Assessing Rectal Cancer Treatment Response Using Coregistered Endorectal Photoacoustic and US Imaging Paired with...

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Conventional radiology struggles to differentiate rectal cancer recurrence from treatment effects.
• Accurate assessment of treatment response is crucial for optimizing patient management.

Purpose of the Study:

• To develop and evaluate an endorectal photoacoustic microscopy (PAM) and ultrasound (US) imaging system.
• To utilize a convolutional neural network (CNN) for assessing rectal cancer treatment response.

Main Methods:

• Developed an endorectal PAM/US probe with a 1064-nm laser and US transducer.
• Trained and validated PAM and US CNN models on ex vivo and in vivo colorectal tissue data.
• Tested the CNN models on independent patient data to differentiate tumor from normal tissue and assess treatment response.

Main Results:

• The PAM CNN model identified unique markers of complete tumor response, such as restored vascular architecture.
• The PAM CNN achieved an area under the receiver operating characteristic curve of 0.98 in differentiating tissue types.
• The US CNN achieved an area under the receiver operating characteristic curve of 0.71.

Conclusions:

• The integrated PAM/US system with CNN demonstrates high diagnostic performance for rectal cancer treatment response assessment.
• This technology holds potential for optimizing post-treatment management strategies.
• The system effectively differentiates residual tumor from treatment-induced scarring.