Imaging System Distinguishes Residual Rectal Cancer from Treatment Scarring

Assessing rectal cancer treatment response using an endorectal coregistered photoacoustic microscopy and ultrasound system paired with a deep learning model showed high diagnostic performance and potential for optimizing posttreatment management of patients, according to a study published online ahead of print in Radiology.

“Conventional radiologic modalities perform poorly in the radiated rectum and are often unable to differentiate residual cancer from treatment scarring,” researchers explained.

The prospective study reported on the development and performance of an imaging system that consists of an endorectal coregistered photoacoustic microscopy (PAM) and ultrasound system paired with a convolution neural network (CNN) in patients who completed radiation and chemotherapy for rectal cancer. PAM and ultrasound convolution neural network models were trained to distinguish normal from malignant colorectal tissue using ex vivo and in vivo patient data.

In the evaluation of 32 participants, the PAM CNN model was highly capable of differentiating tumor from normal tissue, researchers reported. For data from 5 participants, researchers reported that PAM CNN achieved an area under the receiver operating characteristic curve of 0.98. The ultrasound CNN had an area under the receiver operating characteristic curve of 0.71.

“Coregistered endorectal photoacoustic microscopy and ultrasound imaging quantified by a neural network demonstrated excellent differentiation between rectal tumor beds with residual cancers and those without viable tumors,” researchers reported.

—Jolynn Tumolo

Reference

Leng X, Uddin KMS, Chapman W Jr, et al. Assessing rectal cancer treatment response using coregistered endorectal photoacoustic and US imaging paired with deep learning. Radiology. 2021 March 23;[Epub ahead of print].