Screening w/ AI

Why use AI for screening?

Filed under
Screening w/ AI

Despite being very basic tests, chest X-rays and mammograms are difficult to analyze. Organs, tumors and lesions in the three-dimensional human body overlap into indistinct blurs once compressed into a flat two-dimensional image.

That is why it is very difficult for anyone but an expert radiologist to make an accurate diagnosis, and even a radiologist can fail to identify cancer through human error. The chance to miss cancer on chest X-rays or mammograms can be as high as 30%.¹ ² It basically means that three out of ten cancer patients may not know that they have it even after a test.

AI can solve this limitation.

Artificial intelligence analyzes complicated patterns(correlations) in enormous medical data, and draws conclusions from it.

AI(Artificial Intelligence) is a technology where a deep-learning algorithm, made to emulate the neural network of the human brain, learns, thinks and makes decisions like a human being.

AI far surpasses the human level when it comes to learning capacity and precision.

On average, a radiologist goes through approximately 10,000 X-rays and 2,000 mammograms in a year. On the other hand, AI has gone through a million X-rays and 240,000 mammograms. In other words, AI has already learned as much as a human would in a hundred years.

What if artificial intelligence assists radiologists in reading medical images?

As a result of all this learning, AI has proven to be more accurate compared to humans.

According to papers published in JAMA Oncology, European Respiratory Journal and Lancet Digital Health, AI is better than radiologists at identifying cancer. This is especially true when it comes to cancer that is difficult to identify, whether because it is small and early-stage, or because it is obscured by other organs.³

Another reason that AI is more accurate than humans is its focus. The Go master Lee Sedol said that “the prime advantage of AI over humans is its unwavering focus,” in an interview following his historic game with the AI, AlphaGo. Without any mental or physiological limits, AI can support people with endless consistency.

artificial intelligence increases survival by finding small and subtle early-stage cancers.

Early cancer screening using AI vastly increases the chance of survival.

If lung cancer is left undetected and progresses to stage 3 or 4, the five-year survival rate(the criteria for full recovery from cancer) falls to 18%. However, early AI-based diagnosis of stage 1 or 2 lung cancer can result in a 73% survival rate.

Breast cancer has a higher survival rate than other types. If breast cancer is undetected and reaches stage 3 or 4, the survival rate drops to 65%. However, AI-based early diagnosis of stage 1 or 2 breast cancer results in a survival rate of 96%.

Let’s look at some real-life cases.

what if the patient was diagnosed with lung cancer in 2013 with the assistance of
artificial intelligence?

Above is a chest X-ray of a 54-year-old man. He had chest X-rays taken each year for three years, and he was diagnosed with cancer in the final year. An AI-based analysis of the three test images detected the cancer, even from the first image, two years before the diagnosis.

what if the patient was diagnosed with breast cancer in 2008 with the assistance of
artificial intelligence?

Above is a mammogram of a 59-year-old woman. Once again, an AI-based analysis could find the cancer from the test images prior to the human diagnosis.⁴ This shows that AI can find tiny, early-stage cancer that can slip past human doctors.

What would have happened if AI was used for those diagnoses?

Cancer would have been found early, and cured through simpler methods like surgery.
The patients would have returned to their daily lives in no time. Their mental and physical pain and the financial burden would have been minimal, not to mention the burden on their family.

There we have it. That’s the reason to use AI for cancer screening.

¹ Quekel LG, Kessels AG, Goei R, van Engelshoven JM. Miss rate of lung cancer on the chest radiograph in clinical practice. Chest 1999;115:720-4.
² Forrest JV, Friedman PJ. Radiologic errors in patients with lung cancer. Wes J Med 1981;134:485.
³ Sowon Jang, Hwayoung Song, et al. Deep Learning–based Automatic Detection Algorithm for Reducing Overlooked Lung Cancers on Chest Radiographs. Radiology 2020
⁴ AJCC 8th Edition
Copy Link
chapter 2
Screening w/ AI
Can't find a topic you want?
Please ask us anything.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.