these early pioneers left few blueprints, their work ripples through the sketches of Renaissance inventor, Leonardo da Vinci. Of his sketches, we have an extensive collection.
Leonardo’s robots, particularly the Knight of Milan, were capable of independent movement via a system of pulleys, levers, wheels, and cables which mimicked the anatomical knowledge da Vinci had gathered through regular attendance at dissections. He performed upwards of 30 in his lifetime. His extant sketches and writings would have changed the course of medicine, had they been published during his lifetime. His Anatomical Manuscript, filled with 240 drawings and 13,000 words of accompanying notes informed all of his robotic inventions. The Knight of Milan graced the Duke’s court with its humanoid movements – lifting its visor, standing, sitting, waving, and movings its head and jaw.
In recognition of his work, a surgical system devised by Intuitive Surgical was named Da Vinci in the early 2000s. Da Vinci surgical systems enable surgeons to operate on patients via a console in minimally invasive procedures.
After Da Vinci, French inventor, Jacques de Vaucanson, crafted an automaton able to play a musical pipe. Included in its repertoire were 12 songs. After spending a childhood in the care of Jesuit priests, de Vaucanson struck out on his own in Paris and soon found financial backers and an avenue for the study of anatomy. His most famous creation, The Automaton Flute Player, was first exhibited in 1738 to groups of 10 to 15 at a cost of an average week’s wages. The flute was and still is considered one of the most difficult instruments to play in tune since producing a sound requires perfect harmony between fingerings, breaths, the shapings of the lips, and the volume of air expelled with each breath. Vaucanson succeeded by mimicking by mechanism all the corresponding muscles and motions of a human body – in the same manner as his predecessor Da Vinci.
Fast forwarding to the past century, inventors and scientists switched their focus from the mechanics of muscles to the behaviour of the brain and cells therein. William Gray Walter deserves credit for creating the first electronic autonomous robot in 1948 which he called Machina Speculatrix. The name is a nod to the latin verb speculare – to explore. He built many of these turtle-like machines, three-wheeled contraptions, equipped with a light sensor, touch sensor, propulsion motor, steering motor, and vacuum tube analog computers. Walter’s machines were able to navigate around obstacles and detect light and darkness. His awkward turtle-like robots became the ancestors of the modern Roomba. Walter hoped to demonstrate that connections between even a small number of brain cells could produce unexpected and complex behaviours.
The disparate fields and ideas which would convalesce into AI took off in the 1950s. John McCarthy coined the term “artificial intelligence” in 1955 and founded the field of AI research with a set of highly gifted colleagues at Dartmouth College. As computers became capable of increasingly complex problem solving, the US Department of Defense took interest and began to seriously funnel funds into new avenues of interdisciplinary research. The public became familiar with AI through the much publicized chess competition between world champion Gary Kasparov and the computer Deep Blue in 1997.
The disciplines which harness and develop AI owe much to Norbert Weiner and the origin of cybernetics. He refined the idea of feedback which had ripples throughout engineering systems controls, biology, neuroscience, computer science, systems theory, sociology, psychology, and philosophy.
Artificial Intelligence has gained credibility and is increasingly used in medical applications. These applications will be outlined below.
The virtual component of AI depends on Deep Learning or Machine Learning which is the idea that mathematical algorithms can be improved through trial and error. Algorithms can learn independently or under supervision
- Unsupervised (ability to find patterns)
- Supervised (classification and prediction algorithms based on previous examples.
- Reinforcement Learning (use of sequences of rewards and punishments to form a strategy for operation in a specific problem space)
Pattern finding is essential in uncovering protein-protein interactions. Understanding these interactions and the conditions in which they occur allow scientists to develop novel therapies. Researchers were able to predict upwards of 5000 protein complexes, out of which, over two-thirds were “enriched by at least one gene ontology function term”.
DNA variants like single nucleotide polymorphisms (SNPs) which may be indicative of diseases or traits use evolutionary embedded algorithms such as Markov clustering.
Another application is “systems thinking” which aggregates data and finds correlations between health care records, large scale organizations, and cycles. Systems thinking seeks to turn entire health care systems into entities capable of self-learning and self-improvements. A key example of this is a multi-agent system (MAS) approach to treating chronic mental diseases. Such a system tries to understand how multiple factors interact with each other to produce the conditions which lead to mental diseases.The system tries to incorporate the dynamics of individual patients (responses to medication and behavioural interactions) within the greater societal ecosystem.
MAS could help governments focus on actionable health data, rather than on plain data-sets, devoid of behaviour. Data on mere costs is unactionable. Coordination between health care and other city systems such as criminal justice and education allows for better process mapping, controls, and changes to decrease overall costs and increase positive patient responses to medication.
Another virtual application stem from the digitization of electronic medical records which provide algorithms with a wealth of data from which to detect patterns and correlations. These correlations arm patients and physicians with actionable data. Patients may be able to avoid certain lifestyle choices which contribute to the development and progression of diseases. Digitization enables medical offices, clinics, laboratories, and hospitals to pool patient information into one repository. It is critical that more offices jump on the digital bandwagon so as to eliminate lost links within the chain of information.
The scope of action is macro as well as micro. Patient data stripped of identifying factors is critical for epidemiological research and planning. Tracking outbreaks becomes a far more manageable task when epidemiologists have access to simple and readable data sets.
In the field of psychotherapy, softbots – emotionally sensitive and teachable avatars – are changing outcomes in child medicine. They are able to detect pain levels in children with cancer and detect the first signs of emotional disturbances
The physical branch of AI involves medical devices and objects which aid in the care of patients. These carebots may deliver medication in hospitals, as is is the case in the University of Nagasaki hospital, aid those with mobility issues, or provide positive cognitive companionship.
Robots are also used to assist in surgeries or may perform surgeries completely unassisted by a human surgeon. In many surgeries, robotics have become the preferred method. Due to the precision and smaller incisions, robotic procedures tend to be less invasive leading to decreased blood loss, less pain, and quicker healing time. The burden on patient and the healthcare system is alleviated when patients spend fewer days recovering in beds. The Da Vinci Surgical System has been lauded by Medicare as converting many ‘large incision’ procedures into minimally invasive procedures.
Artificial Intelligence is bringing us closer to a future in which all medicine may be preventative. Through pattern recognition, trends in healthcare will be recognizable in infant stages both on the micro and macro levels. For more information on the use of robots in Nagoya’s Hospital, please see: Smart Hospitals.