Google AI technology could be better at diagnosing lung cancer than human experts, suggests a new study

Scientists at Google have developed an artificial intelligence (AI) model which y claim is better at diagsing lung cancer than human experts, an vance that could le to earlier treatments for dely disease.

Deep learning -- a form of AI -- was able to detect malignant lung dules on low-dose chest computed tomography (LDCT) scans with a performance meeting or exceeding that of expert riologists, researchers said.

system, described in journal Nature Medicine, provides an automated im evaluation system to enhance accuracy of early lung cancer diagsis that could le to earlier treatment.

deep-learning system was compared against riologists on LDCTs for patients, some of whom h biopsy confirmed cancer within a year. In most comparisons, model performed at or better than riologists.

Deep learning is a technique that teaches computers to learn by example.

deep-learning system also produced fewer false positives and fewer false negatives, which could le to fewer unnecessary follow-up procedures and fewer missed tumours, if it were used in a clinical setting.

RE | 'Artificial intelligence (AI) systems could help robots assist soldiers in battlefield', say US Scientists

"Riologists generally examine hundreds of two-dimensional ims or 'slices' in a single CT scan but this new machine learning system views lungs in a huge, single three-dimensional im," said Mozziyar Etemi, a research assistant professor at rthwestern University in US.

"AI in 3D can be much more sensitive in its ability to detect early lung cancer than human eye looking at 2D ims. This is technically '4D' because it is t only looking at one CT scan, but two over time," Etemi said.

"In order to build AI to view CTs in this way, you require an ermous computer system of Google-scale. concept is vel but actual engineering of it is also vel because of scale," he said.

This research is incredibly important, as lung cancer has highest rate of mortality among all cancers, and re are many challenges in way of bro option of screening, said Shravya Shetty, technical le at Google.

"Our work examines ways AI can be used to improve accuracy and optimise screening process, in ways that could help with implementation of screening programs. results are promising, and we look forward to continuing our work with partners and peers," said Shetty.

Large clinical trials across US and Europe have shown that chest screening can identify cancer and reduce death rates, researchers said.

However, high error rates and limited access to se screenings mean that many lung cancers are usually detected at vanced sts, when y are hard to treat, y said.

deep-learning system utilises both primary CT scan and, whenever available, a prior CT scan from patient as input.

Prior CT scans are useful in predicting lung cancer malignancy risk because growth rate of suspicious lung dules can be indicative of malignancy.

computer was trained using fully de-identified, biopsy-confirmed low-dose chest CT scans.

vel system identifies both a region of interest and wher region has a high likelihood of lung cancer.

model outperformed six riologists when previous CT imaging was t available and performed as well as riologists when re was prior imaging.

" system can categorise a lesion with more specificity. t only can we better diagse someone with cancer, we can also say if someone doesn't have cancer, potentially saving m from an invasive, costly and risky lung biopsy," Etemi said.

RE | Artificial intelligence (AI) bias: Face recognition researcher fights Amazon

Google scientists developed deep-learning model and applied it to 6,716 de-identified CT scan sets to validate accuracy of its new system.

y found Ai-powered system was able to spot sometimes-minuscule malignant lung dules with a model of 0.94 test cases.

researchers cautioned that se findings need to be clinically validated in large patient populations.

However, y said this model may assist in improving manment and outcome of patients with lung cancer.