A medical expert system approach using artificial neural networks for standardized treatment planning

PURPOSE: Many radiotherapy treatment plans involve some level of standardization (e.g., in terms of beam ballistics, collimator settings, and wedge angles), which is determined primarily by tumor site and stage. If patient-to-patient variations in the size and shape of relevant anatomical structures for a given treatment site are adequately sampled, then it would seem possible to develop a general method for automatically mapping individual patient anatomy to a corresponding set of treatment variables. A medical expert system approach to standardized treatment planning was developed that should lead to improved planning efficiency and consistency. METHODS AND MATERIALS: The expert system was designed to specify treatment variables for new patients based upon a set of templates (a database of treatment plans for previous patients) and a similarity metric for determining the goodness of fit between the relevant anatomy of new patients and patients in the database. A set of artificial neural networks was used to optimize the treatment variables to the individual patient. A simplified example, a four-field box technique for prostate treatments based upon a single external contour, was used to test the viability of the approach. RESULTS: For a group of new prostate patients, treatment variables specified by the expert system were compared to treatment variables chosen by the dosimetrists. Performance criteria included dose uniformity within the target region and dose to surrounding critical organs. For this standardized prostate technique, a database consisting of approximately 75 patient records was required for the expert system performance to approach that of the dosimetrists. CONCLUSIONS: An expert system approach to standardized treatment planning has the potential of improving the overall efficiency of the planning process by reducing the number of iterations required to generate an optimized dose distribution, and to function most effectively, should be closely integrated with a dosimetric based treatment planning system.     

A new genetic algorithm technique in optimization of permanent 125I prostate implants

Real time optimized treatment planning at the time of the implant is desirable for ultrasound-guided transperineal 125I permanent prostate implants. Currently available optimization algorithms are too slow to be used in the operating room. The goal of this work is to develop a robust optimization algorithm, which is suitable for such application. Three different genetic algorithms (sGA, sureGA and securGA) were developed and compared in terms of the number of function evaluations and the corresponding fitness. The optimized dose distribution was achieved by searching the best seed distribution through the minimization of a cost function. The cost function included constraints on the periphery dose of the planned target volume, the dose uniformity within the target volume, and the dose to the critical structure. Adjustment between the peripheral dose, the dose uniformity and critical structure dose can be achieved by varying the weighting factors in the cost function. All plans were evaluated in terms of the dose nonuniformity ratio, the conformation number and the dose volume histograms. Among these three GA algorithms, the securGA provided the best performance. Within 2500 function evaluations, the near optimum results were obtained. For a large target volume (5 cm x 4 cm x 4.5 cm) including urethra with 20 needles, the computer time needed for the optimization was less than 5 min on a HP735 workstation. The results showed that once the best set of parameters was found, they were applicable for all sizes of prostate volume. For a fixed needle geometry, the optimized plan showed much better dose distribution than that of nonoptimized plan. If the critical structure was considered in the optimization, the dose to the critical structure could be minimized. In the cases of irregular and skewed needle geometry, the optimized treatment plans were almost as good as ideal needle geometry. It is concluded that this new genetic algorithm (securGA) allows for an efficient and rapid optimization of dose distribution, which is suitable for real time treatment planning optimization for ultrasound-guided prostate implant.     

Optimization of the deposited power distribution inside a layered lossy medium irradiated by a coupled system of concentrically placed waveguide applicators

A method is proposed for controlling the deposited power distribution in a layered cylindrical lossy model, irradiated by a phased-array hyperthermia system consisting of four waveguide applicators. A rigorous electromagnetic model of the heated tissue, which takes into account coupling phenomena between system elements, is used for predicting the electric field at any point inside tissue. The relative amplitudes and relative phases of the array elements are optimized in order to attain desired specific absorption rate (SAR) distributions inside and outside malignant tissues. A constrained nonlinear optimization problem is solved by using the penalty function method and the resulting unconstrained minimization of the penalty function is carried out by the downhill simplex method. Two practical phased-array hyperthermia systems have been studied and numerical results are presented.     

Optimization of permanent 125I prostate implants using fast simulated annealing

PURPOSE: Treatment planning of ultrasound-guided transperineal 125I permanent prostatic implants is a time-consuming task, due to the large number of seeds used and the very large number of possible source arrangements within the target volume. The goal of this work is to develop an algorithm based on fast simulated annealing allowing consistent and automatic dose distribution optimization in permanent 125I prostatic implants. METHODS AND MATERIALS: Fast simulated annealing is used to optimize the dose distribution by finding the best seed distribution through the minimization of a cost function. The cost function includes constraints on the dose at the periphery of the planned target volume and on the dose uniformity within this volume. Adjustment between peripheral dose and the dose uniformity can be achieved by varying the weight factor in the cost function. RESULTS: Fast simulated annealing algorithm finds very good seed distributions within 20,000 iterations. The computer time needed for the optimization of a typical permanent implant involving 60 seeds and 14 needles is approximately 15 min. An additional 5 min are necessary for isodose distribution computations and miscellaneous outputs. CONCLUSION: The use of fast simulated annealing allows for an efficient and rapid optimization of dose distribution. This algorithm is now routinely used at our institution in the clinical planning of 125I permanent transperineal prostate implants for early stage prostatic carcinoma.     

Elastic Constants Identification of Shear Deformable Laminated Composite Plates

A constrained minimization method is presented for the identification of elastic constants of shear deformable laminated composite plates. Strains and∕or displacements obtained from static testing of laminated composite plates are used in the proposed method to identify the elastic constants of the plates. In the identification process, the trial elastic constants of a laminated composite plate are used in a finite-element analysis to predict the strains and displacements of the plate. An error function is established to measure the differences between the experimental and theoretical predictions of strains and∕or displacements. A constrained minimization technique is used to minimize the error function and update the trial elastic constants. The best estimates of the elastic constants of the plate are then determined by subsequently reducing the size of the feasible region of the elastic constants and making the error function a global minimum. The accuracy and applications of the proposed method are demonstrated by means of a number of examples. A sensitivity analysis is also performed to study the effects of variations of experimental data on the accuracy of the identified elastic constants.     

Fast iterative algorithms for three-dimensional inverse treatment planning

Three types of iterative algorithms, algebraic inverse treatment planning (AITP), simultaneous iterative inverse treatment planning (SIITP), and iterative least-square inverse treatment planning (ILSITP), differentiated according to their updating sequences, were generalized to three dimension with true beam geometry and dose model. A rapid ray-tracing approach was developed to optimize the primary beam components. Instead of recalculating the dose matrix at each iteration, the dose distribution was generated by scaling up or down the dose matrix elements of the previous iteration. This significantly increased the calculation speed. The iterative algorithms started with an initial intensity profile for each beam, specified by a two-dimensional pixel beam map of M elements. The calculation volume was divided into N voxels, and the calculation was done by repeatedly comparing the calculated and desired doses and adjusting the values of the beam map elements to minimize an objective function. In AITP, the iteration is performed voxel by voxel. For each voxel, the dose discrepancy was evaluated and the contributing pencil beams were updated. In ILSITP and SIITP, the iteration proceeded pencil beam by pencil beam instead of voxel by voxel. In all cases, the iteration procedure was repeated until the best possible dose distribution was achieved. The algorithms were applied to two examples and the results showed that the iterative techniques were able to produce superior isodose distributions.     

Cardiac parasympathetic stimulation via QRS-synchronous low-energy shocks in humans

PURPOSE: Four different three-dimensional planning techniques for localized radiotherapy of prostate cancer were compared with regard to dose homogeneity within the target volume and dose to organs at risk, dependent upon tumor stage. PATIENTS AND METHODS: Six patients with stage T1, 7 patients with stage T2 and 4 patients with stage T3 were included in this study. Four different 3D treatment plans (rotation, 4-field, 5-field and 6-field technique) were calculated for each patient. Dose was calculated with the reference point at the isocenter (100%). The planning target volume was encompassed within the 95% isodose surface. All the techniques used different shaped portal for each beam. Dose volume histograms were created and compared for the planning target volume and the organs at risk (33%, 50%, 66% volume level) in all techniques. RESULTS: The 4 different three-dimensional planning techniques revealed no differences concerning dose homogeneity within the planning target volume. The dose volume distribution at organs at risk show differences between the calculated techniques. In our study the best protection for bladder and rectum in stage T1 and T2 was achieved by the 6-field technique. A significant difference was achieved between 6-field and 4-field technique only in the 50% volume of the bladder (p = 0.034), between the 6-field and rotation technique (all volume levels) and between 5-field and rotation technique (all volume levels). In stage T1, T2 6-field and 4-field technique in 50% (p = 0.033) and 66% (p = 0.011) of the rectum volume. In stage T3 a significant difference was not observed between the 4 techniques. The best protection of head of the femur was achieved by the rotation technique. CONCLUSION: In the localized radiotherapy of prostate cancer in stage T1 or T2 the best protection for bladder and rectum was achieved by a 3D-planned conformal 6-field technique. If the seminal vesicles have been included in the target volume and in the case of large planning target volume other techniques should be taken for a better protection for organs at risk e. g. a 3D-planned 4-field technique box technique.     

Simplified method for three-dimensional evaluation of interstitial brachytherapy applications

Although three-dimensional (3-D) treatment planning has primarily been used for external beam radiation therapy, the advantages of 3-D treatment planning can be realized for brachytherapy applications. As with teletherapy, the use of 3-D treatment planning for brachytherapy can provide both superior dose distribution as well as detailed evaluations of the relationship of dose and volume in critical structures and target tissues. Conventional 3-D treatment planning uses computed tomography (CT) scans to localize structures; however, localizing individual brachytherapy sources on each CT slice can be impractical for routine clinical use. In the transition from two-dimensional to 3-D localization and dose evaluation of interstitial perineal templates in particular, a practical method of seed localization on a postimplant CT dataset has been developed. This method does not utilize dummy sources and, as such, does not require individual seed locations to be identified. Instead, the position of the afterloading catheter is defined as a reference line by connecting its location as seen on the axial CT slices and seed locations defined along its length. Full volumetric calculations can then be performed, including dose-volume histograms (DVH) for critical organs and tumor volumes. Source localization and normal tissue doses were calculated using both orthogonal films and the 3-D method for a series of perineal template guided implants. Point dose calculations of the rectum and bladder were obtained from orthogonal films and were then compared to the corresponding DVHs for these organs.     

Easy graphical display of beam directions in three-dimensional converging radiation therapy: proposal for a radiation map

In stereotactic radiosurgery, non-coplanar isocentric beams are employed to concentrate the dose distribution on the planning target volume (PTV). However, the directions of incident beams must be determined with great care by using a digitally reconstructed beam's eye view (BEV) to prevent the irradiation of organs at risk. We present a new method of 2-dimensional graphical representation (radiation map) to facilitate the understanding of 3-dimensional relationships between incident beams and critical organs. After determining the isocenter and beam diameter, beam directions and critical organs are projected onto the imaginary sphere centered on the isocenter. The coordinate of the beam directions and the organs at risk can be expressed by latitude and longitude on the sphere. The contours of the organs at risk are displayed with a margin of the the radius of the radiation beam. Mirror images of the critical organs are also displayed to prevent irradiation by the opposing beams. The radiation map could be produced within 5 minutes using a workstation. Radiation maps, like DVH, will be very useful in the evaluation of radiation treatment planning.     

Evaluation of optimized compensators on a 3D planning system

A commercially available treatment planning system contains several functions that allow for the automation of missing tissue and optimized compensators, where the former retracts the bolus toward the source, and the latter attempts, by iteration, to establish a uniform dose at some user defined depth. The intent of this paper is to report on the compensators designed by the system and to compare them to those devised through conventional techniques. It is demonstrated that the system can model the dosimetric effects of compensators with a high degree of accuracy; measured and predicted doses agree to within 3%. Optimized compensators show slightly improved dose uniformity over thickness reduced compensators. Both show significantly improved uniformity over compensators that simply retract the bolus geometry. In cases where internal inhomogeneities exist, however, the dose uniformity from the optimized compensators vary by as much as 6% at the target depth. These deviations are comparable to the errors of the inhomogeneity algorithm itself. The pathlength reduction technique has been applied to both missing tissue and inhomogeneity compensation, and it has been found that for inhomogeneity compensation, the pathlength reduced compensators produce more uniform distributions than those generated by the optimization algorithm.     

Return Mapping Procedure for Frictional Force Calculation: Some Insights

The return mapping procedure is studied to gain insights for its numerical implementation to calculate the frictional force during contact analysis. A simple quasistatic truss–wall frictional contact analysis problem is used in the study. The problem has closed-form solutions which provide exact target solutions for a numerical algorithm. The penalty method and a true augmented Lagrangian method that automatically determines an accurate value of the penalty parameter are employed in the numerical study. It is determined that the return mapping procedure is not applicable unless the contacting node is constrained to stay at the initial contact point, and the total normal reaction force, the tangential reaction force and the friction limit have been determined. If these requirements are not met, inaccurate or even incorrect solutions are obtained. This characteristic of the procedure is studied by solving slip and stick cases with several different load increments. It is concluded that the return mapping procedure for friction force calculation should be implemented carefully to obtain accurate solutions for contact analysis problems.     

Three-dimensional optimization of treatment planning for gamma unit treatment system

During a treatment using the Leksell gamma unit, the physician and physicist need to determine a treatment plan by changing the parameters such as collimator sizes, the position of isocenters and isocenters' weights. This is a complex problem because the set of parameters is large, especially when targets are geometrically close to a critical structure. For this reason, we present here an optimization algorithm, namely the multiplier penalty method, to mathematically determine those parameters. Two cases are presented in this article: the first one is really planned by a physicist in a clinical treatment, and is redone in our optimization algorithm to show the effectiveness of this method; the second one is theoretical where a critical structure is placed close to the target volume. The results show that this method achieves an excellent conformation to the specified isodose curve with the contour of the target volume, allowing minimal damage to surrounding healthy tissue.     

Radiation dose in critical organs due to non-coplanar irradiation of the hypophysis

BACKGROUND: In order to estimate the somatic and genetic risk associated with a non-coplanar linac-based radiation technique of the pituitary gland, systematic secondary-dose measurements in a phantom and sample measurements of the dose near critical organs of patients were performed. PATIENTS AND METHODS: For measurements of the dose outside the primary radiation field an acrylic-PVC phantom was used which was irradiated with a single field (4 x 4 cm2). Eight patients with pituitary tumors were treated isocentrically with a combination of sagittal and transverse rotational arcs. To measure the dose in critical organs. LiF thermoluminescence dosimeters (TLD) in chip form were placed onto 1 eyelid, the skin over the thyroid, and the patient's clothes covering the region of breasts and ovaries of female patients and the testicles of male patients. Measurements were performed for all patients during 1 sagittal irradiation and for the majority of patients during 1 transverse irradiation. RESULTS: The phantom measurements demonstrated that the secondary dose measured on the patients surface can be considered as a good approximation for the dose in adjacent organs. The median dose in critical organs for sagittal irradiation was in the range of 25.8 mGy (eyes) to 1.9 mGy (testicles), and for transverse irradiation in the range of 23.3 mGy (eyes) to 1.3 mGy (testicles). The ratio of median organ doses for sagittal and transverse irradiation was 2.1 for the thyroid gland, 1.1 for the eyes, and 1.5 for the other organs. CONCLUSIONS: The dose in critical organs due to non-coplanar irradiation of the pituitary gland is only a small fraction of the dose delivered to the reference point of the planning target volume. The risk of a radiation-induced tumor and a genetic consequence associated with these small doses is generally less than 1% and 0.1%, respectively.     

Significance of three-dimensional radiation treatment planning

A tumour and its environment constitute a three-dimensional (3D) phenomenon. Consequently, adequate management of the target volume (tumour + safety zone) is feasible only with a 30 treatment planning and irradiating system. The more precise planning and dose delivery involved in such a 30 treatment lead to an improved therapeutic effectiveness.     

Conformal treatment planning for interstitial brachytherapy

PURPOSE: Quality of a brachytherapy application depends on the choice of the target volume, on the dose distribution homogeneity and radiation injury on critical tissue, which should be postulated by advanced brachytherapy treatment planning systems. MATERIAL AND METHODS: Basic imaging method for conformal treatment planning is the cross-sectional imaging. The clinical applicability of a new type 3D planning system using CT and/or MRT-simulation or US-simulation for planning purposes was studied. The planning system developed at Kiel University differs from usual brachytherapy planning systems because of the obligatory use of cross-sectional imaging as basic imaging method for reconstruction of structures of interest. Dose distribution and normal anatomy can be visualized on each CT/MRT/US slice as well as coronal, sagittal, axial and free chosen reconstruction (3D), as well as dose-volume histogram curves and special colour-coded visualization of dose homogeneity in the target can be analyzed. RESULTS: Because of the experience in the clinical routine, as well as on the base of 30 simultaneous planning procedures on both 2D (semi-3D) and 3D planning systems we observed similar time consumption. Advantages of 3D planning were the better interpretation of target delineation, delineation of critical structures as well as dose distribution, causing more accurate volume optimisation of dose distribution. CONCLUSION: Conformal brachytherapy treatment planning for interstitial brachytherapy means significant advantages for the clinical routine compared to 2D or semi-3D methods.     

Optimal Channel Cross Section with Composite Roughness

For channels with composite roughness, an equivalent uniform roughness coefficient and flow geometric elements are used in an optimal design method using the Manning equation. The optimal design problems are formulated in a nonlinear optimization framework with the objective function being a cost function per unit length of the canal. Constraints are the Manning equation, positive values for design variables, and specified values of side slopes or top width. The constrained problem is transformed into an unconstrained problem using the Lagrangian multipliers. To obtain an optimal solution for the resulting unconstrained problem, the first-order necessary conditions for optima are applied. The resulting simultaneous nonlinear equations are solved using the computational methodology developed. This technique is applied to illustrative numerical examples. The evaluations establish the potential applicability of the developed computational methodology for optimal design of open channel cross sections with composite roughness.     

Use of GA to Determine Areas of Accretionto Semiconfined Aquifer

The goal of any groundwater inverse problem is to identify the distribution of an input function or certain other variables describing the unique flow dynamics of an aquifer system. A genetic algorithm (GA) combined with a numerical modeling technique is useful in determining both the spatial distribution and the flux represented by the accretion component of the groundwater flow equation. The GA technique was compared to a modified Gauss-Newton iterative technique. Binary and hexadecimal representations provided mapping of parameters from genetic operators to the numerical model. The technique used the patterns that developed in the string representations to determine probability regions. Two synthetic test cases were used to test the effectiveness of the technique. The stability of the technique was evaluated by introducing random error into the observation data. The technique was capable of locating the accretion area and tended to converge to a flux most representative of the flux entering the aquifer. However, the technique was susceptible to typical problems affecting the inverse problem, such as nonuniqueness.     

Treatment planning for radio-immunotherapy

To foster the success of clinical trials in radio-immunotherapy (RIT), one needs to determine (i) the quantity and spatial distribution of the administered radionuclide carrier in the patient over time, (ii) the absorbed dose in the tumour sites and critical organs based on this distribution and (iii) the volume of tumour mass(es) and normal organs from computerized tomography or magnetic resonance imaging and appropriately correlated with nuclear medicine imaging techniques (such as planar, single-photon emission computerized tomography or positron-emission tomography). Treatment planning for RIT has become an important tool in predicting the relative benefit of therapy based on individualized dosimetry as derived from diagnostic, pre-therapy administration of the radiolabelled antibody. This allows the investigator to pre-select those patients who have 'favorable' dosimetry characteristics (high time-averaged target: non-target ratios) so that the chances for treatment success may be more accurately quantified before placing the patient at risk for treatment-related organ toxicities. The future prospects for RIT treatment planning may yield a more accurate correlation of response and critical organ toxicity with computed absorbed dose, and the compilation of dose-volume histogram information for tumour(s) and normal organ(s) such that computing tumour control probabilities and normal tissue complication probabilities becomes possible for heterogeneous distributions of the radiolabelled antibody. Additionally, radiobiological consequences of depositing absorbed doses from exponentially decaying sources must be factored into the interpretation when trying to compute the effects of standard external beam isodose display patterns combined with those associated with RIT.     

Description and evaluation of a new 3-D computerized treatment planning dose compensator system

Evaluation has been performed of compensators generated by means of a computerized three-dimensional treatment planning system that can utilize either digitized slice profiles or CT scans. Two methods of calculating compensator thickness are used: the modified Batho power law (dSAR) method for digitized profiles and the equivalent TAR (eqTAR) method for CT scans. This system not only compensates for patient surface contours but also compensates for internal inhomogeneities. In addition, any required wedging will be incorporated in the compensator generation. This system has been tested for a number of extreme cases with inhomogeneities and sloping contours. Good agreement was obtained between the measured and computer calculated dose profiles especially along the central axis of the beam. A "Profile Uniformity Index" was defined to quantify the goodness of dose compensation in three dimensions. Compensation using this system can achieve good dose uniformity within the target volume in all clinical cases and is definitely an improvement over systems based solely on tissue deficit.