AI helps doctors better predict cancer treatment outcomes - study

An international research team in the I3LUNG project trained AI models on data from 2,396 patients across six medical centers in Italy, Germany, Spain, Greece, the United States and Israel, raising doctors' accuracy in predicting disease control from immunotherapy in advanced non-small cell lung cancer from 57% to 65%, according to Qazinform News Agency. Among 20 recruited physicians, correct identification of patients whose treatment achieved disease control rose to 87% with AI support versus 72% without it, though doctors became slightly more likely to wrongly predict a positive outcome for patients whose disease was not controlled. The team is now running further validation with more than 2,000 patients before the system can be introduced into clinical practice.

AI helps doctors better predict cancer treatment outcomes - study An international team of researchers has developed an artificial intelligence system to help doctors assess immunotherapy outcomes in advanced non-small cell lung cancer https://qazinform.com/tag/cancer t12201 . When reviewing patient records with its support, doctors’ accuracy in predicting disease control rose from 57% to 65%, Qazinform News Agency https://qazinform.com/ reports. Immunotherapy helps the immune system fight tumors, but not all patients experience lasting benefits. Identifying in advance who will benefit from treatment remains difficult. The indicators doctors currently rely on do not always provide accurate predictions. As part of the I3LUNG project https://www.nature.com/articles/s41591-026-04488-2 , researchers collected data from 2,396 patients at six medical centers in Italy, Germany, Spain, Greece, the United States and Israel. The patients received immunotherapy either alone or in combination with chemotherapy. The researchers trained AI models to predict the likelihood of disease control and patient survival. The systems analyzed clinical information and blood test results, while more advanced versions also incorporated computed tomography scans, images of tumor tissue samples and genetic data. To assess the tool’s usefulness to doctors, the researchers recruited 20 physicians. Half specialized in lung cancer https://qazinform.com/tag/cancer t12201 , while the others were oncologists without a lung cancer https://qazinform.com/tag/cancer t12201 specialization or doctors in training. They first assessed cases independently, then received AI predictions alongside explanations of which data had influenced the results. With the system’s support, doctors more accurately predicted whether treatment would achieve disease control, meaning tumor shrinkage or stabilization. Among patients whose treatment achieved this outcome, doctors correctly identified 87% of cases, compared with 72% without AI assistance. However, they also became slightly more likely to incorrectly predict a positive outcome for patients whose disease was not controlled by treatment. The team is now conducting further validation involving more than 2,000 patients. The system’s reliability and value in supporting treatment decisions must be confirmed before it can be introduced into clinical practice. Earlier, Qazinform https://qazinform.com/ News Agency reported https://qazinform.com/news/kazakhstan-considers-ai-for-lung-cancer-detection-851bd4 that deputy Sergey Ponomarev had called for the nationwide rollout of a domestic AI system for early lung cancer https://qazinform.com/tag/cancer t12201 detection in Kazakhstan.