Artificial Intelligence In Medical Field

Contribution:

1) Sakshi: Introduction and what makes algorithm intelligent.

2) Priyanka: History of AI in Medical, AI: A network tool, AI in the lifecycle of pharmaceutical products.

3) Kunal: Diagnosed diseases.

4) Samruddhi: Develop drugs faster.

5) Anurag: AI in drug screening, Personalized treatment, improve gene editing.

                                                                                           (Image credits: www.forbes.com)

Artificial intelligence (AI) in varying forms and degrees has been accustomed to developing and advancing a good spectrum of fields, like banking and financial markets, education, supply chains, manufacturing, retail and e-commerce, and healthcare. Within the technology industry, AI has been a crucial enabler for several new business innovations. These include web search (e.g., Google), content recommendations (e.g., Netflix), product recommendations (e.g., Amazon), targeted advertising (e.g., Facebook), and autonomous vehicles (e.g., Tesla).

Multidisciplinary proficiency and degrees are used to develop and develop multiple zones, such as banking and finance markets, education, supply chain, manufacturing, marketing and e-commerce, and health care. This includes web search, content recommendations, product suggestions, targeted advertising, and private vehicles.

From the spam mails, we get in our inboxes, to smartwatches that use input from accelerometer sensors to discriminate between common and aerobic activities, to buy products on online shopping websites, like Amazon or Flipkart recommending products based on our previous purchase records.

Smart computer systems are widely used in the medical field. Although computer systems often perform tasks more efficiently than humans, in recent years computerized computer algorithms have gained increasingly enlightened accuracy in the field of medical science. The purpose of this blog is to discuss how artificial intelligence changes the nature of medical science and distinguishes hype from reality.

Collected data generated in clinics and stored in electronic medical records for routine examination and medical imaging allow for increased use of artificial intelligence and more efficient data-driven drugs. Because while these algorithms can have a significant impact on medicine and strengthen the effectiveness of medical interventions, many regulatory concerns need to be addressed first.

What makes an algorithm intelligent?

Just as physicians are trained in medical school years, performing practical tasks and practicing tests, getting marks, and learning from mistakes, AI algorithms must also learn to perform their tasks. In general, tasks that can be performed by AI algorithms are tasks that require human ingenuity to master, such as pattern and speech recognition, image analysis, and decision making. However, people need to clearly tell the computer what they can look at in the image they are giving the algorithm, for example. In short, AI algorithms are ready to perform complex tasks, and sometimes they can surpass people in the tasks they are trained to perform.

To produce an effective AI algorithm, computer systems are first fed standardized data, which means that each data point has a label or annotation that appears in the algorithm. After the algorithm has been presented with insufficient sets of data points and their labels, performance is evaluated to ensure accuracy, just as the tests are given to students. These "tests" of the algorithm usually include test data input where programmers already know the answers, which allows them to test the algorithm's ability to determine the correct answer. Based on test results, the algorithm can be modified, provided with additional data, or extracted to assist in the decision-making of the author of the algorithm.

There are various algorithms that can learn from data. Algorithms then read the data and extract the possible subdivisions. In clinical applications, the effectiveness of the algorithm in diagnostic work is compared with the performance of a physician to determine its efficacy and clinical value.

                                                                                          (Image credits: sitn.hms.harvard.edu)

The above image shows an example of an algorithm that learns the essential anatomy of a hand and may recreate where a missing digit should be. The input may be a sort of hand x-rays, and therefore the output may be a trace of where missing parts of the hand should be. The model, during this case, is that the hand outlines which will be generated and applied to other images. This could leave physicians to ascertain the right place to reconstruct a limb or put a prosthetic.

History of AI in Medical Field 

Artificial intelligence (AI) was first described in 1950; however, several limitations in early models prevented widespread acceptance and application to medicine within the early 2000s, many of those limitations were overcome by the arrival of deep learning. Now that AI systems are capable of analyzing complex algorithms and self-learning, we enter a replacement age in medicine where AI is often applied to clinical practice through risk assessment models, improving diagnostic accuracy and workflow efficiency. AI in medicine even goes back to 1964 with Eliza, the very first chatbot, which was a conversational tool that recreated the conversation between a psychotherapist and a patient. That also was the early days of applying artificial intelligence and rules-based systems to the interaction between patients and their caregivers.

Significant advances have been made in the use of smart artificial intelligence in the event of a patient's diagnosis. developing differentiation models to assist physicians in diagnosing skin cancer, skin lesions, and psoriasis. Their research has shown that AI programs are able to differentiate skin cancer at the level of expertise compared to dermatologists and require only a shorter half of model training time compared to doctors who spend years in medical school and rely on advanced patient experience. diagnosis for decades. A lot of work has also been done in the area of ​​AI and patient prediction. For example, Google researchers developed and trained DCNN using 128,175 retinal fundus images to classify images such as diabetic retinopathy and macular edema in adults with diabetes.

Great strides have been made in implementing AI programs in drug acquisition and in providing personalized treatment options. Intelligent architectural systems are also used in the healthcare sector to improve patient knowledge, patient care, and to help physicians with AI assistants. Today, predictive models can diagnose diseases, predict therapeutic responses, and potentially prevent medicine within the future. AI has the potential to enhance diagnostic accuracy, provider workflow, clinical operations efficiency, disease, and therapeutic monitoring, procedure accuracy, and overall patient outcomes. With AI systems capable of analyzing complex algorithms and self-learning, we enter a replacement era in medicine, where AI is often utilized in clinical practice to enhance diagnostic accuracy and workflow efficiency through risk assessment models.

 (Image credits: news.mit.edu)

AI: A network tool

AI involves several method domains, like reasoning, knowledge representation, solution search, and, among them, a fundamental paradigm of machine learning (ML). ML uses algorithms that will recognize patterns within a group of knowledge. A subfield of the ML is deep learning (DL), which engages artificial neural networks (ANNs). These comprise a group of interconnected sophisticated computing elements, mimicking the transmission of electrical impulses within the human brain. ANNs involve various types, including multilayer perceptron (MLP) networks, recurrent neural networks (RNNs), and convolutional neural networks (CNNs), which utilize either supervised or unsupervised training procedures.

AI in the lifecycle of pharmaceutical products

Involvement of AI within the development of a pharmaceutical product from the bench to the bedside is often imagined as long as it can aid rational drug design assist in decision making, which uses ML algorithms alongside an easy-to-use interface to make analytical roadmaps supported competitors, key stakeholders, and currently held market share to predict key drivers in sales of pharmaceuticals, thus helping marketing executives to allocate resources for max market share gain, reversing poor sales and enabled them to anticipate where to form investments.

                                    

1. Diagnose diseases 

Proper diagnosis can take years of medical training. Even then, diagnosing is often a difficult, time-consuming process. In many areas, the need for specialists exceeds that available. This puts doctors under pressure and often delays a patient's life-saving diagnosis. Machine learning - especially in-depth Learning algorithms - has recently made great strides in diagnosing diseases automatically, making diagnoses cheaper and more accessible.

For diagnosis of coronavirus disease, 2019 (COVID-19), a SARS-CoV-2 virus-specific reverse transcriptase-polymerase chain reaction (RT–PCR) test is routinely used. However, this test can take up to 2 d to complete, serial testing may be required to rule out the possibility of false-negative results and there is currently a shortage of RT–PCR test kits, underscoring the urgent need for alternative methods for rapid and accurate diagnosis of patients with COVID-19. Chest computerized tomography (CT) may be a valuable component within the evaluation of patients with suspected SARS-CoV-2 infection. Nevertheless, CT alone may have limited negative predictive value for ruling out SARS-CoV-2 infection, as some patients may have normal radiological findings at the first stages of the disease. In this study, we used artificial intelligence (AI) algorithms to integrate chest CT findings with clinical symptoms, exposure history, and laboratory testing to rapidly diagnose patients who are positive for COVID-19. Among a total of 905 patients tested by real-time RT–PCR assay and next-generation sequencing RT–PCR, 419 (46.3%) tested positive for SARS-CoV-2. In a test set of 279 patients, the AI system achieved a neighborhood under the curve of 0.92 and had equal sensitivity as compared to a senior thoracic radiologist. The AI system also improved the detection of patients who were positive for COVID-19 via RT–PCR who presented with normal CT scans.

How machines learn to diagnose

Machine learning algorithms can learn to recognize patterns in the same way that doctors perceive them. The main difference is that algorithms require a lot of concrete examples - many thousands - in order to learn. And these examples need to be digitally digitized - machines cannot read between the lines in textbooks.

So Mechanical Learning is especially considerate in areas where the diagnostic details the doctor is examining have already been digitized.

The difference is: the algorithm can reach conclusions in a fraction of a second and can be reproduced economically around the world. Soon everyone, everywhere he would be able to accomplish the same quality of top specialists in radiology diagnosis, and at a lower cost.

More advanced AI diagnostics are coming soon

With a lot of good data available in these cases, algorithms are good at diagnosing as a professional. The application of Machine Learning and Artificial Intelligence in diagnostics is just beginning – more aspiring systems involve the fusion of multiple data sources (CT, MRI, genomics and proteomics, patient data, and even handwritten files) in appraising a disease or its progression.

It is unlikely that AI will replace a doctor directly. Instead, AI programs will be used to highlight potentially injurious lesions or cardiac patterns in a specialist - permitting the physician to focus on interpreting those symptoms.

2. Develop drugs faster  

Drug building is a very expensive process. Many of the analytical processes entangled in drug development can be made more productive by Machine Learning. This has the potential to end years of work and hundreds of millions of investments. The usual method of determining which differences are clinically relevant consists of a manual review of thousands of documents. This approach is extremely time-consuming and unsustainable due to the speed at which the number of healthcare information is growing. To handle the vast amount of knowledge available, the method must be automated, which can require the implementation of AI technologies. Automation will successively release researchers to specialize in identifying many new biomarkers, targets, and medicines instead of spending an inordinate amount of your time validating them through manual literature views.

                                                                         (Image credits: www.pharmaphorum.com)

AI has already been used successfully in all  4 main stages in drug development:

Stage 1: Diagnosis aimed at intervention

Stage 2: Finding people who choose drugs

Stage 3: Accelerate clinical trials

Stage 4: Detecting Biomarker Diagnosis

 Stage 1: Diagnosis aimed at intervention

The first step in drug development is to recognize the biological origin of the sickness and its mechanisms of resistance. The widespread availability of high-performance techniques, such as short-term RNA hairpin testing and in-depth sequencing, has greatly increased the amount of data available to determine the most effective targeted methods. Machine learning algorithms can easily analyze all accessible data and can involuntarily read and target the targeted proteins.

Stage 2: Finding people who choose drugs

Next, you need to find a combination that can interact with the targeted molecule the way you want. These compounds can be natural, synthetic, or bioengineered.

However, the current software is often flawed and produces a lot of negative suggestions - so it takes a lot of time to reduce it to those who choose the best drugs.

Mechanical learning algorithms can help here: They can learn to predict molecular suitability based on fingerprints and cell definitions.

Stage 3: Accelerate clinical trials

It is difficult to find suitable candidates for clinical trials. Mechanical Learning can accelerate the design of clinical trials by automatically identifying suitable candidates and ensuring appropriate distribution of research study groups.

                                                                         (Image credits: www.netscribes.com)

Stage 4: Detecting Biomarker Diagnosis

You can only treat patients for the disease once you are sure of your diagnosis. Biomarkers are molecules found in body fluids that provide complete assurance that a patient is infected. They make the diagnostic process safer and cheaper. You can also use them to signal the progression of the disease - making it easier for doctors to choose the right treatment and to monitor the effectiveness of the medication.

But finding the right Biomarkers for a particular disease is difficult.

Biomarkers can be used to identify:

• The presence of the disease as soon as possible - a diagnostic biomarker

• The risk of the patient becoming infected - risk biomarker

• Possible development of the disease - prognostic biomarker

• Whether the patient will respond to the drug - a predictive biomarker

3. AI in drug screening

The process of discovering and developing a drug can take over a decade and costs US$2.8 billion on average. Even then, nine out of ten therapeutic molecules fail phase II clinical trial clinical trials and regulatory approval. Algorithms, like Nearest-Neighbour classifiers, RF, extreme learning machines, SVMs, and deep neural networks (DNNs), are used for VS-supported synthesis feasibility and may also predict in vivo activity and toxicity. Several biopharmaceutical companies, such as Bayer, Roche, and Pfizer, have teamed up with IT companies to develop a platform for the discovery of therapies in areas such as immuno-oncology and cardiovascular diseases. The aspects of VS to which AI has been applied

4. Personalize treatment 

You can only treat patients for the disease once you are sure of your diagnosis. Other methods are more expensive and include more sophisticated laboratory equipment and expert knowledge - such as genetic sequence.

Biomarkers are molecules found in body fluids that provide complete assurance that a patient is infected. They make the diagnostic process safer and cheaper.

You can also use them to signal the progression of the disease - making it easier for doctors to choose the right treatment and to monitor the effectiveness of the medication. But finding the right Biomarkers for a particular disease is difficult. It is another expensive, time-consuming procedure that involves inspecting tens of thousands of potential human molecules.

AI can do a lot of manual work and speed up the process. Algorithms divide molecules into good and bad people - helping physicians to focus on analyzing the best prospects.

5. Improve gene editing 

Clustered Regularly Interspaced Short Palindromic Repeats, the CRISPR-Cas9 genetic system, is a major step in our ability to effectively estimate DNA costs - and precisely, as a surgeon.

This process relies on short-stranded RNAs to direct and organize a specific area of ​​DNA. But RNA guidance can be equivalent to many areas of DNA - and that can lead to unintended negative consequences. Careful selection of the RNA guide with the least harmful side effects is a major bottle in the implementation of the CRISPR system.

Machine Learning Models have been demonstrated to produce the best results when it comes to predicting your level of both the interlinkage of the guide and the unintended consequences of a particular sgRNA. This can greatly accelerate the development of guideline RNA in all regions of human DNA.

                                                                          (Image credits: www.drugtargetreview.com)

Summary

AI has advanced over several decades to incorporate more complex algorithms that perform similarly to the human brain. Today, predictive models can diagnose diseases, predict therapeutic responses, and potentially prevent medicine within the future. AI has the potential to enhance diagnostic accuracy, provider workflow, clinical operations efficiency, disease, and therapeutic monitoring, procedure accuracy, and overall patient outcomes. With AI systems capable of analyzing complex algorithms and self-learning, we enter a replacement era in medicine, where AI is often utilized in clinical practice to enhance diagnostic accuracy and workflow efficiency through risk assessment models.

AI already helps us to diagnose diseases effectively, develop drugs, treat treatments, and even genetically engineer. But this is just the beginning. When we compile and integrate our medical data, this is where we can use AI to help us identify important patterns - patterns that we can use to make accurate, less expensive decisions in complex analytical processes.

                                                                         (Image credits: www.healthcareweekly.com)

References: 1) www.ncbi.nlm.nih.gov

                         2) https://sitn.hms.harvard.edu/

                         3) https://healthitanalytics.com/

                         4) https://www.frontiersin.org/

Editors: Group 3- Kunal Shivam, Priyanka Lokhande, Sakshi Manmode, Anurag Saraf, and Samruddhi Zaware.