Artificial Intelligence can detect diseases including COVID-19

Artificial Intelligence
Image credit: source

By Sumit Pandey

Taoyuan City (Taiwan), Sep 20 (UNI) Even as the world awaits a COVID vaccine, Artificial intelligence (AI) can be used for detecting pneumonia caused by the pandemic which has claimed nearly a million lives globally.

The dataset commonly used for this work is open source chest X-ray images from Kaggle or other open-source websites. Some of these models have reported an accuracy even greater than 98 percent, experts have said.

The experts while calling for integrating the AI systems into the medical practice, said it would build a mutually-beneficial relationship between AI and Medicine.

In future AI would offer greater efficiency or cost-effectiveness and Doctors (or Medical Staff) would offer AI the essential medical exposure of complex cases.

In this process, it will be necessary to ensure that AI does not hide the human face of medicine because the biggest obstacle to its adoption will be the public’s hesitation to embrace this technology.

According to Prof (Dr.) Andrew Ng, computer scientist from Stanford University, Artificial Intelligence will change all the ‘’paradigms of our life as electricity did around 200 years ago.’’

Medical Prognosis is used for predicting the expected development or likelihood of the disease.

Currently one of the most recent advancements of Artificial Intelligence is in three fields of medicine: Diagnosis, Prognosis, and Treatment. Medical Diagnosis is used for determining the disease or condition that explains the symptoms of the patient.

The core of these Artificial Intelligence models is Machine Learning (ML) or Deep Learning (DL) algorithms. Both (ML and DL) algorithms are called data-driven which means that instead of defining the situation, these algorithms learn from previous examples. Generally, by increasing the amount of the data, the accuracy of these models also increases.

Researchers from Stanford University have developed a Deep Learning based Artificial Intelligence CheXNet that can detect 14 different diseases by analyzing the input front view chest X-ray images.

To train this model, researchers used an open source dataset released by the National Institute of Health (NIH) which contains 112,120 frontal-view X-ray images of 30,805 unique patients, annotated with up to 14 different thoracic pathology labels using Natural Language Processing (NLP) methods on radiology reports.

In 2019, a team of researchers at MIT-CSAIL developed an Artificial Intelligence model that could predict breast cancer in a person five years before its development. Based on the prediction, this model can also develop personalized treatment.

Researchers from Stanford University have developed a Machine Learning-based Artificial Intelligence system that can predict the survival of a person in the next five years. The system takes inputs in the form of body parameters (blood pressure, age, sugar level etc ) and results in the survival rate of a person in the next five years, based on the prognosis report the model recommends the personalized cure to the person.

Researchers have also developed many Artificial Intelligence models for health sectors and if implemented in hospitals, these can help doctors and clinicians in diagnosis, prognosis, and treatment.

However, it was observed that most of these models have limitations and were highly accurate in labs but failed to perform better in real life.

According to a report published in MIT technology review (on April 2020), Artificial Intelligence was accurate in lab testing but the real-life story was different. Google developed a Medical Diagnosis Artificial Intelligence for detecting diabetic retinopathy. To test it in real life Google got permission from Thailand’s Health Ministry.

The model performed with 90 percent accuracy but more than one-fifth of the images were rejected by the Artificial Intelligence.

The problem is related to data generalization and the AI model in the lab was trained and tested on good high-quality images, whereas in real life it was tested on both good and poor quality images.

UNI XC RB 1211

(Excerpt) Read more Here | 2020-09-20 11:14:00

Leave a Reply

Your email address will not be published. Required fields are marked *