2211 125I prostate implants: Automatic optimization of dose distribution using fast simulated annealing

I. J. Radiation OncoIogy*Biology*Physics

346

Volume 39, Number 2, Supplement, 1997

2211 ‘*‘I PROSTATE

IMPLANTS:

AUTOMATIC

OPTIMIZATION

OF DOSE

DISTRIBUTION

USING

FAST

SIMULATED

ANNEALING

Pouliot Jean, Ph.D., Tremblay Daniel, Ph.D., Roy Jean MD, M.Sc., and Cot& Carl, B.Sc.A Centre Hospitalier

Universitaire

de QuCbec, Department

of Radiation Oncology,

Qdbec,

Canada

To automatically determine seed and needle numbers and dispositions within the target volume to achieve optimized of tramperineal “‘1 permanent implants.

Purpose/Objective:

distribution

dose

& Methods: Target volume is initially determined from a set of ultrasound (or CT) transverse slices. Clinical criteria describing implant quality are expressed by a mathematical cost function. The first term of cost function takes into account the minimum peripheral prescribed dose on the prostate contours. The second term insures dose uniformity throughout the volume. Finally, the third term, analogous to a c&al potential, governs the transverse needle scheme distribution and minimizes their numbers. The last term also adds penalties on needles located too close to each other or containing only one seed. To begin the optimization, the user defines all possible needle positions covering the entire prostate at each 0.5 cm on the transverse template, except in the vicinity of the urethra which is devoided of needles. Then, the program starts and an initial number of seeds is randomly distributed within the allowed needle positions. The simulated annealing algorithm allows seed and needle positions to be varied and their number reduced to optimize the cost function. Materials

Proper weightings between the three terms of the cost function were determined emperically optimization. The optimization has been used in the planning of 50 patients with prostate volume ranging from an average size prostate, the optimization can be completed within 50000 iterations, approximately 45 minutes The obtained dose distribution is characterized by the prescribed &dose following the shape of the prostate The prostate DVH for the prescribed dose is generally larger than 98 %.

Results:

and use 19 to 56 on a Sun contours

to perform the dose cc (median 33 cc). For SPARCS workstation. with a proper margin.

Conclusion: The use of a cost fimction closely related to the clinical criteria and fast simulated anxaling allow for consistent and automatic determination of seed and needle distributions resulting in an optimized dose distribution customized for each patient and independent of the dosimetrist experience. The algorithm has been fully implemented clinically and was used for the planning of 50 patients treated for prostate cancer with transperineal ‘*‘I permanent prostate implants. The outcome is a better dose distribution and obtained much faster than what can be achieved without automation.

2212 POST IMPLANT Chandrasekhar

DOSIMETRIC

EVALUATION

OF I-125 PERMANENT

Kota’ , Mark Yudelev’, Suzanne Chungbin’, Peter Litrrup’,

PROSTATE

IMPLANTS

David Wood’, Jeffrey Forman’

‘Gershenson Radiation Oncology Center, *Department of Radiology, ‘Department of Urology Karmanos Cancer Institute, Harper Hospital and Wayne State University, Detroit, MI, 48201. Purpose I Objective: To evaluate the dosimetric quality of ultrasound (US) guided I-125 permanent seed implants for prostate cancer planned using preimplant US images and reconstructed using postimplant CT images. Materials and Methods: Preimplant US images were used to calculate I-125 seed locations on a 0.5 cm grid with parallel needle positions, to obtain the prescription dose to a 5 mm margin around the prostate. Patients were implanted with I-125 seeds in the operating room using a Mick applicator and free hand guidance of the needles/seeds using a 0.5 cm grid template on the ultrasound images. The clearly visualized needle tip in the saggital images was used to determine the seed drop off position. Patients were simulated on a conventional and a CT simulator within 24 hours of the implant. Prostate contours were outlined on the CT images by the radiation oncologist. Seed positions reconstnrcted from CT images and verified with localization films were used to generate post implant dose distributions. Dose volume histograms of the implant and the prostate were used to obtain the following parameters which were subsequently used to evaluate the quality of the implants: (i) prostate volume covered by prescription dose (PVPD), (ii) implant volume of prescription dose (IVPD), (iii) minimum peripheral dose (MPD), and (iv) heterogeneity index (HI).

1 - Preimplant Ultrasound;

2- Postimplant CT;

In spite of the observed deviations in the seed locations from the preplan, the implant volumes of the prescription dose ware comparable in the pre and post plans. The postimplant CT volumes were consistently larger than the preimplant US volumes. Even though the implant volumes were consistently larger than the postimplant CT prostate volumes, the fraction of the prostate receiving the prescription dose ranged from 75% to 95%, which is consistent with other studies published in the literature. This could be due to either a difference in the US and CT prostate shapes and volumes, or a geometrical r&positioning of the seeds. The areas of the prostate receiving less than the prescribed dose were found to be randomly distributed in all cases. Our implant procedure, based on US based preplanning and CT based postplanning is seen to be comparable to those practised at other centers. Deviations from planned dose distributions most likely indicate the limitations of this concept, which could be overcome by on-line postimplant dose verification and correction in the OR or m-implantation at a later time. Conclusions: