A new AI program, based on nearly 500,000 tissue images, can accurately diagnose and predict the severity of lung adenocarcinoma, the most common form of lung cancer. Developed by researchers at NYU Langone Health and the University of Glasgow, this self-taught algorithm provides a reliable second opinion for oncologists, distinguishing lung cancer types with 99% accuracy and predicting cancer recurrence with 72% accuracy.
Main Points:
Accurate Diagnosis: The AI program diagnoses lung adenocarcinoma with high precision using structural features from patient tissue samples.
Predictive Power: It predicts cancer recurrence better than pathologists, offering significant prognostic insights.
Self-Learning Capability: The algorithm independently identifies significant features for diagnosis and prognosis, improving over time with more data.
Summary:
Researchers at NYU Langone Health and the University of Glasgow have created an AI program that accurately diagnoses lung adenocarcinoma, the most common type of lung cancer, by analyzing structural features from nearly half a million tissue images. This self-taught algorithm provides a reliable second opinion for oncologists and has shown a 99% accuracy rate in distinguishing between lung cancer types and a 72% accuracy rate in predicting cancer recurrence, outperforming traditional pathologist assessments.
The program, called histomorphological phenotype learning (HPL), leverages the AI’s ability to learn independently, identifying key features that impact diagnosis and prognosis. It offers a detailed and unbiased analysis, significantly aiding in treatment decisions. The researchers aim to further refine the tool and extend its application to other cancers.
