AI traces mysterious metastatic cancers to their source

🌈 Abstract

The article discusses the development of an artificial intelligence (AI) tool that can outperform pathologists in identifying the origins of metastatic cancer cells that circulate in the body. This proof-of-concept model could help doctors improve the diagnosis and treatment of late-stage cancer, and extend people's lives.

🙋 Q&A

[01] Developing an AI Tool for Identifying Cancer Origins

1. What is the key purpose of the AI tool developed by the researchers? The key purpose of the AI tool is to help identify the origins of metastatic cancer cells that circulate in the body, which can be challenging for pathologists. This could improve the diagnosis and treatment of late-stage cancer.

2. How did the researchers train and test the AI model? The researchers trained their AI model on 30,000 images of cells found in abdominal or lung fluid from 21,000 people with known tumor origins. They then tested the model on 27,000 images and found it had an 83% accuracy in predicting the source of the tumor, and a 99% chance of including the correct source in its top three predictions.

3. How did the AI model perform compared to human pathologists? When tested on 500 images, the AI model was found to be better than human pathologists at predicting a tumor's origin, and this improvement was statistically significant.

4. What were the key findings from the retrospective assessment of study participants? The researchers retrospectively assessed a subset of 391 study participants some four years after they had had cancer treatment. They found that those who had received treatment for the type of cancer that the model predicted were more likely to have survived, and lived longer, than participants for whom the prediction did not match.

[02] Potential for Improving Metastatic Cancer Diagnosis and Treatment

1. What are the challenges in diagnosing the origins of metastatic cancers? Up to 5% of all tumors have an unidentified origin, and the prognosis for people whose primary cancer remains unknown is poor. Clinicians currently rely on examining images of tumor cells to try to determine their origin.

2. How could the AI tool help address these challenges? The AI tool could help reduce the number of additional, often intrusive, tests needed to identify a tumor's origins. Having a top-three list of potential origins from the AI model could guide clinicians in their diagnosis and treatment decisions.

3. What are the potential next steps for improving metastatic cancer diagnosis and treatment? Combining the AI model's predictions with data from tissue samples and genomic data could further improve outcomes for people with metastatic cancers of unknown origins, according to the expert quoted in the article.