top of page

Doctor AI






Doctors lead a pretty hectic life, and they need to be precise in their job, be it giving accurate test results or providing treatment to their patients. Hence they cannot afford any mistakes as they have to deal with life or death situations many times. There must be a way in providing assistance to doctors to reduce human errors. This is exactly where AI comes into the picture! With the help of Artificial Intelligence, we can develop algorithms which assist healthcare professionals in such tasks, improve diagnosis, and provide better treatment plans.


Support vector machines, decision trees and neural networks are some of the AI techniques which are used for a variety of different diseases.


Clinical Applications of AI


Diagnosis of dermatological diseases

Dermatology is a branch of medicine which relies highly on image interpretation, and image processing using deep learning is an up and coming branch of AI. Image processing naturally becomes useful in dermatology. Today we can detect keratinocyte skin cancer just by face images of patients[1] by Convolution Neural Network, and skin cancer by lesion images using Deep Neural Networks.[2]


Radiology

A well known application of AI is in radiology. Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) are helpful when the demand for human expertise exceeds supply, or where data is too complex to be efficiently interpreted by human readers. Deep learning in MRI focuses on image segmentation and classification of the reconstructed images.



Reducing the spread of infections

Infectious diseases are a risk to society, but with the help of AI, we can reduce the chances of spread of such diseases by rapid diagnosis. Talking about the recent Covid-19 virus, machine learning helped researchers and practitioners to predict the spread of the virus and hence prepare for emergencies. Other applications include detection of malaria by analyzing blood smears using machine learning, where models are trained to identify patterns in the blood, hence allowing the practitioners to treat the patients as soon as possible.


Detection of heart diseases

AI algorithms have been proven to be successful in early diagnosis of cardiovascular diseases. Wearables and smartphones help these algorithms by collecting huge amounts of data from the users and thus improving the accuracy of such algorithms. One of the upcoming techniques include classifying heart sounds[3] and detecting valvular heart diseases at an early stage, where they are easily treatable, using methods like Correlation based Feature Selection technique (CFS) and Support Vector Machine classifier (SVM).

In the latest research, physician-scientists from Smidt Heart Institute have developed an algorithm using AI which can differentiate between two almost indistinguishable heart conditions - hypertrophic cardiomyopathy and cardiac amyloidosis[4]. These conditions are often onerous to identify, but with this new algorithm, it is possible to be more accurate while diagnosing such life-threatening diseases.



Early diagnosis of cancer

In cancer diagnosis time is a very important component, the faster and earlier it is detected, the more are the chances of the patient being cured. AI plays a huge role in speeding up cancer detection. In January 2020, an algorithm was developed and claimed to surpass human experts in breast cancer prediction, using image processing.[5]


On the other hand, AI has large applications in pathological analysis of cells and tissues using several deep learning and neural network techniques. It has been studied that the accuracy of pathologists with assistance of a deep learning program is more as compared to the accuracy of any of them alone.




System applications of AI


Electronic health records

An EHR is an organized collection of patients’ health information stored electronically, which can be shared across different healthcare settings easily. EHRs are an important and reliable source of data for model-building. Natural Language Processing (NPL) can be used to make the data collected more accurate, by reducing the variation in medical terms. For example, certain doctors unlike the others, might enter “heart attack” instead of “myocardial infarction”, even though both the terms mean the same thing. Using NPL we are able to consolidate such differences and make the data stronger.


Telemedicine

Telemedicine refers to virtual medical consultation, i.e. when the patient and the doctor are not physically present for consultation. It is a great way of diagnosis and treatment for patients who are unable to travel or have to be isolated due to some health conditions, or elderly people.

Keeping a track of patients’ health is a concern in long distance treatment, but thanks to Artificial Intelligence it is possible easily today.[6]

There are wearables which can monitor the patient and at the same time report the results to the physician on a regular basis (or even real time) so that they can observe it for any irregularities. The information can be compared with other data that has already been collected using artificial intelligence algorithms used to alert physicians if there are any issues to be aware of. Commonly such devices measure oxygen level, heart rate, blood pressure, sugar level, et cetera, and can be used by patients without any difficulty. Besides these, advanced smartwatches and smartphones also have some of these features, making it easier for people to keep track of their own health and avoid any health problems.



Drug-drug interaction refers to disturbance in a drug's mechanism of action by substances such as food, beverages or other drugs. Advanced NPL techniques can also help to better understand and identify drug-drug interactions.[7]



Conclusion

Even though AI has impressive applications in healthcare today, what we know is just the tip of the iceberg. As we advance into the age of technology, it is believed that there is still a lot that can be done which will change the course of medical research, diagnosis, and treatment as we know it, making it better, faster and more reliable.



Interesting articles regarding a few applications mentioned above:




Written by,

Saachi Gogate, MSc Statistics and Data Science.








15 views0 comments

Comments


bottom of page