Evaluation of partitional algorithms for clustering medical documents

There are large quantities of information about patients and their medical conditions. The discovery of trends and patterns hidden within the data could significantly enhance understanding of disease and medicine progression and management by evaluating stored medical documents. Methods are needed to facilitate discovering the trends and patterns within such large quantities of medical documents. Clustering medical documents into small number of meaningful clusters is one of these methods; because dealing with only the cluster that will contain relevant documents should improve effectiveness and efficiency. The produced clusters must be in high-quality because it will be used for further processing to discover the hidden trends and patterns. The focus of this paper is to experimentally evaluate the clusters' quality of partitional clustering algorithms that use different criterion functions in the context of clustering medical documents. Our experimental results show that E1 leads to the best solution using repeated bisection as clustering method in term entropy. And I1 is the best using direct clustering methods in term of both entropy and purity

Portal-image IMRT verification

The aim of the present work was to develop and utilizing the use of an amorphous silicon electronic portal imaging device [EPID] in IMRT dosimetric verification. Neither pre-irradiation nor extra build-up materials was needed for EPID dosimetry applications. Accurate absolute dosimetry [output] obtained using the EPID up to 250 cGy with mean deviation of 0.39% and 0.62% while the maximum observed deviation was 0.7% and 1.9% at 250 MUs for 6 and IS MV photon energy respectively. Beam-by-beam fluence profiles obtained from portal images were used in air [i.e. without phantom presence] The EPID estimated the relative dose up to 1.5% accuracy. The in-air absolute dose [i.e. number of MUs] of an arbitrary clinical breast [aperture-based] IMRT test fields were extracted, the difference rose to 1.7% in the most severe tested field. One disadvantage of beam-by-beam verification is that the cumulative effect of dose errors from all beams is not quantified, however, it allows the potential origin of dose errors to be isolated more easily. The EPID could estimate the dose at each segment with average accuracy of around 1% for central axis positions and up to 1.7% for off-axis positions in the tested fields. The absolute dose verification plus the fluence map verification of IMRT fields may represent a sufficient procedure to examine the step-and shoot-IMRT treatment. Multi-leaf-collimator [MLC] related QA also tested using EPED. The mean difference between EPID effective penumbra results and both ion chamber and film measurements was 0.06 and 0.01 cm respectively. These results could justify the use of EPID in dosimetric applications including aperture-based IMRT verification, and quality control programs

MLC leaf position and dynalog file accuracy QA procedure using the EPID

A main concern about the IMRT dose validation tool using Monte Carlo [MC] simulation and R and V-system/Dynalog file is the potential inconsistency between the actual leaf-end positions and those recorded by the Dynalog file. The present study investigates a method to validate the accuracy of the Dynalog tiles using amorphous silicon electronic portal imaging device [aSi-EPID] images. A computer program was developed to detect the MLC segmented field edges in EPID images [l024x768 pixels, pixel size: 0.392 mm], Standard reference MLC segmented fields were designed and leaf-end positions were measured accurately. EPID images for these reference MLC fields were recorded and the leaf-end positions were calculated as the locations where the image intensity is 50% of the maximum. Small corrections were made to minimize the effect of scattered photons [background]. Daily EPII] images of the same MLC segmented fields were compared to the original images and to check the accuracy of the Dynalog files. A daily QA tool was developed to check the accuracy of the Dynalog file and MLC leaf end positions as part of the comprehensive IMRT-QA procedure. This ensures the accuracy of the MC based patient-specific IMRT dose verification using the information recorded in the Reeord and Verify system/Dynalog files