ABSTRACT Colorectal cancer remains one of the leading causes of cancer mortality, and its prognosis is strongly dependent on early detection. Conventional diagnostic approaches, such as colonoscopy and capsule endoscopy, primarily rely on visual inspection and often fail to capture subsurface or biomechanical abnormalities. To address this limitation, an AI‐enhanced vibration‐assisted capsule is developed for in situ biomechanical sensing of bowel tissues. The self‐contained capsule integrates an eccentric vibration motor, triaxial accelerometer, power supply, and wireless communication unit within a swallowable format. During operation, motor‐induced oscillations deform surrounding tissue, and the resulting acceleration responses are wirelessly transmitted for real‐time analysis. Experiments conducted on ex vivo porcine colon specimens and a soft pneumatic in vitro colon simulator demonstrate that vibration‐induced signals capture reproducible mechanical contrasts between normal tissue and lesions of varying relative stiffness. A one‐class support vector machine, trained exclusively on normal tissue data, is employed to classify abnormal tissues without requiring extensive annotated datasets. Complementary dynamic modeling further identifies equivalent stiffness parameters, correlating lesion stiffness with abnormal classifications. This work demonstrates a minimally invasive, AI‐driven modality for non‐visual colorectal cancer detection, establishing the potential of vibration‐assisted capsules to provide real‐time, patient‐friendly screening and to complement existing imaging‐based diagnostics.
Fang et al. (Tue,) studied this question.