Main Session
Sep 29
PQA 06 - Genitourinary Cancer, Gynecological Cancer, and Health Care Access and Engagement

3410 - Wearable Artificial Intelligence Powered Device for Optimal Pelvic Radiation Therapy

02:15pm - 03:30pm ET
Poster Hall - Exhibit Hall A
Screen: 21
POSTER

Presenter(s)

Fan Zhu, MD, BS - UMass Memorial Medical Center, Worcester, MA

F. Zhu1, M. Sealander2, E. S. Paulson1, and W. A. Hall1; 1Department of Radiation Oncology, Medical College of Wisconsin, Milwaukee, WI, 2Sonoptima, Milwaukee, WI

Purpose/Objective(s):

Pelvic RT sometimes requires bladder filling to specific levels, which can be difficult for patients to accomplish accurately and repeatedly. Patients may be taken off the treatment table or have their treatments canceled due to insufficient bladder filling, resulting in patient dissatisfaction, additional resource consumption (i.e. therapist and physician time), and machine idling. We hypothesize that a wearable artificial intelligence (AI)-equipped ultrasound (US) device can be designed to better inform bladder readiness for pelvic RT.

Materials/Methods:

A wearable US prototype was designed and consists of two main components:

  1. A transducer assembly (TA) with a belt for placement on the subject
  2. An electronics package containing processing hardware
The TA consists of four piezoelectric elements operating at 2.27 MHz, with one piezo element oriented perpendicular to the body surface and three piezo elements angled 10° outward. These elements cast acoustic windows with tissue-like acoustic impedance and cast matching layers with intermediate acoustic impedance. The TA also has a custom transducer printed circuit board and coaxial cables connecting to the electronics package. The electronics package consists of a custom circuit board with power supplies, ultrasound driver chip, buffer amplifier, transmit pattern generator module, red Pitaya digitizer module (125 MHz sampling capability), Field-Programmable Gate Array (FPGA) module, 7.4V Li-Ion battery (IEC 62133 certified), Wi-Fi antenna, cooling fan and power button with LED indicator. AI is used to process the data to predict bladder readiness.

Results: The prototype is intended to be worn by patients during CT/MR simulation, and while in the waiting room before daily RT treatments. The device will be calibrated during CT/MR simulation once optimal anatomy and bladder filling have been achieved. Each day while in the waiting room before treatment, the device continuously monitors various bladder state variables, their rates of change, and estimates time remaining until maximal overlap with treatment planning anatomy. A feasibility trial (NCT06820385) is currently underway, with target enrollment of five healthy volunteers and ten patients, to evaluate the feasibility of acquiring and evaluating relative bladder volume (RBV) using the wearable imaging device, when compared to CT (for patients) or MRI (for volunteers). RBV is defined as the ratio of BV at each treatment to the BV at simulation. Feasibility is defined as successful data acquisition in the majority of patients (>50%). A successful data acquisition event refers to the device’s ability to approximate RBV to that measured by the daily cone beam CT (for patients) or MRI (for healthy volunteers) to within 30%.

Conclusion: A wearable AI-equipped US device was successfully designed and a prototype was constructed. A feasibility trial is currently underway. Final trial results will be reported at the 2026 ASTRO meeting.