Make no mistake, algorithms are everywhere. From dating websites to policing and credit checks, these well-defined sets of computer-implementable instructions form an ubiquitous part of our everyday lives.
Every time you Google search something, AI is involved. In fact, the tech giant’s ever-changing algorithm for ranking is so important to modern business that it has led to the creation of the Skillshop—a provider of free online courses designed to help digitally-driven learners crack Google’s coveted code.
It’s no secret that the influence of computer algorithms is somewhat controversial. Some people argue that they put too much power into the hands of big corporations and perpetuate bias. But algorithms also promote positive change, such as that seen in cancer research. Could its new use in tumor detection change the way we think about AI?
This brief insight into the recent advances in cancer diagnosis using AI explains exactly how early detection could become the new norm, and how these algorithms could revolutionize the future of healthcare.
Are screening methods reliable?
When it comes to cancer, catching it in its early stages is key. Not only does it improve an individual’s chances of survival and full recovery, but it can also reduce the need for invasive surgery and chemotherapy.
In many countries, screening programs are already in place to help catch cancer early among risk groups. However, it’s reported that the reading of these images can produce false negatives, with physicians failing to detect the presence of cancerous cells or tumors.
Since doctors are forced to rely on the naked eye alone, it’s easy to understand how mistakes can be made.
Can AI boost detection rates?
When it comes to breast cancer, around one in every five cases are missed by screening mammograms. However, a recent AI tool developed by Google is thought to be able to improve accuracy when it comes to detection.
This AI—which is designed specifically to analyze X-ray images in breast cancer patients—cuts false negatives by almost 10% in US patients. By being trained to detect issues that may go undetected by the human eye, the algorithm is thought to be able to produce a more accurate diagnosis.
Developed by Assad Oberai, this breakthrough discovery involved training a machine to interpret real-world data by using synthetic data. A total of 12,000 synthetic images were used to train the machine-learning algorithm in much the same way as photo identification software.
The tool produced by Oberai and his team is thought to be able to distinguish between the features of benign and malignant tumors with 80% accuracy. It’s now hoped that by refining the computer algorithm using real-world images, 100% accuracy can be achieved in the not too distant future.
What about physicians?
The rise of algorithms in healthcare is hugely advantageous. It’s proven that in certain instances, these AI tools even outperform healthcare professionals. But does this pose a threat to the medical profession?
The short answer is no—or at least not for now. Even though advances in cancer screening may reduce the time doctors are forced to dedicate to reading X-ray images, the treatment of such diseases shows no signs of being handed over to machines.
In fact, it’s thought that these tools will be merely used to help radiologists reach more accurate conclusions. For example, if professionals are unsure of whether a mass is cancerous or not, they can turn to the AI tool. In other words, it’s a more reliable, second medical opinion.
Is it just for breast cancer?
Though Oberai’s tool was engineered for the detection of breast cancer alone, it can be easily adapted to suit other types of cancer. In fact, progress is already being made on a similar AI for detecting renal cancer in CT images.
Since the effect cancer has on tissue varies depending on the organ or area of the body in question, there is, unfortunately, no “one-size-fits-all” tool. Instead, it’s necessary to train a machine to detect each specific type of cancer.
So is it just a question of time? Not exactly. A lack of reliable data is also a big issue when it comes to medical research. There can also be obstacles caused by patient data protection, resulting in not having a big enough sample size for tools to be trained properly for tumor detection.
What does the future hold?
All research indicates that cancer mortality rates will continue to drop as medical advances continue to be made. With millions of dollars being put into cancer research every year it seems inevitable that a cure will be found at some point. But what role will computer algorithms play in all of this?
Jo Bhakdi, Quantgene’s CEO and founder, believes that wide-scale AI implementation could lead to a fully proactive healthcare system that responds preemptively to diseases. This would signify a break away from the current model which focuses on the treatment of sick individuals.
Ultimately, this breakthrough in computer-vision AI will save millions of lives once it is adopted as a routine means of image analysis by doctors across the world. By being able to detect cancerous cells before they are visible to the human eye, it seems clear that cancer mortality rates will drop dramatically.
In other words, cancer may no longer be viewed as the death-sentence diagnosis it is today. When it comes to the implementation of AI in cancer detection, it seems the sky really is the limit.