Finding and treating cancer can be difficult. The right equipment and tools are needed to help detect the cancer so medical staff can choose the appropriate method of treatment. New methods, which utilise technology, are required to ensure that cancers are found quickly and efficiently, especially in cancers that have high rates of mortality.
One form of cancer that has a high mortality rate is lung cancer, the cause of one in five cancer-related deaths worldwide. In England, over one-third of cases are diagnosed after the patient has progressed to a state of emergency, whereby the cancer has already reached a critical state. To stop this, medical institutes and charities have come together to try to improve the detection of lung cancer of UK citizens, including recent news of screening in shopping centre car parks.
To effectively diagnose lung cancer and its progression, hospital staff use Positron Emission Tomography (PET) and Computed Tomography (CT) scans to view the lung and assess the anatomical structure and functional characteristics, alongside the lung’s lesions. This process can be complicated, costly, and time consuming.
However, to simplify the process of detecting lung cancer, Future Processing and NVIDIA have partnered to make the diagnosis process more affordable, accessible, and accurate. By working with medical imaging experts, including research institutions and clinics worldwide, the two companies have simplified the use of the lung cancer detecting tools and have developed an artificial intelligence (AI)-based software to make better sense of CT scans.
A particular aspect of lung cancer detection that Future Processing and NVIDIA have focussed on is CT imaging, whereby medical staff can analyse CT scan images only. Whereas clinicians typically view both PET and CT scans to determine a patient’s condition, this progression in medical imaging—including using convolutional neural networks—can ultimately lead to diagnoses based solely on CT scans. By using the AI enabled-software, medical professionals can segment the lesions automatically, saving precious time for both professionals and staff, while also measuring lesion progress as it develops.
“Before, the segmentation and analysis of active lesions in lungs required co-registering PET and CT sequences in a time-consuming procedure,” explains Dr. Jakub Nalepa, senior research scientist at Future Processing. “In our paper, we showed how convolutional neural networks can be effectively used to automate the segmentation of lesions and make it much easier in practice.”
This technology will not only simplify the process, but it could also save a lot of money for medical sites. On average, a CT scan costs between $1,200 to $3,200, whereas a PET scan costs $3,000 to $6,000. Alongside cost, patient comfort will also be improved, as patients will only need to prepare for one scan, rather than both the CT and PET scans. The AI-enabled software has also proven to reduce the rate of false positives, when studying lung data without active lesions, from 90.14% to 6.6%.
However, the Future Processing and NVIDIA teams haven’t stopped there, as they plan to develop this solution further and to apply its findings to other forms of cancer. AI in healthcare has progressed quickly in recent years, and has the potential to improve the accuracy and time it takes to detect and diagnose cancers, among other forms of illness.