Milestones of Research for Artificial Intelligence in Healthcare that Shaped What AI is Today
If we were to guess as novices to identify the time when AI in healthcare was discovered, the money would be on this very decade. But it happens to be half a century years back. John McCarthy named the concept of machine behavior with human cognitive thinking characteristic as artificial intelligence(AI).
An AI project named Dendral was in works during the 60s and 70s. Its function was to analyze hypothesis formation and discovery in science. This was a necessary tool for organic chemists to identify unknown organic molecules, by studying their mass spectra.
This research was taking place at Stanford University by a team of very skilled research associates and students.
With Dendral opening gates with the possibility of more inventions for AI in Healthcare. There are some projects we must know about that pushed the boundaries and evolved artificial intelligence to where it is today.
Medical Diagnostic Decision Support System (‘54–93)
MDDS Predicted and proliferated computerized ECG analysis, automated arterial blood gas interpretation, automated protein electrophoresis reports, and automated differential blood cell counters.
Computer Diagnosis of Primary Bone Tumors (‘63)
GWILYM S. LODWICK & Team researched and gathered the requirements for developing a digital computer program to determine probable histologic diagnosis from roentgenograms include:
a) assurance that the diagnosis can actually be made
b) a method of digitizing roentgen findings
c) determination of probability values for the findings
d) a method assessing the probability relationships between the
findings and the disease (in this study, Bayes’ formula for the probability of causes has been employed)
e) testing and modification of the program
The initial testing was observed on a limited user group, the system was believed to have the potential to predict improve the correct histologic diagnosis in 77.9 percent of the time. Further studies are to be directed toward enlarging the scope of the project to include prediction of behavior and prognosis of tumors from roentgenograms.
Computerized Health Records in Ambulatory Care(2008)
Most U.S. physicians practiced electronic-records systems that were developed by a panel of experts, the study previously indicated that electronic health records were available in the office setting to only a small minority of U.S. physicians. Only 4% of physicians have what the expert panel considered a fully functional electronic-records system.
This study exposed the practice proving efficiency in data collection and access to fast track treatment and diagnosis in ambulatory care, and 49% had no electronic-records system at all.
Categorical and probabilistic reasoning in medical diagnosis (‘78)
Some observations during the study of probability concept were concluded. The physicians and caregivers do not prefer to depend on any numerical computation involving the likelihood of a diagnosis or the prognosis for treatment. Even when the official blessing is bestowed upon Bayesian techniques. Doctors certainly have a strong impression of their confidence in the diagnosis or treatment, but that impression must arise more from recognizing a typical situation or comparing the present case to their credibility. An experienced physician, in his domain of expertise, to give arbitrarily many complex potential explanations for a patient’s condition. In the teaching hospital environment, this serves the useful pedagogical purpose of discouraging pat answers from students. Because so many diagnostic possibilities appear to be available for the expert to consider, we suspect that the rapid generation and equally rapid modification or elimination of many explicit hypotheses plays a significant role in his reasoning. The probabilistic techniques will be best in some well-defined domains, they should not be applied arbitrarily to making other decisions where the development of precise categorical models could lead to significantly better performance. The development and aggregation of a number of different approaches, both categorical and probabilistic, into a coherent program that is well suited to its application area remain a fascinating and difficult challenge. When thinking about the effectiveness of a computerized medical consultant, it is essential to recognize the difference between impressive expert-like and truly expert behavior. A vehement critic of early work in Artificial Intelligence accused the practitioners of this “black art” of trying to reach the moon by climbing the tallest tree at their disposal.
Image Quality Algorithm (2018)
Artificial intelligence (AI)-based algorithms to detect Diabetic Retinopathy from retinal images have been examined in laboratory settings. Recent advances incorporate improved machine learning into these algorithms have led to higher diagnostic accuracy. The image quality algorithm implemented as multiple independent detectors for retinal area validation as well as focus, color balance and exposure, and is used interactively by the operator to detect, in seconds, sufficient image quality for the Diagnostic algorithm to rule out (or in) mtmDR, and thus maximize the number of subjects that can be imaged successfully. As its input, it takes four retinal images, and its output is whether the quality is sufficient and if not, whether this is due to the field of view or image quality.
Modern AI software covers wide-ranging use cases, from cybersecurity to radiographic imaging. As AI applications continue to improve, the entire healthcare industry is undergoing a shift. AI excels at categorizing data, especially once it has been exposed to large amounts of data on the subject. This creates enough potential for AI while it involves medicine — medical imaging analysis and patient medical records, genetics, and more can all be combined to improve diagnostic outcomes.
Moreover, AI tools will use similar information to craft distinctive treatment approaches and supply recommendations to doctors.
Robotic surgeries allow surgeons to use smaller tools and make more precise incisions. Surgeons (and patients) could also benefit from AI by combining medical records with real-time data during operations, as well as drawing on data from previous successful surgeries of the same type. Accenture, a technology consulting firm, estimates that AI-enabled, robot-assisted surgery could save the U.S. healthcare industry $40 billion annually by 2026.
Think of virtual nursing assistants like an Alexa for your hospital bedside. These virtual assistants replicate the typical behavior of a nurse by assisting patients with their daily routines, reminding them to take medications or go to appointments, helping answer medical questions and more.
