Distinct CT imaging features of new liver metastases from primary genitourinary cancers

Tae-Hyung Kim ¹, Wyanne Law ¹, Bora Kalaycioglu ²

⁰Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, NY, USA.

Abdominal Radiology (new York) +

|

May 20, 2024

View abstract on PubMed

Summary

Natural language processing (NLP) identified CT imaging features of liver metastases in genitourinary cancers. These liver metastases are typically hypovascular, with distinct features varying by primary cancer type.

Area Of Science

Radiology
Oncology
Medical Informatics

Background

Genitourinary cancers can metastasize to the liver.
Characterizing liver metastases is crucial for diagnosis and treatment planning.
Distinguishing imaging features may aid in identifying the primary cancer type.

Purpose Of The Study

To apply natural language processing (NLP) to structured radiology reports.
To investigate CT imaging features of new liver metastases from primary genitourinary cancers.
To identify potential differences in imaging features based on the primary genitourinary cancer type.

Main Methods

A retrospective study utilizing a previously reported NLP model on CT reports (2016-2022).
Validation of NLP predictions using pathology or imaging follow-up.
Qualitative assessment of liver metastasis imaging features by two radiologists.
Comparison of imaging feature proportions across different primary genitourinary cancers using Chi-square or Fisher's exact test.

Main Results

In 112 patients, most liver metastases were hypovascular (73.2%), well-defined (76.6%), homogenous (66.9%), and without necrosis/cysts (73.2%).
Iso- to hyperdense metastases were more common in kidney cancer (42.5%) compared to other genitourinary cancers (p < 0.05).
Ill-defined margins were more frequent in prostate cancer metastases (44.0%) than in others (p < 0.05).

Conclusions

New liver metastases from genitourinary cancers are predominantly hypovascular.
Distinct CT imaging features of liver metastases vary significantly between primary genitourinary cancer types.
NLP analysis of radiology reports can aid in characterizing liver metastases and potentially inferring primary cancer origin.

Keywords:

Computed tomography Liver metastasis Urogenital neoplasm