Universal Technical Institute of Illinois

Biopsy procedures are essential for definitive cancer diagnosis but remain limited by the risk of accidental blood vessel puncture, which can lead to hemorrhage and procedural failure. Conventional imaging guidance often provides insufficient vascular contrast, making vessel avoidance during needle insertion challenging. A rotational oblique spectral ultrasound (ROSUS) imaging system was developed to improve vessel detection and needle guidance during biopsy procedures. The device integrates a high-frequency PMN-PT 1-3 composite transducer mounted at a 45° angle within the 18-gauge needle tip, enabling simultaneous forward- and side-looking capability. By combining synchronized rotational-axial scanning with multifrequency signal ratio (MFSR) processing, ROSUS achieved volumetric images with blood-tissue contrast ratio improvement over 1.2 dB compared to conventional B-mode signal processing while maintaining high spatial resolution of 85 µm and 424 µm in axial and lateral directions, respectively. These results demonstrate that frequency-domain spectral processing can improve vessel and tissue differentiation, offering an 18-gauge needle-integrated platform for safer and more accurate biopsy needle-based procedures.

We introduce CollabToolBuilder, a flexible multiagent LLM framework with expert-in-the-loop (HITL) guidance that iteratively learns to create tools for a target goal, aligning with human intent and process, while minimizing time for task/domain adaptation effort and human feedback capture. The architecture generates and validates tools via four specialized agents (Coach, Coder, Critic, Capitalizer) using a reinforced dynamic prompt and systematic human feedback integration to reinforce each agent's role toward goals and constraints. This work is best viewed as a system-level integration and methodology combining multi-agent in-context learning, HITL controls, and reusable tool capitalization for complex iterative problems such as scientific document generation. We illustrate it with preliminary experiments (e.g., generating state-of-the-art research papers or patents given an abstract) and discuss its applicability to other iterative problem-solving.