It’s obvious that it takes years to train doctors, especially those who handle serious and complicated medical issues – pathologists, cardiologists, dermatologists and the rest, that’s why there’s always a shortage of these lifesaving experts. Thanks to artificial intelligence because now, machines can be trained to help fill that shortage.
In fact, already, we have AI tools that can diagnose pneumonia, fungi, depression and certain eye infections — all with an average accuracy rate of over 92 percent. And you know what, the list is expanding further! Chinese researchers have managed to develop a new system that diagnoses prostate cancer, as accurately as pathologists do.
It Means Improvement in Efficiency During Cancer Diagnosis
This new application of machine intelligence is a clear example of how AI can be used to better healthcare. Owing to the fact that there is a great shortage of pathologists and equipment used to tackle cancer, the new system can be served online at a tap, freeing the specialists, to focus on analyzing the diagnostic results and issue the more serious advice.
In most cases, processes delay in healthcare centers because diagnosis takes the most time, and it’s also the most expensive stage of treatment. But with such systems, it would mean less waiting, for results, and lower charges.
The Chinese scientists believe that their new system will streamline, but more so, eliminate variations (which often bring doubts) in the process of a cancer diagnosis.
Prostate Cancer Diagnosis
Like several other types of cancers, testing prostate cancer requires samples, which doctors call, biopsy. This is what the specialist examines to declare the status of the concerned.
Likewise, the Chinese AI system, presented at the EAUC (European Association of Urology Congress) this week in Copenhagen uses biopsies to state the fate of the concerned — exactly the way a highly trained pathologist would offer their verdict.
However, Hongqian Guo the AI expert who steered the research made it clear that “This will not replace human pathologist because the final decision of the diagnosis needs to rest with an experienced specialist. A machine cannot take full responsibility. In essence, the system’s work is to promote faster and better diagnoses – as well as cut-off variations during judgment (a common trend seen in human evaluations.)”
Testing the New Algorithm
The system was trained and set so that it learns on its own when it’ll be in real life application. Now, Guo and his team engaged over 280 patients, who donated 918 prostate samples. Like with most AI diagnostics that use image recognition, the samples were further split into smaller images — where 30,000 of them were fed into the system during training and 10,000 were used to test the system’s intelligence and progress.
Writing the summary, Guo and his team agreed that on average, the system displayed a 99 percent accuracy in telling whether a sample is affected or not. He added that “The system is as accurate as a trained pathologist because it can also perfectly classify the malignancy level or stage of the prostate cancer.”
Real Life Application
Ideally, the hardest decision currently exhibited with most algorithms is adopting them for real life use. Until now, scientists can’t tell whether it’s ready for an agent to fully take up a clinical task or not. Maybe this is because there are still less AI testing services.
Nonetheless, a tool like this at work would mean accuracy and consistency in cancer diagnosis across specialists, hospitals and even the globe. And besides being fast and more available than human pathologists, the software can also be served to remote areas with less medical advantages or specialist through the cloud.
In an overview, training a human to a level of becoming doctor is expensive, time demanding and sometimes less predictive -because the person might decide to swap careers, from being a pathologist to something else.
But with training a machine, the time and resource become less demanding and more interesting cheaper in comparison, because a trained software can be shared from a platform. In simple words, it takes less to develop a machine pathologist and the benefits are long term because the automation cannot change mind.
Automating prostate cancer diagnosis will also lead to less reliance on human specialists. Nonetheless, we’ll still need a good number of trained and experienced pathologists because machines, as intelligent as they appear, cannot take full responsibility when it comes to decision making.