Cancer of unknown primary (CUP) is a heterogeneous group of metastatic cancers whose primary anatomical site of tumor origin cannot be determined despite an appropriate pretreatment diagnostic evaluation. These carcinomas have a poor prognosis in about 80% of patients due to their early dissemination, aggressiveness, late presentation, and unpredictability of metastatic pattern. Their median overall survival (OS) ranges from 3 to 10 months. This poses a considerable challenge, as modern therapeutics are predominantly specific to the primary tumor.
CUP has an average age at diagnosis of 60 to 75 years and accounts for 2% to 9% of all tumors. Multiple sites of involvement are observed in greater than 50% of these patients. Common sites of involvement are the liver, lungs, bones, and lymph nodes. Although certain patterns of metastases suggest possible primary sites, CUP can metastasize to any site. Therefore, physicians should not rely on patterns of known metastases to determine the primary site in patients with CUP.
Scientists have now developed an artificial intelligence (AI) tool that outperforms pathologists at identifying the origins of metastatic cancer cells that circulate in the body. “That’s a pretty significant finding — that it can be used as an assistive tool,” says Faisal Mahmood, who studies AI applications in health care at Harvard Medical School in Boston, Massachusetts.
One method used to diagnose tricky metastatic cancers relies on tumor cells found in fluid extracted from the body. Clinicians examine images of the cells to work out which type of cancer cell they resemble. For example, breast cancer cells that migrate to the lungs still look like breast cancer cells. The researchers trained their AI model (TORCH) on some 30,000 images of cells found in abdominal or lung fluid from 21,000 people whose tumor of origin was known. They then tested their model on 27,000 images and found there was an 83% change that it would accurately predict the source of the tumor. And there was a 99% chance that the source of the tumor was included in the model’s top three predictions. Having a top-three list is useful because it can help clinicians to reduce the number of extra — often intrusive — tests needed to identify a tumor’s origins.
The researchers also 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. “This is a pretty convincing argument” for using the AI model in a clinical setting, says Mahmood.
This AI model could become a valuable tool in differentiating between malignant tumor and benign disease, localization of cancer origins and aiding clinical decision making in patients with CUP. It is a challenging task to identify the origins of metastatic free tumor cells using limited clinical information and cytological images. TORCH achieved robust performance across five testing sets and outstanding accuracy versus a group of four pathologists. The high technical performance and potential clinical benefits of TORCH warrant further investigation in prospective randomized clinical trials.
