Faculty Bibliography

Automatically measuring a muscle’s cross-sectional area is an important application in clinical practice that has been studied extensively in recent years for its ability to assess muscle architecture. Additionally, an adequately segmented cross-sectional area can be used to estimate the echogenicity of the muscle, another valuable parameter correlated with muscle quality. This study assesses state-of-the-art convolutional neural networks and vision transformers for automating this task in a new, large, and diverse database. This database consists of 2005 transverse ultrasound images from four informative muscles for neuromuscular disorders, recorded from 210 subjects of different ages, pathological conditions, and sexes. Regarding the reported results, all of the evaluated deep learning models have achieved near-to-human-level performance. In particular, the manual vs. the automatic measurements of the cross-sectional area exhibit an average discrepancy of less than 38.15 mm2, a significant result demonstrating the feasibility of automating this task. Moreover, the difference in muscle echogenicity estimated from these two readings is only 0.88, another indicator of the proposed method’s success. Furthermore, Bland−Altman analysis of the measurements exhibits no systematic errors since most differences fall between the 95% limits of agreements and the two readings have a 0.97 Pearson’s correlation coefficient (p < 0.001, validation set) with ICC (2, 1) surpassing 0.97, showing the reliability of this approach. Finally, as a supplementary analysis, the texture of the muscle’s visible cross-sectional area was examined using deep learning to investigate whether a classification between healthy subjects and patients with pathological conditions solely from the muscle texture is possible. Our preliminary results indicate that such a task is feasible, but further and more extensive studies are required for more conclusive results.

Despite the success of hand-crafted features in computer visioning for many years, nowadays, this has been replaced by end-to-end learnable features that are extracted from deep convolutional neural networks (CNNs). Whilst CNNs can learn robust features directly from image pixels, they require large amounts of samples and extreme augmentations. On the contrary, hand-crafted features, like SIFT, exhibit several interesting properties as they can provide local rotation invariance. In this work, a novel scheme combining the strengths of SIFT descriptors with CNNs, namely SIFT-CNN, is presented. Given a single-channel image, one SIFT descriptor is computed for every pixel, and thus, every pixel is represented as an M-dimensional histogram, which ultimately results in an M-channel image. Thus, the SIFT image is generated from the SIFT descriptors for all the pixels in a single-channel image, while at the same time, the original spatial size is preserved. Next, a CNN is trained to utilize these M-channel images as inputs by operating directly on the multiscale SIFT images with the regular convolution processes. Since these images incorporate spatial relations between the histograms of the SIFT descriptors, the CNN is guided to learn features from local gradient information of images that otherwise can be neglected. In this manner, the SIFT-CNN implicitly acquires a local rotation invariance property, which is desired for problems where local areas within the image can be rotated without affecting the overall classification result of the respective image. Some of these problems refer to indirect immunofluorescence (IIF) cell image classification, ground-based all-sky image-cloud classification and human lip-reading classification. The results for the popular datasets related to the three different aforementioned problems indicate that the proposed SIFT-CNN can improve the performance and surpasses the corresponding CNNs trained directly on pixel values in various challenging tasks due to its robustness in local rotations. Our findings highlight the importance of the input image representation in the overall efficiency of a data-driven system.

Automatically delineating the deep and superficial aponeurosis of the skeletal muscles from ultrasound images is important in many aspects of the clinical routine. In particular, finding muscle parameters, such as thickness, fascicle length or pennation angle, is a time-consuming clinical task requiring both human labour and specialised knowledge. In this study, a multi-step solution for automating these tasks is presented. A process to effortlessly extract the aponeurosis for automatically measuring the muscle thickness has been introduced as a first step. This process consists mainly of three parts. In the first part, the Attention UNet has been incorporated to automatically delineate the boundaries of the studied muscles. Afterwards, a specialised post-processing algorithm was utilised to improve (and correct) the segmentation results. Lastly, the calculation of the muscle thickness was performed. The proposed method has achieved similar to a human-level performance. In particular, the overall discrepancy between the automatic and the manual muscle thickness measurements was equal to 0.4 mm, a significant result that demonstrates the feasibility of automating this task. In the second step of the proposed methodology, the fascicle's length and pennation angle are extracted through an unsupervised pipeline. Initially, filtering is applied to the ultrasound images to further distinguish the tissues from the other muscle structures. Later, the well-known K-Means algorithm is used to isolate them successfully. As the last step, the dominant angle of the segmented muscle tissues is reported and compared with manual measurements. The proposed pipeline is showing very promising results in the evaluated dataset. Specifically, in the calculation of the pennation angle, the overall discrepancy between the automatic and the manual measurements was less than 2.22° (degrees), once more comparable with the human-level performance. Finally, regarding the fascicle length measurements, the results were divided based on the muscle properties. In the muscles where a large portion (or all) of the fascicles are located between the upper and lower aponeuroses, the proposed pipeline exhibits superb performance; otherwise, overall accuracy deteriorates due to errors caused by the trigonometric approximations needed for the length calculation.

Human assistive technology and computer-aided diagnosis is an emerging field in the area of medical imaging. Following the recent advances in this domain, a study for integrating machine learning techniques in musculoskeletal ultrasonography images was conducted. The goal of this attempt was to investigate how feature extraction techniques, that capture higher-level information, perform in identifying human characteristics. The potential success of these techniques could lead to significant improvement of the current assessment methods-as the gray-scale image analysis-for distinguishing healthy and pathologic conditions, that are heavily dependent on the image-acquisition system. The contribution of this work is threefold. First, a new privately held data set of 74 healthy patients was presented. This data set included musculoskeletal ultrasound images from four muscles of the human body, namely the biceps brachii, tibialis anterior, gastrocnemius medialis and rectus femoris, recorded in the transverse and longitudinal plane. Second, two classification tasks were performed, namely, gender and muscle-type recognition, to assess the performance of the proposed method for successfully identifying differences in the texture of the examined muscle sections. Third, a novel method used with great success in the computer vision domain was presented, allowing the extraction of a high-level feature representation, by encoding the distribution of locally invariant texture descriptors. On the muscle-type recognition our method achieved an 87.07% classification rate, and on the task of gender recognition it surpassed state-of-the-art textural representations, reported in the literature in almost all the examined muscle sections.

The aim of the present study was to design a microscopy image analysis (MIA) system for predicting the 5-year survival of patients with laryngeal squamous cell carcinoma, employing histopathology images of lesions, which had been immunohistochemically (IHC) stained for p63 expression. Biopsy materials from 42 patients, with verified laryngeal cancer and follow-up, were selected from the archives of the University Hospital of Patras, Greece. Twenty six patients had survived more than 5 years and 16 less than 5 years after the first diagnosis. Histopathology images were IHC stained for p63 expression. Images were first processed by a segmentation method for isolating the p63-expressed nuclei. Seventy-seven features were evaluated regarding texture, shape, and physical topology of nuclei, p63 staining, and patient-specific data. Those features, the probabilistic neural network classifier, the leave-one-out (LOO), and the bootstrap cross-validation methods, were used to design the MIA-system for assessing the 5-year survival of patients with laryngeal cancer. MIA-system accuracy was about 90 % and 85 %, employing the LOO and the Bootstrap methods, respectively. The image texture of p63-expressed nuclei appeared coarser and contained more edges in the 5-year non-survivor group. These differences were at a statistically significant level (p < 0.05). In conclusion, this study has proposed an MIA-system that may be of assistance to physicians, as a second opinion tool in assessing the 5-year survival of patients with laryngeal cancer, and it has revealed useful information regarding differences in nuclei texture between 5-year survivors and non-survivors.

BACKGROUND:P63 immunostaining has been considered as potential prognostic factor in laryngeal cancer. Considering that P63 is mainly nuclear stain, a possible correlation between the texture of P63-stained nuclei and the tumor's grade could be of value to diagnosis, since this may be related to biologic information imprinted as texture on P63 expressed nuclei. OBJECTIVE:To investigate the association between P63 stained nuclei and histologic grade in laryngeal tumor lesions. METHODS:Biopsy specimens from laryngeal tumour lesions of 55 patients diagnosed with laryngeal squamous cell carcinomas were immunohistochemically (IHC) stained for P63 expression. Four images were digitized from each patient's IHC specimens. P63 positively expressed nuclei were identified, the percentage of P63 expressed nuclei was computed, and 118 textural, morphological, shape, and architectural features were calculated from each one of the 55 laryngeal lesions. Data were split into the low grade (21 grade I lesions) and high grade (34 grade II and grade III lesions) classes for statistical analysis. RESULTS:With advancing grade, P63 expression decreased, P63 stained nuclei appeared of lower image intensity, more inhomogeneous, of higher local contrast, contained smaller randomly distributed dissimilar structures and had irregular shape. CONCLUSION/CONCLUSIONS:P63 expressed nuclei contain important information related to histologic grade.

OBJECTIVE:To design a pattern recognition (PR) system for discriminating between low- and high-grade laryngeal cancer cases, employing immunohistochemically stained, for p63 expression, histopathology images. STUDY DESIGN/METHODS:The PR system was designed to assist in the physician's diagnosis for improving patient survival. The material comprised 55 verified cases of laryngeal cancer, 21 of low-grade and 34 of high-grade malignancy. Histopathology images were first processed for automatically segmenting p63 expressed nuclei. Fifty-two features were next extracted from the segmented nuclei, concerning nuclei texture, shape, and physical topology in the image. Those features and the Probabilistic Neural Network classifier were used to design the PR system on the multiprocessors of the Nvidia 580 GTX graphics processing unit (GPU) card using the Compute Unified Device Architecture parallel programming model and C++ programming language. RESULTS:PR system performance in classifying laryngeal cancer cases as low grade and high grade was 85.7% and 94.1%, respectively. The system's overall accuracy was 90.9%, using 7 features, and its estimated accuracy to "unseen" by the system cases was 80%. CONCLUSION/CONCLUSIONS:Optimum system design was feasible after employing parallel processing techniques and GPU technology. The proposed system was structured so as to function in a clinical environment, as a research tool, and with the capability of being redesigned on site when new verified cases are added to its repository.

Bottom and fly ash samples (BASH and FASH) from the APOTEFROTIRAS S.A. medical waste incinerator (Athens, Greece) were investigated. Powder-XRD data and geochemical diagrams showed BASH to be an amorphous material, analogous to basaltic glass, and FASH consisting of crystalline compounds (mainly CaClOH). Bulk analyses by ICP-MS and point analyses by SEM-EDS indicated a high content of heavy metals, such as Fe, Cu and Cr, in both samples. However, BASH was highly enriched in Ni while FASH was additionally enriched in Zn and Pb. Gamma-ray measurements showed that the radioactivity of both ash samples, due to natural and artificial radionuclides ((137)Cs, (57)Co), was within the permissible levels recommended by IAEA. According to EN-type leaching tests, BASH was practically inert with regard to the mobility of the hazardous elements in aqueous media. FASH, however, showed a relatively high EN (and TCLP) leachability with regard to Pb and Zn. Finally, the stabilisation method, suggested for the treatment of FASH, included compression of the powder into briquettes using an appropriate machine and embedding the briquettes into pozzolanic cement blocks. After this treatment, TCLP and EN-type tests showed minimal release of Pb and Zn, thereby demonstrating a reliable management of ash waste.

Hyperthermic isolated limb perfusion with TNF-alpha and melphalan (TM-HILP) is a complicated surgical procedure. Herein, we present the experience of the Hellenic collaborating centers with TM-HILP for inoperable in-transit melanoma and soft tissue sarcoma (STS) of the extremities to examine safety and feasibility of collaborating as a multi-institutional group for future research studies. From 2001 to 2009, twenty patients (median age 63.5 years) underwent TM-HILP for locally advanced in-transit melanoma (n=14) or unresectable STS (n=6). All patients underwent a 90-min isolated limb perfusion with melphalan (10 mg/l limb volume) and TNF-alpha (1-2 mg) under mild hyperthermia (39-40 degrees C). No major intra-operative complications occurred and all patients completed the procedure successfully. One patient developed postoperative ischemic necrosis of the limb necessitating amputation. All melanoma patients showed a response to TM-HILP with 7 (62%) of them experiencing complete response. All STS patients attained complete response after excision of residual tumor. The median disease specific and limb-relapse-free survival was 15 and 12 months, respectively. TM-HILP can be safely applied even in low volume tertiary hospitals provided that technology to minimize intraoperative systemic leakage is available. Future prospective studies can be performed reproducibly by this multi-institutional collaborative group.

This paper proposes a methodology that incorporates principles from cluster analysis and graph representation to achieve efficient image segmentation results. More specifically, a feature-based, inter-region dissimilarity relation is considered here in order to determine the dissimilarity matrix in a graph-based segmentation scheme. The calculation of the dissimilarity function between adjacent elementary image regions is based on the proximity of each region's feature vector to the main clusters that are formed by the image samples in the feature space. In contrast to typical segmentation approaches of the literature, the global feature space information is included in the spatial graph representation that was derived from the initial Watershed partitioning. A region grouping process is applied next to form the final segmentation results. The proposed approach was also compared to approaches that use feature-based, or spatial information exclusively, to indicate its effectiveness.