A data model for handling whole slide microscopy images in picture archiving and communications systems.

Extremely large medical images, like ones of virtual slide microscopy, are beyond some limitations of the DICOM standard (e.g., a 4 Gbyte barrier, caused by 32-bit architecture). Some solutions and trade-offs have been already proposed and included in the DICOM standard (e.g., usage of JPEG2000 image compression standard, JPEG2000 interactive protocol [JPIP] and division of the images into smaller parts for placing them in the PACS). These new features lead to implementation of alternative interaction solutions simultaneously in the same PACS to serve both images of typical size (e.g., radiological) and large size virtual slide microscopy images. The paper deals with problems of (a) constructing a data and interaction model of images within PACS, (b) searching for criteria to assess image content complexity and to reach an efficient division of virtual slide microscopy images into tiles; (c) providing both the conventional DICOM services and the image interchange using JPIP.

Myocardium tissue analysis based on textures in ultrasound images.

The heart disease is related to the alterations in the biological tissue of its muscles. The myocardium alterations have been analysed and evaluated on ultrasound images. The statistical parameters of the heart texture on the defined region of interest present information about the state of myocardium. There are hundreds of the texture parameters, which could be used to evaluate the heart tissue structure. The goal was to choose the most informative ones by applying the quantitative analysis technique. The Fisher statistics has been applied for hypothesis testing to define the most informative texture parameters. The quantitative analysis proved that the texture parameters of the myocardium tissue can be estimated significantly, when the sample size of ROI is not less than 900-1000 pixels. The sample cannot be enhanced by combining images on in the diastolic and systolic phases. To evaluate the state of the myocardium the following statistical parameters can be used on four chamber view images: the percentiles of order 0.01, 0.1, 0.5, grey level run length non-uniformity (in all directions: horizontally, vertically, 45(0), 135(0)), except a lateral wall of the left ventricle, the entropy of the two-dimensional probability distribution, the parameter of the two-dimensional probability distribution, which defines the mean distance between grey levels. On parasternal images the informative statistical parameters are: grey level run length non-uniformity (in all directions: horizontally, vertically, 45(0), 135(0)), grey level non-uniformity, the entropy of the two-dimensional probability distribution applied to the septum only.

Myocardium structure analysis on echocardiographic images.

The heart disease is caused by the alterations in the biological tissue of its muscles. The myocardium alterations have been analysed and evaluated on echocardiographic images. The statistical parameters of the heart texture on the defined region of interest present information about the state of myocardium. There are hundreds of the texture parameters, which could be used to evaluate the heart tissue structure. The goal was to choose the most informative ones by applying the statistical analysis techniques. The Fisher statistics has been applied for hypothesis testing to define the most informative texture parameters. The data for the certain position of a transducer has been compiled for sound and heart patients. Two kinds of echocardiographic images (native and hormonic ones) have been used for analysis. The most informative texture parameters on evaluating changes in myocardium structure proved to be: local mean intensity, the form of local histogram, local variance and run-length nomuniformity. The statistical analysis was limited to the measurements of the most informative position of the transducer from the compiled database.

[Statistical study of the wavelet-based lossy medical image compression technique]. / Statistinis tyrimas medicininiu vaizdu suspaudimui "bangeliu" transformacijos metodu ivertinti.

Medical digital images have informational redundancy. Both the amount of memory for image storage and their transmission time could be reduced if image compression techniques are applied. The techniques are divided into two groups: lossless (compression ratio does not exceed 3 times) and lossy ones. Compression ratio of lossy techniques depends on visibility of distortions. It is a variable parameter and it can exceed 20 times. A compression study was performed to evaluate the compression schemes, which were based on the wavelet transform. The goal was to develop a set of recommendations for an acceptable compression ratio for different medical image modalities: ultrasound cardiac images and X-ray angiographic images. The acceptable image quality after compression was evaluated by physicians. Statistical analysis of the evaluation results was used to form a set of recommendations.

Comparative analysis of cartilage structure on MR and CT images.

The magnetic resonance (MR) imaging device outperforms the quality of CT images for diagnosis related to soft tissues. The cartilage structure is intermediary between soft tissue and bones. The methodology to compare visualisation quality of two different CT and MR techniques was developed. It enabled to evaluate the equipment priority. Before carrying out the comparative analysis of cartilage structure images the method of reducing radiation dose for a CT examination has been developed. It was based on relationship between CT adjustment parameter (reconstruction increment) and image quality. The method enabled to carry out CT examination with a considerably reduced radiation dose for a patient and the medical personnel, taking part in the experiment.