projects

Active EMI Suppression for MRI-Guided Thermal Ablation

Problem: Microwave ablation devices flood the MRI scanner with electromagnetic interference, corrupting the temperature images clinicians need to guide the procedure.

Approach: I developed a software-based active EMI suppression system that estimates and removes device interference directly from MRI raw data using a separate receiver coil array as a noise reference — requiring zero modifications to the ablation hardware.

Outcome: The system achieves 91% EMI suppression and real-time capability (~15 ms/slice), enabling continuous MR thermometry during active ablation with temperature accuracy of 1.3°C validated against fiber-optic probes.

Related:

Volumetric MR Thermometry in Moving Tissues

Problem: Monitoring temperature during liver ablation requires 3D coverage, but the liver moves with every breath, corrupting standard thermometry.

Approach: I developed a volumetric PRF thermometry framework combining golden-angle 3D stack-of-radial MRI acquisition with an image-navigated multi-baseline method that tracks respiratory motion directly from reconstructed images — eliminating the need for external navigators.

Outcome: The system achieves full-liver coverage (~72 mm) refreshing every ~1 second with 1.44°C accuracy, reducing temperature measurement error by over two-thirds compared to conventional single-baseline methods.

Related:

Computational Modeling for Microwave Ablation Treatment Planning

Problem: Ablation treatment planning relies on computational models to predict ablation zone geometry, but existing models use generic tissue parameters that limit predictive accuracy — especially for pulsed protocols with complex cooling-reheating dynamics.

Approach: I built a calibrated EM–bioheat computational model in COMSOL Multiphysics, with tissue-specific parameters derived via inverse-problem optimization against experimental measurements rather than literature values.

Outcome: The calibrated model is validated against independent MR thermometry data acquired during ex vivo ablations with active EMI suppression, enabling direct comparison of simulated and measured spatiotemporal temperature evolution.

Related:

MRI-Guided Thermal Ablation in Oncopig Cancer Models

Problem: MR thermometry can estimate ablation zones during treatment, but whether those predictions match actual tissue damage requires systematic validation in a biologically realistic model.

Approach: I designed and executed an end-to-end MRI-guided microwave ablation workflow in Oncopig subjects with induced liver tumors — from planning MRI and real-time needle targeting through respiratory-triggered MR thermometry with thermal dose evaluation to post-ablation imaging and gross pathology confirmation.

Outcome: Ablation zone dimensions from thermometry, post-ablation MRI, and pathology showed agreement across all procedures, establishing a preclinical framework directly translatable to clinical deployment.

Related:

MRI-Compatible Surgical Robotics

Problem: MRI-guided liver procedures face a practical constraint: the scanner bore limits the clinician’s physical access to the patient during imaging.

Approach: I collaborated with the UCLA Mechanical Engineering Department on developing and validating MRI-compatible surgical robots designed to overcome this limitation for liver biopsy and thermal ablation. My contributions spanned real-time device tracking and visualization on MRI, workflow optimization, and prototype testing — validated through in-bore, ex vivo, and in vivo experiments.

Outcome: The integrated system demonstrated successful device localization and procedure execution within the MRI bore, establishing feasibility for robot-assisted MRI-guided liver interventions.

Related:

Focused Ultrasound-Mediated Drug Delivery with MRI Confirmation

Problem: When delivering a therapeutic agent to a tumor using focused ultrasound, confirming that the drug reached its target during the procedure remains a key challenge.

Approach: I develop MRI-based acquisition and monitoring methods for verifying focused-ultrasound-mediated nanoparticle delivery across stimuli-responsive platforms. In one line of work, I designed a modulation enhancement mapping framework (“Spotlight MRI”) that uses cycled HIFU to toggle temperature-sensitive nanoparticle contrast on and off. In a second, I used MRI and MR thermometry to confirm targeted drug release from a mechanosensitive nanoparticle platform triggered by HIFU.

Outcome: The Spotlight MRI framework achieved 139-fold signal amplification localized to the HIFU focal zone. The mechanosensitive platform demonstrated improved therapeutic outcomes versus controls in vivo.

Related:

Renal Tumor AI and Open Dataset

I curated a large multi-phase contrast-enhanced CT dataset of renal mass patients at UCSF and developed deep learning-based models for tumor detection, classification, and aggressiveness prediction. The dataset has been prepared for public release as an open benchmark for the research community.

Related:

Hyperpolarized ¹³C MRI for Renal Tumors

I built data-driven kinetic models and metabolic quantification pipelines for dynamic hyperpolarized ¹³C pyruvate MRI in patients with renal tumors, supporting non-invasive assessment of tumor metabolism.

Related: