Enhancing prediction accuracy for muscle invasion in bladder cancer using a dual-energy CT-based interpretable model incorporating habitat radiomics and deep learning

Comprehensive Summary

In developing personalized treatment plans within the field of urology, a topic of interest is the ability to accurately predict muscle invasion status in bladder cancer (BCa). In a study of 200 BCa patients, the use of a dual-energy computed tomography (DECT) image model was tested for perioperative assessments. A habitat analysis and 2.5- dimension (2.5D) deep learning system was used in the development of DECT; and results revealed that an integrated model combining effective atomic number (Zeff), habitat features, and ResNet 101-basedDL features presented the optimal performance in predicting muscle invasion in BCa. Reliability of the model in clinical applications was also confirmed through calibration curves.

Outcomes and Implications

The use of predictive models such as DECT allow for the enhancement of preoperative predictions of muscle invasion status in patients with bladder cancer. This can then aid in the development of more personalized patient treatment plans, where disease development is more accurately identified and appropriate actions are taken in response.