Deep Learning Can Be Useful in Eliminating Shortage of Cancer Doctors

Deep Learning Can Be Useful in Eliminating Shortage of Cancer Doctors

by November 5, 2019 0 comments

The national health service in Britain is facing a critical shortage of cancer doctors currently. This makes it extremely difficult to provide specialized care to seriously ill patients.

The stats from the Royal College of Radiologists depicts that nearly 7.5 percent of consultant roles at 62 major UK-based cancer centers are vacant. This has made them reliant on overtime.

Counter to this, Peltarion, a Swedish AI company worked with a radiotherapy company in order to develop a deep learning model. This model targets tumors for radiotherapy and helps eradicate doctor shortages. It saves time and improves patient care through faster, more accurate diagnostics and customized treatment plans.

Björn Brinne, chief AI officer at Peltarion said, “Deep learning technology is a key AI technique that can support the radiotherapy treatment process. With more than one billion radiologic examinations taking place every year, being able to quickly and accurately determine the disease or condition which explains a person’s symptoms is central to successful health service.”

He further stated, “Radiologists, for example, use data, and especially images, to form the core basis of their assessments, determining what the disease is and where it is displaying. There are over 1 billion radiologic examinations taking place every year.”

Radiologists create a segmentation mask to diagnose and treat brain tumors. These masks are essentially an image that is used to mark the exact location of cancerous growths. It also helps in creating a treatment plan.

Making the use of deep learning models, Peltarion is assisting radiologists to create these masks quickly and accurately. The technology used can help with the segmentation process. It will also enable doctors to work more efficiently when it comes to radiotherapy treatment.

According to Brinne, “In simple terms, you train a deep learning model with data—in this area, you train the models with a lot of images which are well labeled with data on the brain tumor, its characteristics and where it is displayed.”

The model is later optimized to ensure precision and accuracy. As Brinne explains these models can operate at the same level or at least equivalent to a medical professional.

These models are quick to run and are basically free from errors. For a new case, the deep learning model forecasts the appropriate segmentation mask and suggest where the tumor is present and how treatment can be applied to it.

Although this approach is quite new, Brinne believes that it can significantly assist radiologists and medical professionals by improving their ability to determine the segmentation mask and speed the diagnostic process. It can also act as a “second opinion.”

Brinne continued to say, “It is not about replacing experts, but helping them get quick and accurate results. First world health services struggle to provide radiology and pathology service at the level expected or demanded by their citizens. In many parts of the world, these services are simply not provided or are non-existent. The World Health Organisation estimates up two-thirds of the global population does not have access to basic radiology services.”

The company hopes that more and more organizations will release its platform approach to worldwide health systems. According to Brinne, deep learning can support and improve almost every diagnostics process based on images. Peltarion’s objective is to make deep learning more mainstream and accessible in more countries and with more providers.

Additionally, Peltarion wants to help organizations utilize different data types including language and audio, in deep learning-based diagnostic and treatment solutions. This could benefit several patients by simply transcribing patient data accurately or help flag potential diseases.

Brinne added, “Additionally, there are some interesting applications of deep learning with audio data like detecting stress from breathing patterns, diagnosing post-traumatic stress disorder and diagnosing asthma from cough waveforms. There is even a service that analyses audio cues from emergency calls to predict when someone might be having a heart attack.”

No Comments so far

Jump into a conversation

No Comments Yet!

You can be the one to start a conversation.

Your data will be safe!Your e-mail address will not be published. Also other data will not be shared with third person.