Artificial Intelligence in Medical Field
For the past 2 years, the world has been in an era of pandemics because of Covid-19. Although now conditions have become better, still new variants of Covid are rising. Doctors are working 24x7 to tackle this problem. Besides Covid, there are other diseases for which doctors are needed, but can there be a solution, which can make the task of doctors easy? Yes, there is Artificial Intelligence. But what is AI ? and how can it be used in the medical field?
AI is a wide-ranging branch of computer science concerned with building smart machines capable of performing tasks that typically require human intelligence. The medical sciences make substantial use of artificially intelligent computer systems. Common uses include remote patient treatment, prescription transcription, increasing doctor-patient communication, medication discovery and development from beginning to end, and patient diagnosis.
Background Research
According to a survey conducted by Deloitte of 1,100 US companies that were using Artificial Intelligence. With AI, you can enhance the organizational aspects of your industry. Regular US nurses spend 25% of their working hours on regulatory and administrative work. This technology makes it easy to take over these routine tasks such as billing, revenue cycle management, clinical documents, and records management.
In another Harvard Business Review study, more than 300 medical and clinical executives said they had problems with patient involvement. More than 70% of the surveyed individuals said that less than half of the patients were deeply involved in the treatment process, and 42% of the surveyed subjects said that less than a quarter of the patients were highly involved. Increasing patient involvement will undoubtedly lead to better patient health outcomes. Machine learning can provide automated messaging alerts and relevant targeted content that triggers actions at critical moments. In general, there are many ways AI can personalize and improve the treatment process.
One of the best uses of artificial intelligence for medical treatment is a bot system that greatly simplifies treatment time. The virtual patient-nurse acts as a voice-controlled health assistant that provides information about many illnesses, health problems, and medications. Data engineers work on solutions for all medical activities that cover not only general health monitoring but also treatment and prevention of illness.
The existence of an artificial intelligence model has a number of benefits, including:
- Automated diabetic retinopathy grading, which improves efficiency and speeds up patient diagnosis.
- Providing second opinions for optometrists.
- Early detection of diabetic retinopathy thanks to the model’s ability to analyze images at a granular level, which is something a human opthalmologist is unable to perform.
- A broad range of screening initiatives that lower entry obstacles.
In the healthcare industry, artificial intelligence technologies are also being used to improve patient care, and the patient experience, and offer assistance to doctors through the employment of AI assistants. Companies like BotMD have developed systems that can assist with clinically related problems relating to:
- The AI system can quickly determine which doctors are on call and book the next available appointment; it can also scan several scheduling systems across various institutions.
- Answering inquiries about prescriptions, such as the availability of medications and affordable alternatives.
- Using a smartphone application to help clinicians find hospital policies, a list of available clinical tools and available medications can improve hospital workflow.
Applications of AI in the medical field
Applications of AI in medicine are endless and cannot be covered in a single blog. Here are a few applications mentioned below
- Diagnosis via Image analysis
Here are two examples of current applications of precise and therapeutically useful algorithms that can help patients and doctors by simplifying diagnosis.
The first of these algorithms is just one of several examples of existing algorithms that perform better than doctors in image categorization jobs. Researchers at Seoul National University Hospital and College of Medicine created the AI system DLAD (Deep Learning-based Automatic Detection) in the fall of 2018 to analyze chest radiographs and find aberrant cell development, such as possible malignancies. On the same set of photos, the system outperformed 17 out of 18 doctors when its performance was compared to that of different physicians’ detecting skills.
The second of these algorithms was developed by Google AI Healthcare researchers in the fall of 2018, who developed LYNA (Lymph Node Assistant), a learning system that examined histology slides stained tissue samples to recognize metastatic breast cancer tumors from lymph node biopsies. Although this is not the first application of AI to undertake histological analysis, it is important to note that this system was able to identify problem spots in the provided biopsy samples that were indistinguishable from the human eye. On two datasets, LYNA was put to the test, and it successfully identified samples as malignant or noncancerous 97% of the time. Additionally, LYNA cut the customary slide viewing time in half when used in conjunction with doctors’ regular analysis of stained tissue samples.
- Prediction of hotspots of Covid-19 using Contact Tracing
Government agencies utilize contact tracing as a disease prevention strategy to stop the spread of disease. Contact tracing works by getting in touch with, educating, and ordering those who have been exposed to a person who has contracted the disease to quarantine themselves in order to stop the sickness from spreading. According to Apple Newsroom, tech goliaths Google and Apple have teamed up to develop a platform for contact tracing that would employ artificial intelligence through the use of application programming interfaces, or APIs, as they are known on mobile devices. Users who wish to enlist on the platform will be able to report their lab findings. The platform will then be able to reach everyone who may have been close to the infected person thanks to location services.
- Drug development and discovery
Deep learning in medicine can produce new chemical structures and speed up the drug discovery process. It has been utilized in conjunction with other ML-based methods to evaluate biological activity, absorption, distribution, metabolism, and excretion (ADME) features to choose molecules with advantageous physicochemical and biological characteristics.
The druGAN program is a nice illustration of this application of ML in healthcare. On the basis of predetermined anticancer drug qualities, it aims to produce new molecular fingerprints and drug designs that incorporate necessary attributes. It has already demonstrated a noticeable advancement in creating novel medication designs with particular features.
- Medical Record Management
Any doctor’s time is heavily consumed with record keeping. However, ML in healthcare can address this problem. It can relieve doctors of numerous repetitive activities along with natural language processing (NLP), another branch of AI. For example, NLP algorithms can convert human speech during a patient visit into text, eliminating the need for clinicians to manually enter clinical notes. Additionally, by unlocking valuable unstructured data from EHR, NLP and optical character recognition (OCR) algorithms can enable doctors to utilize this data for analytics and decision-making. Additionally, these algorithms can organize and sort clinical documentation, making it better suited for machine learning.
The Dutch business MedInReal, which offers a virtual care assistant for doctors based on AI, is a fantastic example. They are able to update EHRs using NLP capabilities and automate tedious procedures. It uses machine learning to identify the components of structured data and check that they correspond to medical terms. Another illustration is Google’s Cloud Vision API, which already employs handwriting recognition technology to organize data in electronic health records.
- Mental health trends tracking
Learning and predicting mental health concerns globally or among particular demographic groups is one of the applications of machine learning in healthcare. The demographic groups that are more susceptible to stressors like pandemics or natural disasters may be identified by mental healthcare specialists using the results of this investigation.
For instance, by examining the language people used to convey their worries online, MIT and Harvard University researchers utilized machine learning to gauge the consequences of the global pandemic on mental health. They discovered that subjects related to suicidality and loneliness had nearly doubled after their machine learning algorithm examined 800,000 Reddit messages. The results may make it easier for psychiatrists to spot and assist those whose mental health is in need.
- Robotic Surgery
It’s still too early to discuss robots doing all surgical procedures, but they can help doctors a lot when it comes to manipulating surgical instruments and carrying out certain jobs. Suturing automation, surgical skill evaluation, and workflow modeling improvements have all been accomplished with the use of machine learning.
For instance, the smart tissue autonomous robot (STAR) from Johns Hopkins University has already proven that it can do surgical tasks like suturing and knot-tying better than human surgeons.
Companies Using AI in Medical Sciences
Limitations of Artificial Intelligence in Medical Science
- Creating biased models
The data is placed aside for testing while a subset of the obtained data is used to train artificially intelligent systems (also known as training data set) (also known as testing data set). Therefore, the final model will be biased if the data is skewed, that is if it targets a particular race, gender, or age group. Therefore, the population for which the data’s use is intended must accurately reflect the data acquired data.
- Data Preprocessing
It is still feasible to develop a biassed model even after objective data have been gathered. Before the data can be utilized to train an algorithm, it must be preprocessed. The manually entered data or a number of other factors can lead to inaccuracies in the raw data that has been gathered. These entries are occasionally changed mathematically justified or omitted entirely. Data preparation must be carefully avoided to prevent the creation of a biassed data set.
- Fragmented Data
The inability to seamlessly transition models (such as regression, classification, clustering, and NLP) that one organization spends time and effort designing and deploying for a specific task to another organization for immediate use without recalibration is another limitation of the application of AI. Data exchange between healthcare companies is frequently unavailable or constrained due to privacy concerns, which leads to fragmented data that reduces the accuracy of a model.
- Blackboxes
Due to the complexity of the underlying mathematical processes, artificial intelligence systems have a reputation for being “black boxes.” It is necessary to improve the usability and interpretability of models. Despite recent advances in this area, there is still more work to be done.
Conclusion
Although we are still a long way from making AI a reality, it has the potential to help with many of healthcare’s major issues. Data is a significant issue and impediment to making this a reality. Without sufficient and well-represented data, we cannot fully utilize AI in healthcare, no matter how innovative the technology and machine learning algorithms become. The healthcare sector must digitize medical records, come to an agreement on the standardization of the data architecture, and develop a foolproof mechanism to handle patient data consent and secure patient anonymity. It would be difficult to realize the true promise of AI to improve human health without these significant reforms and collaboration in the healthcare sector.
Thank you for reading this post till the end. If you like this post clap, comment, and share with your friends.
References
- https://demigos.com/blog-post/machine-learning-in-the-medical-field-cases-challenges/
- https://www.foreseemed.com/blog/machine-learning-in-healthcare
- https://www.flatworldsolutions.com/healthcare/articles/top-10-applications-of-machine-learning-in-healthcare.php
- https://sitn.hms.harvard.edu/flash/2019/artificial-intelligence-in-medicine-applications-implications-and-limitations/
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6691444/
- https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7640807/#:~:text=Artificially%20intelligent%20computer%20systems%20are,prescriptions%2C%20and%20remotely%20treating%20patients.
