Thermal Ablation Research Paper

Abstract—this paper investigates thermal ablation of cancer tissue as

a controls problem. The objective is to deliver a lethal dose of

thermal energy to the cancer tissue without damaging critical nearby

healthy tissue. This paper proposes a model predictive controller

(MPC) to regulate thermal dose. The model predictive controller

guarantees the safety of the healthy tissue surrounding the tumor. The

cost function rewards killing cancer cells. Simulations comparing the

proposed controller with the bang-bang controller used in practice

demonstrate that the MPC controller kills more of the cancer tissue

and requires less time than the bang-bang controller, while damaging

less healthy tissue.

INTRODUCTION

Ultrasonic thermal

Section III converts the general problem into an

implementable discrete finite-dimension controller.

A. Therapeutic Goal

The therapeutic goal is to thermally ablate all of the

cancer tissue while preserving all critical nearby tissues. Cell

death is a nonlinear function of time and temperature. The

common heuristic used to measure thermal dose is the total

Cumulative Equivalent Minutes (CEM) at some reference

Fig. 1. A basic block diagram level approach to control of thermal

ablation.

An ultrasonic transducer heats the tissue based on feedback from

thermally sensitive magnetic resonance imaging as well as a

calculation of the thermal dose (CEM). Here, the controller is a

model predictive controller.

Cell death is guaranteed when the CEM is above the threshold CEM

= 240 min. In contrast, to guarantee no damage, the thermal dose

CEM(r; t) should be kept below the threshold CEM = 10 min. In

addition to using the heuristic CEM, it is common to also define a

threshold constraint on the tissue temperature T(r; t) of sensitive

healthy tissue.

Thus, here we also consider keeping the healthy tissue below the

reference temperature Tref as part of the therapeutic goal.

B. Bang-Bang Controller

In clinics, a bang-bang controller is often used to solve problems. For

the bang-bang controller, the input uj(k) is determined by the

temperature measured at two pilot points rj1 and rj2. The pilot point

rj1 is heuristically selected to be the point with the highest

temperature response to the j-th ultrasonic transducer. The output

uj(k) is set to the maximum value umax as long as the temperature

T(rj1; tk) is below a threshold Tbang and set to zero when the

temperature is too high.

The pilot point rj2 is used to determine when to permanently shut-off

the j-th ultrasonic transducer. When the temperature T(rj2; tk) rises

above the threshold Tbang the ultrasonic transducer is shut-off for the

remainder of the treatment. The treatment is complete when all the

actuators have been shut off or the treatment time window has

elapsed.

C .Sensitive Nerve Bundle

constrained MPC controller is an improvement over the current state-

of-the-art. In practice, a priori information can be used to intelligently

place control and pilot points, but this does not guarantee constraints

will be met. Additionally, there can be a stark trade-off