Revolutionizing Modern Healthcare
The United States of America allocates a large part of its Federal budget to its healthcare system. Mandatory spending in 2019 included $1.2 trillion on primary health care programs. The Organization for Economic Co-operation and Development (OECD) health care and expenditure report 2019, the US had spent 16.9% of its Gross Domestic Product (GDP) on the national healthcare system, highest among OECD's 36 members. In the same year, an estimated 30 million Americans were uninsured. Expensive healthcare services make healthcare inaccessible to many Americans.
What can be changed?
Healthcare now needs smarter investment, where research on how existing human knowledge of medicine can be combined with technology, requires more resources. Artificial intelligence (AI) can be the answer. The targeted use of AI and Machine Learning (ML) can cut down the expenditure by a significant amount by reducing human efforts and bringing automation. Cutting the cost is the top priority of the US healthcare system. The United States is well aware of it, and hence the President's Financial Year 2021 budget includes a significant increase in non-defense AI Research and Development (R&D). It also states there that the National Institutes for Health will invest $50 million for new research on chronic diseases using AI and ML.
What roles can AI play here?
AI can play an essential role in the healthcare offerings of the future. Many features of patient care, fields of pharmaceutical organizations, administration of healthcare service providers, etc., can successfully use AI.
Genetic algorithms are capable of carrying out actions performed by humans, primarily when repetitive and trained movements are involved. Researches have shown that AI can diagnose diseases better, if not as accurately as humans. Algorithms, by far, have been able to diagnose malignant cancer more successfully than radiologists. Machine learning is already being used by healthcare, like predicting what treatment protocols will work best for the patient based on various patient's attributes and treatment context. It has also helped in predicting a particular disease that a patient might acquire.
Healthcare is quite susceptible to human errors, and AI tools have helped in preventing many by now. More focused development of ML in fields like keeping and maintaining all past histories of patients will help avoid the occurrence of such errors. Surgical robots, for example, have provided superpower to surgeons in head and neck surgeries, gynecologic surgeries, prostate surgeries, etc. These robots have improved the surgeon's ability to see, create precise and minimally invasive incisions, and stitch wounds. Many researchers have sought solutions to automate daily actions within medicine, like diagnosing patients and prescribing medications.
But AI and ML have their limitations:
High potential legal costs related to any errors in medicine have slowed the adoption of innovation in many cases. What is needed is more accurate, controlled, and calculated research and development.
Research done by the National Center for Biotechnology Information (NCBI) shows that the application of machine learning in medicine is a premise with a promising future. The numbers point to a 95% improvement in the area if all studies are successful and implemented. Also, several studies performed at universities such as MIT, suggest the creation of algorithms aimed mainly at the detection of diseases. Such algorithms may remain active along with enormous improvement capacity. The demand for machine learning algorithms has been increasingly high, which is why different forms and different algorithms have emerged.
SVM- The most popular algorithm in medicine:
Support Vector Machine (SVM) has been one of the most widely used algorithms in medicine for a long time to solve data issues with classification and regression, using repetition as input data. Studies have shown that SVM can help in distinguishing early morphological changes in the mind and can deduce significant initial findings for dementia with the help of MRI and other AI tools. Similarly, a research journal has also published that SVM can also classify the right clustered occurrence of the early phase of Diabetes.
The Support Vector Machine works as follows:
Each data is plotted as an n point in n-dimensional, n being a coordinate in dimensional space.
So, the classification is done by finding the hyperplane that differentiates the two classes very well. With the execution of the algorithm and study of the data generated on the hyperplane, SVM can detect the diseases themselves. This algorithm has been very successful in creating medicines and catching diseases.
Zilic- the promising 2020 algorithm:
Zilic helps in the early detection of diseases in a single test through medical images. Zilic uses Generative Adversarial Network (GAN) and autoencoder. These tools, for example, are trained only with healthy images of lungs. When the systems get an unhealthy lung's picture, the autoencoder reconstructs the image, where the diseased part looks different from the original one. GAN, using discriminator, gives slightly higher scores to the image of the lungs with diseases. And Zilic can highlight the affected area of the lungs and hence can detect the rarest disease of lungs in X-ray images.
Zilic has inspiration in GAN using two neural networks that are trained against each other. The first network is the Generator, which works as an algorithm trainer, that is, sending false images that are exactly like the original. The second neural network is the one who identifies whether the images generated by the first are real, identifying patterns.
Experts use this algorithm to identify rare diseases.
Artificial Intelligence's future in healthcare:
In Atrial Septal Defect (ASD), the algorithm has been very successful. With testing on 1000 children, it has 95% effectiveness. Many scientists see the potential for adaptation and development for this algorithm since it has had a high success rate.
But in cases of algorithms for Alzheimer's, the rate of effectiveness has been 50%, which is quite low. It is not very easy to say that it is an excellent tool, as it is in a tie. However, this does not prevent the development of the algorithm. There are still many improvements to be made in the algorithms. However, there is much space for its development. We can and should invest in algorithmic-based healthcare businesses, as this is a branch that has brought many results to the population.
There is no doubt that AI and ML perform better than humans in processing the vast amount of data of drugs, patients, diseases, historical data, and predicting the results. DNA and RNA analysis, for example, of various types of microbes, viruses, and bacteria, are helping researchers in gaining valuable insights into their biological behaviors. Biologists and pharmacologists in big companies often use AI and ML in their routine drug research and testing activities.
The use of AI and ML has helped in reducing the time and effort taken in bringing new medicines and vaccines for some of the epidemics like Ebola. Researchers are making use of available AI and ML techniques for finding the vaccine for the current Covid19 pandemic as well.
We have relied on computer scientists and health experts to develop these points further. And we firmly believe that entrepreneurs who invest in this field will lead to innovation and gain great popularity and financial return. By far, it's clear that AI and ML are not going to replace medical professionals but instead will enhance their efforts in patients' care.