Ioannis Boniatis

Breast Cancer Diagnosis: Analyzing Texture of Tissue Surrounding Microcalcifications

The current study investigates texture properties of the tissue surrounding microcalcification (MC) clusters on mammograms for breast cancer diagnosis. The case sample analyzed consists of 85 dense mammographic images, originating from the digital database for screening mammography. mammograms analyzed contain 100 subtle MC clusters (46 benign and 54 malignant). The tissue surrounding MCs is defined on original and wavelet decomposed images, based on a redundant discrete wavelet transform. Gray-level texture and wavelet coefficient texture features at three decomposition levels are extracted from surrounding tissue regions of interest (ST-ROIs). Specifically, gray-level first-order statistics, gray-level cooccurrence matrices features, and Lawspsila texture energy measures are extracted from original image ST-ROIs. Wavelet coefficient first-order statistics and wavelet coefficient cooccurrence matrices features are extracted from subimages ST-ROIs. The ability of each feature set in differentiating malignant from benign tissue is investigated using a probabilistic neural network. Classification outputs of most discriminating feature sets are combined using a majority voting rule. The proposed combined scheme achieved an area under receiver operating characteristic curve (Az) of 0.989. Results suggest that MCspsila ST texture analysis can contribute to computer-aided diagnosis of breast cancer.

Osteoarthritis severity of the hip by computer-aided grading of radiographic images

A computer-aided classification system was developed for the assessment of the severity of hip osteoarthritis (OA) . Sixty-four radiographic images of normal and osteoarthritic hips were digitized and enhanced. Employing the Kellgren and Lawrence scale, the hips were grouped by three experienced orthopaedists into three OA-severity categories: Normal, Mild/Moderate and Severe. Utilizing custom-developed software, 64 ROIs corresponding to the radiographic Hip Joint Spaces were manually segmented and novel textural features were generated. These features were used in the design of a two-level classification scheme for characterizing hips as normal or osteoarthritic (1st level) and as of Mild/Moderate or Severe OA (2nd level). At each classification level, an ensemble of three classifiers was implemented. The proposed classification scheme discriminated correctly all normal hips from osteoarthritic hips (100% accuracy), while the discrimination accuracy between Mild/Moderate and Severe osteoarthritic hips was 95.7%. The proposed system could be used as a diagnosis decision-supporting tool.

Computer assisted characterization of cervical intervertebral disc degeneration in MRI

A texture-based pattern recognition system is proposed for the automatic characterization of cervical intervertebral disc degeneration from saggital magnetic resonance images of the spine. A case sample of 50 manually segmented ROIs, corresponding to 25 normal and 25 degenerated discs, was analyzed and textural features were generated from each disc-ROI. Students t-test verified the existence of statistically significant differences between textural feature values generated from normal and degenerated discs. This finding is indicative of disc image texture differentiation due to the degeneration of the disc. The generated features were employed in the design of a pattern recognition system based on the Least Squares Minimum Distance classifier. The system achieved a classification accuracy of 94{%} and it may be of value to physicians for the assessment of cervical intervertebral disc degeneration in MRI.

A computer-based system for the discrimination between normal and degenerated menisci from Magnetic Resonance Images

Meniscal myxoid degeneration (MMD) represents a type of degenerative lesion, characterized by histological alterations of the meniscus. In the context of magnetic resonance (MR) imaging evaluation of MMD, the incidence of the condition is indicated by the presence of high intensity signal within the meniscus, while normal menisci are depicted as of homogeneously low intensity. In the present study, a computer based system is proposed for the automatic discrimination between normal and degenerated menisci, employing texture analysis of MR images. The sample of the study consisted of 55 MR images of the knee, corresponding to an equal number of individuals, who were subjected to MR scans. Following a specific protocol T1-weighted sagittal images of the knee joint were acquired, employing a system operating at 1.5 T. The depicted menisci were graded by consensus of two experienced radiologists, employing the scale proposed by Lotysch et al. Accordingly, 15 menisci were characterized as normal (Grade 0) and 40 as degenerated (20 of Grade 1 and 20 of Grade 2). Employing custom developed software a region of interest (ROI), corresponding to the posterior horn of the medial meniscus, was automatically determined on each MR image, on the basis of the region growing segmentation approach. Utilizing custom developed algorithms, a number of textural features, evaluating aspects of spatial variations of pixel intensities, were generated from the segmented ROIs. The calculated features were utilized in the design of a classification system, based on the Bayes classifier. The latter discriminated successfully 49 out of 55 menisci, accomplishing an overall accuracy of 89.1% (specificity accuracy 80%, sensitivity accuracy 92.5%). The proposed system may be of value as a decision support system for the diagnosis of MMD.

The impact of fusion on adjacent levels in cervical spine injuries: Is it really important?

OBJECTIVE Although the literature on degenerative disease of the cervical spine contains numerous articles studying the changes on levels adjacent to a fusion, there exist very few such studies concerning cervical spine stabilization for trauma. METHODS Over a 16-year period (1989-2005), one hundred and twelve patients underwent stabilization of the lower cervical spine (C3-T1) for subaxial cervical spine injuries, either with an anterior or posterior procedure, or both. Eighty-one patients with adequate follow-up were included in the study and 3 groups were identified: Group A, consisting of 8 patients who underwent anterior stabilization and developed Adjacent Level Ossification Development (ALOD), Group B, comprising 53 patients who were anteriorly plated but who did not develop ALOD and Group C, comprising 20 patients who received posterior stabilization. RESULTS Eight out of 61 patients (13.1%) who were anteriorly operated developed ALOD in 11 adjacent levels (Group A). Severe (grade 3) ossification was noted in 6/8 patients at the cranial adjacent level, and in 2/8 patients at the caudal one. Three out of 8 patients presented with early ALOD at 3, 4 and 18 months respectively. Despite the radiographic abnormalities showing ossification, all the patients had an uncomplicated course without symptoms. All the radiographs of Group B and Group C patients demonstrated grade 0 ossification for both the cranial and caudal adjacent levels. CONCLUSION Adjacent-level ossification in cervical spine injuries may appear very early in the postoperative period and it can have a different course than in the degenerative disc disease population, at least in some patients. The first cephalad level adjacent to a fusion appears to be at greater risk. However, even when ALOD is evident radiographically, it very rarely produces any symptoms.

Texture analysis of spinal cord signal in pre- and post-operative T2-weighted magnetic resonance images of patients with Cervical Spondylotic Myelopathy

Cervical spondylotic myelopathy (CSM) represents the most commonly acquired cause of spinal cord dysfunction among individuals over 55 years old. The pathophysiology of the condition involves mechanical factors, which result to injury of the cervical spinal cord. In T-2 weighted magnetic resonance (MR) images of the spine the site of injury is depicted as a region of high intensity signal within the cervical spine cord. The present study aims to investigate whether texture analysis of MR signal in CSM could provide novel quantitative prognostic factors, rendering possible the prognostic estimation of the outcome of a therapeutic surgical intervention for CSM. The sample of the study comprised 12 MR images of the cervical spine, corresponding to 6 CSM patients, who had undergone surgical intervention with anterior cervical discectomy and spinal canal decompression. Following a specific MR imaging protocol a pair of T2-weighted sagittal images of the spine, corresponding to pre- and post-operative MR scans, were obtained for each of the patients. Employing custom developed software, the region of high intensity signal, associated to CSM, was automatically segmented from each MR image. Utilizing custom developed algorithms a number of textural features were extracted from the segmented ROIs and employed in the design of a classification system, based on the Quadratic classifier. The latter was used for the discrimination between pre-operative and post-operative MR images. Statistical analysis revealed the existence of statistically significant differences between textural features, corresponding to pre- and post-operative CSM MR signals. The quadratic classifier characterized correctly all the pre- and post-operative MR images (100% classification accuracy). The results of the present study indicate that textural features, generated from MR images of the spine, may serve as prognostic factors regarding the prediction of the post-operative outcome of CSM patients.

Assessment of osteoarthritis severity by wavelet analysis of the hip joint space radial distance signature

Osteoarthritis (OA) is a major cause of morbidity worldwide, representing the most common form of arthritis. The radiographic assessment of OA-severity is mainly relied on qualitative criteria, evaluating structural alterations of the joint. In the present study a computer-based image analysis method was developed for the grading of hip OA-severity from radiographic images. The sample of the study comprised 64 hips (18 normal, 46 osteoarthritic), corresponding to 32 unilateral and bilateral hip-OA patients. Two experienced orthopaedists assessed OA-severity from pelvic radiographs, employing the Kellgren and Lawrence (KL) grading scale. Accordingly, 3 KL-based OA-severity categories were formed: (i) ldquonormal/doubtfulrdquo, (ii) ldquomild/moderaterdquo, and ldquosevererdquo. After radiographs digitization their contrast was enhanced by means of the contrast limited adaptive histogram equalization method. Employing custom developed algorithms: (i) 64 ROIs, corresponding to patientspsila hip joint spaces (HJSs), were determined on the processed radiographs, and (ii) the radial distance signature (RDS) of each HJS-ROI was generated, as the sequence of the Euclidean radial distances between the ldquocentre of mass (centroid)rdquo and each point of the HJS-ROI contour. The generated RDS was subject to the discrete wavelet transform (Coiflet1 wavelet, Level 2 decomposition). Statistical measures of the generated wavelet coefficients were used for the formation of feature vectors, representative of the HJS-ROIs. These vectors were involved in the design of a grading system, based on the Bayes classifier, which was used for the discrimination between: (i) normal and OA hips, and (ii) hips of ldquoMild / Moderaterdquo and ldquoSevererdquo OA. The classification accuracy achieved regarding the discrimination between normal and OA hips was 95.3%, while the relevant score for the characterization of hips as of ldquomild/moderaterdquo or ldquosevererdquo OA was 91.3%. The proposed system could be of value for the management of hip OA patient.

Texture-based characterization of pre- and post-operative T2-weighted magnetic resonance signals of the cervical spinal cord in cervical spondylotic myelopathy

The utility of texture analysis regarding the provision of quantitative prognostic factors, potentially valuable to the prediction of the post-operative outcome of cervical spondylotic myelopathy (CSM) patients, is investigated. The clinical sample of the study comprised six subjects, who had undergone surgical therapeutic intervention for CSM. Following a specific imaging protocol, a pair of MR images of the cervical spine, corresponding to pre- and post-operative MR scans, was obtained for each of the patients. Accordingly, 12 sagittal T2-weighted magnetic resonance (MR) images were studied. Employing custom developed software, a Region Of Interest (ROI) within the spinal cord, corresponding to the region of the high-intensity CSM MR signal, was segmented on each image, according to the region growing method. Utilizing custom developed algorithms, the following sets of textural features were generated from the segmented ROIs: (i) gradient features, (ii) mean values of features from co-occurrence matrices (co-occurrence features) and (iii) range values of co-occurrence features. Utilizing each one of these sets of features, as well as the least-squares minimum distance and the quadratic classification algorithms, pattern recognition classification schemes were implemented for the discrimination between pre-operative and post-operative MR signals. Statistical analysis revealed the existence of statistically significant differences (p < 0.05) between textural features generated from pre-operative and post-operative high-intensity MR signals. The classification accuracies accomplished ranged from 75% to 100%. Textural features, descriptive of relevant properties of the high-intensity MR signal in CSM, may be considered as quantitative information of potential value for the prediction of the post-operative outcome of CSM patients.

Use of dimensionality reduction for structural mapping of hip joint osteoarthritis data

A visualization-based, computer-oriented, classification scheme is proposed for assessing the severity of hip osteoarthritis (OA) using dimensionality reduction techniques. The introduced methodology tries to cope with the confined ability of physicians to structurally organize the entire available set of medical data into semantically similar categories and provide the capability to make visual observations among the ensemble of data using low-dimensional biplots. In this work, 18 pelvic radiographs of patients with verified unilateral hip OA are evaluated by experienced physicians and assessed into Normal, Mild and Severe following the Kellgren and Lawrence scale. Two regions of interest corresponding to radiographic hip joint spaces are determined and representative features are extracted using a typical texture analysis technique. The structural organization of all hip OA data is accomplished using distance and topology preservation-based dimensionality reduction techniques. The resulting map is a low-dimensional biplot that reflects the intrinsic organization of the ensemble of available data and which can be directly accessed by the physician. The conceivable visualization scheme can potentially reveal critical data similarities and help the operator to visually estimate their initial diagnosis. In addition, it can be used to detect putative clustering tendencies, examine the presence of data similarities and indicate the existence of possible false alarms in the initial perceptual evaluation.

Monitoring of bone regeneration process by means of texture analysis

An image analysis method is proposed for the monitoring of the regeneration of the tibial bone. For this purpose, 130 digitized radiographs of 13 patients, who had undergone tibial lengthening by the Ilizarov method, were studied. For each patient, 10 radiographs, taken at an equal number of postoperative successive time moments, were available. Employing available software, 3 Regions Of Interest (ROIs), corresponding to the: (a) upper, (b) central, and (c) lower aspect of the gap, where bone regeneration was expected to occur, were determined on each radiograph. Employing custom developed algorithms: (i) a number of textural features were generated from each of the ROIs, and (ii) a texture-feature based regression model was designed for the quantitative monitoring of the bone regeneration process. Statistically significant differences (p < 0.05) were derived for the initial and the final textural features values, generated from the first and the last postoperatively obtained radiographs, respectively. A quadratic polynomial regression equation fitted data adequately (r2 = 0.9, p < 0.001). The suggested method may contribute to the monitoring of the tibial bone regeneration process.