ROI Selection Research Articles

Prognostic Analysis Combining Histopathological Features and Clinical Information to Predict Colorectal Cancer Survival from Whole-Slide Images.

Colorectal cancer (CRC) is a malignant tumor within the digestive tract with both a high incidence rate and mortality. Early detection and intervention could improve patient clinical outcomes and survival. This study computationally investigates a set of prognostic tissue and cell features from diagnostic tissue slides. With the combination of clinical prognostic variables, the pathological image features could predict the prognosis in CRC patients. Our CRC prognosis prediction pipeline sequentially consisted of three modules: (1) A MultiTissue Net to delineate outlines of different tissue types within the WSI of CRC for further ROI selection by pathologists. (2) Development of three-level quantitative image metrics related to tissue compositions, cell shape, and hidden features from a deep network. (3) Fusion of multi-level features to build a prognostic CRC model for predicting survival for CRC. Experimental results suggest that each group of features has a particular relationship with the prognosis of patients in the independent test set. In the fusion features combination experiment, the accuracy rate of predicting patients' prognosis and survival status is 81.52%, and the AUC value is 0.77. This paper constructs a model that can predict the postoperative survival of patients by using image features and clinical information. Some features were found to be associated with the prognosis and survival of patients.

Robust Remote Photoplethysmography Estimation With Environmental Noise Disentanglement.

Remote Photoplethysmography (rPPG) has been attracting increasing attention due to its potential in a wide range of application scenarios such as physical training, clinical monitoring, and face anti-spoofing. On top of conventional solutions, deep-learning approach starts to dominate in rPPG estimation and achieves top-level performance. However, most of them try to integrate preprocessing steps such as the ROI selection into an end-to-end network, which may diverge the attention and also limit the generalization in other scenarios with different input skin regions. In this work, we focus on learning the intrinsic rPPG feature and design a lightweight but effective rPPG estimation network based on spatiotemporal convolution. To further improve the robustness, on top of the basic design we propose the Noise-Disentangled DeeprPPG (ND-DeeprPPG) by disentangling the environmental noise from the raw rPPG feature with an adversarial canonical correlation analysis learning strategy. Background regions are employed as a reference to guide the noise disentangling in a self-supervised manner. Extensive experiments show that our ND-DeeprPPG not only outperforms the state-of-the-arts on heart rate estimation but also exhibits promising robustness in cross-skin-region, cross-dataset scenarios and other rPPG-based tasks.

Implementation of Thermal Camera for Human Stress Detection: A Review

Stress has become a major problem that people face today. The high level of competition and environmental demands make people more susceptible to stress. Stress can interfere with a person's ability to work effectively. If left unchecked for a long time, stress can cause various dangerous diseases such as hypertension, heart problems, and others that can lead to death. Research has been conducted for a long time to detect stress. Various technologies have been used to detect and anticipate stress that occurs in humans. One promising technology for detecting stress is the use of thermal cameras. Thermal cameras have several advantages: being non-contact and non-invasive, quick, easy to use, and cost-effective. In general, the architecture of the stress detection system using a thermal camera consists of several stages, including image acquisition, pre-processing, ROI tracking and selection, feature extraction, and statistical analysis or classification. This paper aims to review the use of thermal cameras in detecting stress in humans. This paper also seeks to answer the research question of what analysis can be done to improve stress detection accuracy using thermal camera images. Research shows that ROI selection must be carefully considered to obtain good accuracy. Combining thermal images with other data can improve accuracy in stress detection. Machine learning in classification provides many benefits in recognizing patterns but is highly influenced by the number of datasets used.

Investigating neuronal senescence marker, p19INK4d, in mouse and human brains

AbstractBackgroundThe cellular senescence stress response allows damaged, toxic cells to survive. Clearing senescent cells is a promising therapeutic approach and has moved to clinical trials for Alzheimer’s disease and related dementias (ADRD). We previously identified a link between the cell cycle inhibitor CDKN2D/p19INK4d and neuronal senescence in human postmortem brains with various levels of AD neuropathology. We hypothesized that senescence could be induced by overexpressing this gene in mouse neurons in vivo.MethodAdeno‐associated virus 9 (AAV9) with the neuron specific human Synapsin I (hSyn1) promoter was used to deliver experimental Cdkn2d‐GFP, or control GFP gene expression. Precise delivery of 0.5µl of 1013 IU virus into four neuronal populations per mouse, bilateral hippocampal CA1 and cortical layer 5 neurons was accomplished with the assistance of a stereotaxic robot. Male and female wild type, human tau transgenic and mouse tau knockout mice were used (12‐13 months‐old, n = 47). Mice were euthanized at either two weeks or eight weeks post‐surgery to determine the short‐ and long‐term effects of Cdkn2d expression. GeoMx digital spatial profiler was utilized to quantify protein expression between Cdkn2d infected and uninfected neighboring neurons. ROI selection is displayed in Figure 1.ResultHistological analyses confirmed p19INK4d protein expression in AAV9‐infected cells. GeoMx DSP analyses performed on hippocampus, cortex, and thalamus revealed a protein expression pattern consistent with senescence (i.e., resistance to apoptosis), in Cdkn2d‐expressing neurons of the hippocampus and cortex compared to the un‐infected thalamus (Figure 2). Moreover, proteins associated with apoptosis were higher in the thalamus compared to the cortex and hippocampus.ConclusionsThe data indicates that AAV9‐hSyn1‐Cdkn2d‐GFP is an appropriate method to increase p19INK4d expression in mouse neurons in vivo. Over‐expressing p19INK4d caused neurons to acquire a protein expression signature consistent with senescence. The concomitant expression of apoptosis proteins in the thalamus leads me to hypothesize that the senescent cells in the cortex may cause cell death in the thalamus through the corticothalamic tract. Further studies with increased sample size are needed to confirm the results and continue exploring the effects of Cdkn2d on neuronal senescence, and how tau protein may influence this stress response.

Prediction of explosives by a de-broadening model based on RBF neural network

The de-broadening algorithm based on RBFNN is proposed to solve the problems of overlapping peaks and ROI selection in the traditional peak area ratio method. This algorithm is combined with the tagged neutron method for the detection of hidden explosives inside packages and Geant4 is applied for modeling the experimental setup. The broadened energy spectra from 3.5 MeV to 6.8 MeV are chosen as input for the prediction of the unbroadened energy spectra from 3.5 MeV to 6.5 MeV. The result shows that the de-broadening algorithm improves the degree of separation of the samples compared to the peak area ratio method. In the reliability validation of the model, the trained RBFNN can effectively predict the samples within the database and has 99 % identification accuracy in a blind experiment. Meanwhile, the established model predicts other kinds of samples, different masses of samples, and untrained experimental spectra, the good prediction results verify the universality of the model.

Lane Reconstruction for Self-Driving Vehicles on Dynamic Road Networks

Lane detection is a crucial task that involves identifying lane lines and proper markings on road structures. The field of lane detection has gained significant attention due to the increasing use of autonomous vehicles and ADAS systems, as they require robust lane detection systems to navigate safely in complex environments. While previous studies have focused on lane detection in various scenarios, lane line reconstruction for dynamic road networks has been largely overlooked. To address this gap, we propose a novel approach to reconstruct lane lines for dynamic road networks using roadblocks, specifically red and white striped markers, as cues. To evaluate the effectiveness of our proposed approach, we first built and labeled our own dataset by extracting 1854 frames from recorded videos. We trained a Faster-RCNN V2 model on 80% of the dataset to detect the roadblocks and used the OpenCV DNN module, along with a range of techniques, such as ROI selection, color selection, region masking, and polynomial and linear fittings for lane detection and reconstruction. Our approach successfully reconstructed the left, middle, and right lane lines, achieving an overall accuracy of 68%, a recall of 73%, and a precision score of 73% on an unseen set of 50 images. Our study contributes to the field by proposing a novel approach for reconstructing lane lines in dynamic road networks using roadblocks as markers. This approach has the potential to enhance the accuracy and robustness of lane detection systems in complex environments. Furthermore, our work addresses a gap in the existing literature and provides insights for future research in this area.

Influence of different factors on registration error in a 1.5 T MR-guided linac

Purpose. Accurate image registration is an important step in online image-guided adaptive radiotherapy. The aim of this study was to investigate the effects of different factors on registration accuracy in a magnetic resonance (MR)-guided adaptive radiotherapy workflow. Materials and Methods. A thorax motion phantom was used to obtain computed tomography (CT) simulations in 8 different motion modes and to generate 8 reference plans. Daily pretreatment online MR images were obtained at 5 different positions in each reference plan. Online MR and CT simulations were separately registered using bone structures and the gross tumor volume (GTV) as ROIs, and the image shift distance was recorded by the online treatment planning system. The difference between the shift distance and the real isocentric distance was the registration error. The registration error was analyzed, and the effects of the setup position, motion mode and ROI selection on the registration error were investigated by multivariate analysis of variance. Result. The minimum values of registration error (ΔX, ΔY, ΔZ) were −1.90 mm, −2.70 mm and −2.40 mm, respectively, and the maximum values were 1.70 mm, 4.30 mm and −0.90 mm. ΔY showed the maximum mean standard deviation of 1.25 mm, and ΔZ showed the minimum mean standard deviation of 0.27 mm. The standard deviation of the registration error is largest in the inferior/superior direction. The motion mode of the phantom and ROI selection were significantly correlated with ΔX, ΔY, and ΔZ (p < 0.05). Conclusion. The registration result with the spine as the selected ROI was better than that with the GTV as the ROI. In 1.5 T MR-linac clinical treatment, more attention should be given to patient movement repeatability and to controlling the intrafractional motion as much as possible. It is not recommended to make the GTV-PTV margin expansion less than 2 mm for MR-linac.

A secure framework for medical image by integrating watermarking and encryption through fuzzy based ROI selection

Security, secrecy, and authenticity problems have arisen as a result of the widespread sharing of medical images in social media. Copyright protection for online photo sharing is becoming a must. In this research, a cutting-edge method for embedding encrypted watermarks into medical images is proposed. The proposed method makes use of fuzzy-based ROI selection and wavelet-transformation to accomplish this. In the first step of the process, a fuzzy search is performed on the original picture to locate relevant places using the center region of interest (RoI) and the radial line along the final intensity. The suggested method takes a digital picture and divides it into 4×4 non-overlapping blocks, with the intent of selecting low information chunks for embedding in order to maximize invisibility. By changing the coefficients, a single watermark bit may be inserted into both the left and right singular SVD matrices. The absence of false positives means the suggested technique can successfully integrate a large amount of data. Watermarks are encrypted using a pseudorandom key before being embedded. Discrete wavelet transform saliency map, block mean method, and cosine functions are used to construct an adaptively-generated pseudo-random key from the cover picture. Images uploaded to social media platforms must have a high degree of invisibility and durability. These watermarking features, however, come with a price. The optimal scaling factor is used to strike a balance between the two in the proposed system. Furthermore, the suggested scheme’s higher performance is confirmed by comparison with the latest state-of-the-art systems.

Stability of mental motor-imagery classification in EEG depends on the choice of classifier model and experiment design, but not on signal preprocessing.

Modern consciousness research has developed diagnostic tests to improve the diagnostic accuracy of different states of consciousness via electroencephalography (EEG)-based mental motor imagery (MI), which is still challenging and lacks a consensus on how to best analyse MI EEG-data. An optimally designed and analyzed paradigm must detect command-following in all healthy individuals, before it can be applied in patients, e.g., for the diagnosis of disorders of consciousness (DOC). We investigated the effects of two important steps in the raw signal preprocessing on predicting participant performance (F1) and machine-learning classifier performance (area-under-curve, AUC) in eight healthy individuals, that are based solely on MI using high-density EEG (HD-EEG): artifact correction (manual correction with vs. without Independent Component Analysis [ICA]), region of interest (ROI; motor area vs. whole brain), and machine-learning algorithm (support-vector machine [SVM] vs. k-nearest neighbor [KNN]). Results revealed no significant effects of artifact correction and ROI on predicting participant performance (F1) and classifier performance (AUC) scores (all ps > 0.05) in the SVM classification model. In the KNN model, ROI had a significant influence on the classifier performance [F(1,8.939) = 7.585, p = 0.023]. There was no evidence for artifact correction and ROI selection changing the prediction of participants performance and classifier performance in EEG-based mental MI if using SVM-based classification (71-100% correct classifications across different signal preprocessing methods). The variance in the prediction of participant performance was significantly higher when the experiment started with a resting-state compared to a mental MI task block [X2(1) = 5.849, p = 0.016]. Overall, we could show that classification is stable across different modes of EEG signal preprocessing when using SVM models. Exploratory analysis gave a hint toward potential effects of the sequence of task execution on the prediction of participant performance, which should be taken into account in future studies.

A deep learning based four-fold approach to classify brain MRI: BTSCNet

Incorrect diagnosis of brain tumor types prevent appropriate response to medical assistance and reduces patients' chances of survival. Examining MRI images of the patient's brain manually is one traditional method for distinguishing brain tumors, but it is time consuming and prone to human errors. Thus, an automated and new deep learning based four-fold method, Brain Tumor Segmentation and Classification Network (BTSCNet), is proposed in this article which will be helpful for physicians to classify three types of brain tumors (meningioma, glioma, and pituitary tumor) from T1-weighted contrast-enhanced MRI (CE-MRI) images properly. The proposed four folds includes: segmentation of brain tumor region using Brain Tumor Segmentation Network (BTSNet), ROI selection using morphological operation, feature extraction using multi-region gray level co-occurrence matrix (MR-GLCM), andclassification using Sliding Window Euclidean distance (SWED) measure. The lack of annotated training samples is the main challenge in deep-learning-based brain MRIimage classification. Thus, while training the proposed system, the scale, orientation and flip of the input image is randomly changed in the first fold of BTSCNetso that the network can be trained with varied or augmentedimages with respect toscale, orientation and flip.Four performance measures are used to measure the efficiency of the proposed method. The output of the proposed method is tested on a public database where the correct classification rates obtained are96.6% (meningioma), 98.1% (glioma), and 95.3% (pituitary tumor), when considering MR-Contrast feature, MR-Correlation and MR-Homogeneity feature respectively, which indicates that the efficiency of the proposed method.

Deep learning neural network image analysis of immunohistochemical protein expression reveals a significantly reduced expression of biglycan in breast cancer.

New breast cancer biomarkers have been sought for better tumor characterization and treatment. Among these putative markers, there is Biglycan (BGN). BGN is a class I small leucine-rich proteoglycan family of proteins characterized by a protein core with leucine-rich repeats. The objective of this study is to compare the protein expression of BGN in breast tissue with and without cancer, using immunohistochemical technique associated with digital histological score (D-HScore) and supervised deep learning neural networks (SDLNN). In this case-control study, 24 formalin-fixed, paraffin-embedded tissues were obtained for analysis. Normal (n = 9) and cancerous (n = 15) tissue sections were analyzed by immunohistochemistry using BGN monoclonal antibody (M01-Abnova) and 3,3'-Diaminobenzidine (DAB) as the chromogen. Photomicrographs of the slides were analysed with D-HScore, using arbitrary DAB units. Another set (n = 129) with higher magnification without ROI selection, was submitted to the inceptionV3 deep neural network image embedding recognition model. Next, supervised neural network analysis, using stratified 20 fold cross validation, with 200 hidden layers, ReLu activation, and regularization at α = 0.0001 were applied for SDLNN. The sample size was calculated for a minimum of 7 cases and 7 controls, having a power = 90%, an α error = 5%, and a standard deviation of 20, to identify a decrease from the average of 40 DAB units (control) to 4 DAB units in cancer. BGN expression in DAB units [median (range)] was 6.2 (0.8 to 12.4) and 27.31 (5.3 to 81.7) in cancer and normal breast tissue, respectively, using D-HScore (p = 0.0017, Mann-Whitney test). SDLNN classification accuracy was 85.3% (110 out of 129; 95%CI = 78.1% to 90.3%). BGN protein expression is reduced in breast cancer tissue, compared to normal tissue.

Robust Heart Rate Measurement by Adaptive ROI Selection for Head-Rotation Conditions

Robust Heart Rate Measurement by Adaptive ROI Selection for Head-Rotation Conditions

Diffusion kurtosis imaging and intravoxel incoherent motion imaging parameters in breast lesions: Effect of radiologists’ experience and region-of-interest selection

PurposeTo investigate the influence of ROI placement methods and radiologists’ experience on diffusion kurtosis imaging (DKI) and intravoxel incoherent motion (IVIM) parameters’ diagnostic performance in differentiating benign and malignant lesions based on the mass and non-mass enhancement (NME). MethodsWe evaluated 138 lesions in 131 patients retrospectively. The IVIM and DKI parameter values were measured by three radiologists with different experiences independently using two different ROI placement methods. IVIM parameters include diffusion coefficient (ADCstand), true diffusion coefficient (ADCslow), pseudo-diffusion coefficient (ADCfast) and perfusion fraction (f). DKI parameters include mean diffusivity (MD) and mean kurtosis (MK). Each radiologist measured the lesions twice with a 3-month interval. We utilized intra-class correlation (ICC) to determine the inter- and intra-reader agreement for mass and NME, respectively. ROC analysis compared the diagnostic performance of parameters between different radiologists, ROI methods, and between mass and NME. ResultsIn mass lesions, inter- and intra-observer agreement were perfect for all parameters (ICC: 0.800–989). In NME, the inter-observer agreement was substantial to perfect for all parameters(ICC: 0.703–877), the intra-observer agreement of the senior and intermediate radiologists was substantial to perfect(ICC: 0.748–931) and the intra-observer agreement of the junior radiologist was moderate to substantial(ICC: 0.569–784). The diagnostic performance of ADCslow (Z = 2.209, P = 0.023), MD (mean diffusivity) (Z = 2.887, P = 0.004), and MK (mean kurtosis) (Z = 2.080, P = 0.038) in the small ROI measured by the senior radiologist was better than that of the junior radiologist for NME. The diagnostic performance of ADCslow in the large ROI measured by the senior radiologist (Z = 2.281, P = 0.023) and intermediate radiologist (Z = 2.867, P = 0.0041) was better than the junior radiologist for mass lesions. The diagnostic performance of ADCslow, ADCstand, MD, and MK did not show a significant difference between the two ROI placement methods (P > 0.05). ConclusionThe observers’ experience can influence the ROI selection and the diagnostic performance of ADCslow, ADCstand, MD, and MK measured using different methods show equal diagnostic performance.

Radiomics in Abdominopelvic Solid-Organ Oncologic Imaging: Current Status.

Radiomics is the process of extraction of high-throughput quantitative imaging features from medical images. These features represent noninvasive quantitative biomarkers that go beyond the traditional imaging features visible to the human eye. This article first reviews the steps of the radiomics pipeline, including image acquisition, ROI selection and image segmentation, image preprocessing, feature extraction, feature selection, and model development and application. Current evidence for the application of radiomics in abdominopelvic solid-organ cancers is then reviewed. Applications including diagnosis, subtype determination, treatment response assessment, and outcome prediction are explored within the context of hepatobiliary and pancreatic cancer, renal cell carcinoma, prostate cancer, gynecologic cancer, and adrenal masses. This literature review focuses on the strongest available evidence, including systematic reviews, meta-analyses, and large multicenter studies. Limitations of the available literature are highlighted, including marked heterogeneity in radiomics methodology, frequent use of small sample sizes with high risk of overfitting, and lack of prospective design, external validation, and standardized radiomics workflow. Thus, although studies have laid a foundation that supports continued investigation into radiomics models, stronger evidence is needed before clinical adoption.

Spatial Profiling of Protein and RNA Expression in Tissue: An Approach to Fine-tune Virtual Microdissection

Multiplexing enables the assessment of several markers on the same tissue while providing spatial context. Spatial Omics technologies allow both protein and RNA multiplexing by leveraging photo-cleavable oligo-tagged antibodies and probes, respectively. Oligos are cleaved and quantified from specific regions across the tissue to elucidate the underlying biology. Here, the study demonstrates that automated custom antibody visualization protocols can be utilized to guide ROI selection in conjunction with spatial proteomics assays. This specific method did not show acceptable performance with spatial transcriptomics assays. The protocol describes the development of a 3-plex immunofluorescent (IF) assay for marker visualization on an automated platform, using tyramide signal amplification (TSA) to amplify the fluorescent signal from a given protein target and increase the antibody pool to choose from. The visualization protocol was automated using a thoroughly validated 3-plex assay to ensure quality and reproducibility. In addition, the exchange of DAPI for SYTO dyes was evaluated to allow imaging of TSA-based IF assays on the spatial profiling platform. Additionally, we tested the ability of selecting small ROIs using the spatial transcriptomics assay to allow the investigation of highly-specific areas of interest (e.g., areas enriched for a given cell type). ROIs of 50 µm and 300 µm diameter were collected, which corresponds to approximately 15 cells and 100 cells, respectively. Samples were made into libraries and sequenced to investigate the capability to detect signals from small ROIs and profile-specific regions of the tissue. We determined that spatial proteomics technologies highly benefit from automated, standardized protocols to guide ROI selection. While this automated visualization protocol was not compatible with spatial transcriptomics assays, we were able to test and confirm that specific cell populations can successfully be detected even in small ROIs with the standard manual visualization protocol.

Robustness Analysis of DCE-MRI-Derived Radiomic Features in Breast Masses: Assessing Quantization Levels and Segmentation Agreement

Machine learning models based on radiomic features allow us to obtain biomarkers that are capable of modeling the disease and that are able to support the clinical routine. Recent studies have shown that it is fundamental that the computed features are robust and reproducible. Although several initiatives to standardize the definition and extraction process of biomarkers are ongoing, there is a lack of comprehensive guidelines. Therefore, no standardized procedures are available for ROI selection, feature extraction, and processing, with the risk of undermining the effective use of radiomic models in clinical routine. In this study, we aim to assess the impact that the different segmentation methods and the quantization level (defined by means of the number of bins used in the feature-extraction phase) may have on the robustness of the radiomic features. In particular, the robustness of texture features extracted by PyRadiomics, and belonging to five categories—GLCM, GLRLM, GLSZM, GLDM, and NGTDM—was evaluated using the intra-class correlation coefficient (ICC) and mean differences between segmentation raters. In addition to the robustness of each single feature, an overall index for each feature category was quantified. The analysis showed that the level of quantization (i.e., the ‘bincount’ parameter) plays a key role in defining robust features: in fact, in our study focused on a dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) dataset of 111 breast masses, sets with cardinality varying between 34 and 43 robust features were obtained with ‘binCount’ values equal to 256 and 32, respectively. Moreover, both manual segmentation methods demonstrated good reliability and agreement, while automated segmentation achieved lower ICC values. Considering the dependence on the quantization level, taking into account only the intersection subset among all the values of ‘binCount’ could be the best selection strategy. Among radiomic feature categories, GLCM, GLRLM, and GLDM showed the best overall robustness with varying segmentation methods.

Integrating computational fluid dynamics data into medical image visualization workflows via DICOM.

Communicating complex blood flow patterns generated from computational fluid dynamics (CFD) simulations to clinical audiences for the purposes of risk assessment or treatment planning is an ongoing challenge. While attempts have been made to develop new software tools for such clinical visualization of CFD data, these often overlook established medical imaging/visualization practice and data infrastructures. Here, leveraging the clinical ubiquity of the DICOM file format, we present techniques for the translation of CFD data to DICOM series, facilitating interactive visualization in standard radiological software. Unstructured CFD data (volumetric fields of velocity magnitude, Q-criterion, and pathlines) are resampled to structured grids. Novel raster-based techniques that simulate experimental optical blurring are presented for bringing simulated pathlines into structured image volumes. DICOM series are created by strategically encoding these data into the file's PixelArray tag. Lumen surface information is also strategically encoded into a different range of pixel intensities, allowing hemodynamics and morphology to be co-visualized in a single volume using opacity-based rendering transfer functions. We show that 3D temporal CFD data represented as structured DICOM series can be rendered interactively in Horos, a widely-used medical imaging/radiology software. Our transfer function-based approach allows for representations of scalar isosurfaces, volumetric rendering, and tubular pathlines to be modified in real-time, resembling conventional unstructured visualizations. Careful selection of voxelization ROIs helps to ensure that data are kept lightweight for real-time rendering and minimal storage. While our approach inherently sacrifices some of the advanced visualization capabilities of specialized software tools, we believe our closer consideration of standardization can help to facilitate meaningful clinical interaction. This work opens up possibilities for the complete integration of measured and simulated data in established radiological software environments and workflows from PACS storage to 3D/4D visualization.

Assessment of Boundary Discrimination Performance in a Printed Phantom.

Printed phantoms hold great potential as a tool for examining task-based image quality of x-ray imaging systems. Their ability to produce complex shapes rendered in materials with adjustable attenuation coefficients allows a new level of flexibility in the design of tasks for the evaluation of physical imaging systems. We investigate performance in a fine "boundary discrimination" task in which fine features at the margin of a clearly visible "lesion" are used to classify the lesion as malignant or benign. These tasks are appealing because of their relevance to clinical tasks, and because they typically emphasize higher spatial frequencies relative to more common lesion detection tasks. A 3D printed phantom containing cylindrical shells of varying thickness was used to generate lesions profiles that differed in their edge profiles. This was intended to approximate lesions with indistinct margins that are clinically associated with malignancy. Wall thickness in the phantom ranged from 0.4mm to 0.8mm, which allows for task difficulty to be varied by choosing different thicknesses to represent malignant and benign lesions. The phantom was immersed in a tub filled with water and potassium phosphate to approximate the attenuating background, and imaged repeatedly on a benchtop cone-beam CT scanner. After preparing the image data (reconstruction, ROI Selection, sub-pixel registration), we find that the mean frequency of the lesion profile is 0.11 cyc/mm. The mean frequency of the lesion-difference profile, representative of the discrimination task, is approximately 6 times larger. Model observers show appropriate dose performance in these tasks as well.

Improving the security of medical image through neuro-fuzzy based ROI selection for reliable transmission

Improving the security of medical image through neuro-fuzzy based ROI selection for reliable transmission

Periapical Radiograph Texture Features for Osteoporosis Detection using Deep Convolutional Neural Network

Currently, research for osteoporosis examination using dental radiographic images is increasing rapidly. Many researchers have used various methods from subject data. It indicates that osteoporosis has become a widespread disease that should be studied more deeply. This study proposes a deep Convolutional Neural Network architecture as a texture feature of dental periapical radiograph for osteoporosis detection. The subject of this study is postmenopausal Javanese women aged over 40 and data measurement result of Bone Mineral Density. The proposed model is divided into stages: 1) stage image acquisition and RoI selection, 2) stage feature extraction and classification. Various experiments with the number of convolution layers (3 layers to 6 layers) and various input block sizes and other hyper parameters were used to get the best model. The best model is obtained when the input image size is greater than 100 and less than 150 and a five of convolution layer, as well as other hyper parameters, including epochs=100, dropout=0.5, learning rate=0.0001, batch size= 16 and loss function using Adam's optimization. Validation and testing accuracy achieved by the best model is 98.10%, and 92.50. The research shows that the bigger images provide additional information about trabecular patterns in normal, osteopenia and osteoporosis classes, so that the proposed method using deep convolutional neural network as textural feature of the periapical radiograph achieves a good performance for detection osteoporosis.