C. Choi

Feature extraction algorithm based on adaptive wavelet packet for surface defect classification

This paper proposes a feature extraction method to effectively handle the textural characteristics in images with defects in cold rolled strips. An adaptive wavelet packet scheme is developed to produce the optimum number of features automatically through subband coding gain. Also four classical entropy features in the images with defects are used as local features in the spatial domain. A neural network is used to classify the defects from these features. Experiments with real image data show good training and generalization performances of the proposed method.

Identification of a naturally-occurring 8-[α-D-glucopyranosyl-( 1→6)-β-D-glucopyranosyl] daidzein from cultivated kudzu root

INTRODUCTION Kudzu root (Radix puerariae) is a rich source of isoflavones that are effective in preventing osteoporosis, heart disease and symptoms associated with menopause. The major isoflavonoids in kudzu root extracts were reported as puerarin, daidzin and daidzein. Recently, an unknown isoflavonoid (compound 1) was detected from one-year-old kudzu root cultivated in Vietnam. OBJECTIVE To identify a novel compound 1 in kudzu root extract and determine the structure of the compound by ESI(+) TOF MS-MS, (1)H-, (13)C-NMR and enzymatic hydrolysis. METHODOLOGY Samples were prepared by extraction of one-year-old kudzu root with 50% ethanol and the isoflavonoids were purified using recycling preparative HPLC. Unknown compound 1 was detected using UV-light at 254 nm in TLC and HPLC analyses. The molecular weight of 1 was determined using a TOF mass spectrometer equipped with an electrospray ion source. The structure of 1 was determined from the (13)C and (1)H NMR spectra recorded at 100.40 and 400.0 MHz, respectively. RESULTS ESI(+) TOF MS-MS analysis shows that 1 is a puerarin diglycoside. The interglycosidic linkage of diglycoside determined by (1)H-, (13)C-NMR, and enzymatic hydrolysis suggests that 1 has a glucosyl residue linked to puerarin by an alpha-1,6-glycosidic bond. This compound is the first naturally-occurring 8-[alpha-D-glucopyranosyl-(1-->6)-beta-D-glucopyranosyl]daidzein in kudzu root. The concentration of glucosyl-alpha-1,6-puerarin in kudzu root was 2.3 mg/g as determined by HPLC. CONCLUSION The results indicate that puerarin diglycoside is one of the major isoflavonoids in kudzu root and has a significant impact on the preparation of highly water-soluble glycosylated puerarin.

SU‐E‐T‐609: Treatment Planning Evaluation of Various Modalities for the Tretment of Skin Cancer On Foot

PURPOSE To investigate the characteristics of various treatment modalities for skin cancer on foot. METHODS Right foot was immersed in water reservoir and CT image was taken. Total 6 targets were virtually generated at the skin of foot. Targets are contoured at heel pad (PTV1), inner ankle (PTV2), inner arch (PTV3), dorsal surface (PTV4), inner big toe mound (PTV5) and outer little toe mound (PTV6). Four types of treatment plans using tangential-photon-beams (TPB), electron-beams (EB), high-dose-rate (HDR) brachytherapy technique with Freiburg flap applicator and volumetric-modulated arc therapy (VMAT) technique were generated for each 6 target (single-target plans) and the summed target of every 6 target (multi-target plans). TrueBeam STx™ and Microselectron™ were used for external beam and brachytherapy planning, respectively. Plan normalization was performed to cover 95% of target volume by 95% of prescribed dose. Conformity indices (CI) and homogeneity indices (HI) and the mean doses to the foot except targets were investigated. RESULTS The averaged CI of single-target plans using TPB, EB, HDR and VMAT was 1.97, 2.39, 1.6 and 0.8, respectively. The averaged HI of single-target plans using aforementioned 4 techniques was 1.05, 1.1, 1.52 and 1.04, respectively. For the multi-target, CI was 3.99, 2.39, 1.6 and 0.8. The HI was 1.1, 1.36, 1.43 and 1.04. The averaged mean dose to the foot except target was 2.48 Gy, 2.66 Gy, 3.57Gy and 2.91 Gy for the single-target while it was 21.3 Gy, 14.6 Gy, 14.2 Gy and 13.1 Gy for the multi-target. CONCLUSION The VMAT technique showed better target conformity and homogeneity not only at the single-target plan but also at the multi-target plan than the other techniques to treat skin cancer on foot. In case of single-target treatment, the mean dose to normal tissue was lowest when using tangential-photon-beams while VMAT was best for multi-target treatment.

SU‐E‐T‐372: A New P‐32 Ophthalmic Applicator and Dosimetry

Purpose: Dosimetric characteristics of new P‐32 ophthalmic applicator as substitute for Sr/Y‐90 irradiation were evaluated. The integrity and safety were confirmed. The feasibility for 32P ophthalmic applicators was evaluated. Method and Materials:According to optimal design of P‐32 ophthalmic applicator, Monte Carlo simulations were performed to calculate dosimetric characteristics: reference dose rate, central axis depth doses, transaxial dose profiles. The reference dose rate at the surface of an ophthalmic applicator is measured by using an extrapolation ionization chamber (EC) and radiochromic film (RCF). Depth dose distributions and dose profiles were measured using RCF. The source non‐uniformity was calculated. The source leakage was examined for the safety. The effective shielding of the applicator was verified. Results: The reference dose rates of EC, RCF was 4.13 ± 0.21 cGy/s and 3.84 ± 0.25 cGy/s, respectively. The axial depth dose rate was reduced into approximately 1/10 as 32P betas penetrate every 2 mm depth. Measured data sets in depths of 1 mm to 3 mm agreed with Monte Carlo data. The dose profiles were not uniform due to non‐uniform activity distribution. At the surface (depth=0.1 mm), source non‐uniformities were 27.8%, 11.3% for film, MCNP, respectively. The exposure rate of the handle grip toward a clinician was 1.4 mR/hour (background: 0.1mR/hour). Conclusion: The 32P applicator with a 20 mCi (740 MBq) activity can deliver the therapeutic doses to the surface of the conjunctiva within 6 min, while sparing the lens better than 90Sr/Y applicators. The integrity and safety of an new P‐32 applicator was confirmed. However, prior to the clinical application of every new applicator, safety, dose uniformity, and absorbed dose rate at the reference point should be carefully evaluated by the method developed in this study.

SU-E-T-422: Monte Carlo Simulations for Photon Energy Modulation in Radiotherapy

Purpose: A device to modulate photon energy during beam‐on time was designed and simulated with the Monte Carlo(MC) method. Method and Materials:A device named energy modulator was a cylinder made of aluminum. The frame of that device was connected with 20 tubes through which mercury could be injected or drained. The mercury thickness inside the device could modulate the photon energy of beam penetrated. Monte Carlo simulations of 6 and 10 MV linac using the EGSnrc code were conducted to evaluate the performance of the device. After the design study, a flattening filter of the linac was replaced with the energy modulator in the MC simulations. At least 5 billion histories were followed for each simulation to create phase‐space files at 100 cm SSD. The thickness of mercury inside the energy modulator varied from 2 mm to 40 mm at the field size of 10 cm by 10 cm. The changes of beam quality were expressed in terms of percentage depth dose (PDD) at the depth of 10 cm and 20 cm in water. Results: As increasing the thickness of mercury, the PDD of 6 MV beam at 10 cm and 20 cm was changed from 63.8% to 68.2% and from 34.5% to 40.2%, respectively. The PDD of 10 MV beam was changed from 68.7% to 73.1% at the depth of 10 cm and 40.2% to 45.7% at the depth of 20 cm. Conclusions: The results of this study showed the feasibility of the energy modulation during beam‐on time, which could be advantageous in dose conformity.

SU‐FF‐T‐586: Evaluation of Dose Distribution According to Various CT Slice Thickness

Purpose: We investigated the variation of dose distributions in the treatment plans based on CTimage sets of different slice thicknesses. Method and Materials: An anthropomorphic phantom was scanned with a slice thickness of 1 mm. The acquired images were reconstructed at the slice thicknesses of 2 mm, 3 mm, 5 mm, and 10 mm. For each image set of slice thicknesses, the treatment plan was generated by using a commercial RTP system. Treatment sites were on the breast and H&N (Head & Neck). The dose distributions of 3‐axes (sagittal, transverse, and coronal) were exported at the iso‐center. The image set of 2 mm slice thickness was used as the reference. Dose differences were evaluated in terms of gamma‐index at various criteria. Results: Passing ratio of evaluating gamma‐index decreases with increasing slice thickness of images. The ratios of 3 mm, 5 mm, and 10 mm thicknesses were 96.3%, 95.9%, and 95.3%, respectively on coronal images of H&N and were 100%, 99.9%, and 93.8% on those of breast. On transversal images of both H&N and breast, the ratios were less than 90%. Conclusion: The slice thickness of CTimages affects dose distributions on treatment plans. There might be little dose difference in treatment plans based on CTimage sets of 2 to 5 mm slice thickness. However, such difference becomes considerable in the organ having a steep slope and/or high dose gradients.