Optical Imaging Techniques and Machine-Learning Models for Diagnosis of Breast, Ovarian, and Endometrial Cancers

This talk will focus on our recent progress in cancer imaging across three major research areas:
Ultrasound-guided diffuse optical imaging combined with machine learning models to improve the accurate diagnosis of breast cancer and reduce unnecessary breast biopsies;
Dual-modality photoacoustic and ultrasound imaging with machine learning models for accurate diagnosis of ovarian cancer and reduction of unnecessary surgeries;
Optical coherence tomography (OCT) integrated with machine learning models for potential endometrial cancer screening and accurate diagnosis.

In our first study on ultrasound-guided diffuse optical imaging, we completed a 5-year clinical trial involving 300 patients. The results showed a 25% reduction in unnecessary benign biopsies. A fusion machine learning model demonstrated its potential to enhance radiologists’ performance in further reducing these unnecessary procedures.  In our second study on dual-modality photoacoustic and ultrasound imaging, recent diagnostic and machine learning results from over 70 patients demonstrated significantly improved sensitivity and specificity in ovarian lesion diagnosis.  In the third study, we demonstrated that optical coherence tomography (OCT), combined with machine learning algorithms, can be used as a screening and diagnostic tool for endometrial cancer.
Finally, I will highlight our next steps toward advancing these technologies and AI tools to support robust clinical translation.