Robust Image Analysis Method Research Articles

Pan-tumor T-lymphocyte detection using deep neural networks: Recommendations for transfer learning in immunohistochemistry

The success of immuno-oncology treatments promises long-term cancer remission for an increasing number of patients. The response to checkpoint inhibitor drugs has shown a correlation with the presence of immune cells in the tumor and tumor microenvironment. An in-depth understanding of the spatial localization of immune cells is therefore critical for understanding the tumor’s immune landscape and predicting drug response. Computer-aided systems are well suited for efficiently quantifying immune cells in their spatial context. Conventional image analysis approaches are often based on color features and therefore require a high level of manual interaction. More robust image analysis methods based on deep learning are expected to decrease this reliance on human interaction and improve the reproducibility of immune cell scoring. However, these methods require sufficient training data and previous work has reported low robustness of these algorithms when they are tested on out-of-distribution data from different pathology labs or samples from different organs. In this work, we used a new image analysis pipeline to explicitly evaluate the robustness of marker-labeled lymphocyte quantification algorithms depending on the number of training samples before and after being transferred to a new tumor indication. For these experiments, we adapted the RetinaNet architecture for the task of T-lymphocyte detection and employed transfer learning to bridge the domain gap between tumor indications and reduce the annotation costs for unseen domains. On our test set, we achieved human-level performance for almost all tumor indications with an average precision of 0.74 in-domain and 0.72–0.74 cross-domain. From our results, we derive recommendations for model development regarding annotation extent, training sample selection, and label extraction for the development of robust algorithms for immune cell scoring. By extending the task of marker-labeled lymphocyte quantification to a multi-class detection task, the pre-requisite for subsequent analyses, e.g., distinguishing lymphocytes in the tumor stroma from tumor-infiltrating lymphocytes, is met.

Framework for Machine Learning of CT and PET Radiomics to Predict Local Failure after Radiotherapy in Locally Advanced Head and Neck Cancers.

Context:Cancer Radiomics is an emerging field in medical imaging and refers to the process of converting routine radiological images that are typically qualitatively interpreted to quantifiable descriptions of the tumor phenotypes and when combined with statistical analytics can improve the accuracy of clinical outcome prediction models. However, to understand the radiomic features and their correlation to molecular changes in the tumor, first, there is a need for the development of robust image analysis methods, software tools and statistical prediction models which is often limited in low- and middle-income countries (LMIC).Aims:The aim is to build a framework for machine learning of radiomic features of planning computed tomography (CT) and positron emission tomography (PET) using open source radiomics and data analytics platforms to make it widely accessible to clinical groups. The framework is tested in a small cohort to predict local disease failure following radiation treatment for head-and-neck cancer (HNC). The predictors were also compared with the existing Aerts HNC radiomics signature.Settings and Design:Retrospective analysis of patients with locally advanced HNC between 2017 and 2018 and 31 patients with both pre- and post-radiation CT and evaluation PET were selected.Subjects and Methods:Tumor volumes were delineated on baseline PET using the semi-automatic adaptive-threshold algorithm and propagated to CT; PyRadiomics features (total of 110 under shape/intensity/texture classes) were extracted. Two feature-selection methods were tested for model stability. Models were built based on least absolute shrinkage and selection operator-logistic and Ridge regression of the top pretreatment radiomic features and compared to Aerts' HNC-signature. Average model performance across all internal validation test folds was summarized by the area under the receiver operator curve (ROC).Results:Both feature selection methods selected CT features MCC (GLCM), SumEntropy (GLCM) and Sphericity (Shape) that could predict the binary failure status in the cross-validated group and achieved an AUC >0.7. However, models using Aerts' signature features (Energy, Compactness, GLRLM-GrayLevelNonUniformity and GrayLevelNonUniformity-HLH wavelet) could not achieve a clear separation between outcomes (AUC = 0.51–0.54).Conclusions:Radiomics pipeline included open-source workflows which makes it adoptable in LMIC countries. Additional independent validation of data is crucial for the implementation of radiomic models for clinical risk stratification.

A Deep Learning Semantic Segmentation-Based Approach for Field-Level Sorghum Panicle Counting

Small unmanned aerial systems (UAS) have emerged as high-throughput platforms for the collection of high-resolution image data over large crop fields to support precision agriculture and plant breeding research. At the same time, the improved efficiency in image capture is leading to massive datasets, which pose analysis challenges in providing needed phenotypic data. To complement these high-throughput platforms, there is an increasing need in crop improvement to develop robust image analysis methods to analyze large amount of image data. Analysis approaches based on deep learning models are currently the most promising and show unparalleled performance in analyzing large image datasets. This study developed and applied an image analysis approach based on a SegNet deep learning semantic segmentation model to estimate sorghum panicles counts, which are critical phenotypic data in sorghum crop improvement, from UAS images over selected sorghum experimental plots. The SegNet model was trained to semantically segment UAS images into sorghum panicles, foliage and the exposed ground using 462, 250 × 250 labeled images, which was then applied to field orthomosaic to generate a field-level semantic segmentation. Individual panicle locations were obtained after post-processing the segmentation output to remove small objects and split merged panicles. A comparison between model panicle count estimates and manually digitized panicle locations in 60 randomly selected plots showed an overall detection accuracy of 94%. A per-plot panicle count comparison also showed high agreement between estimated and reference panicle counts (Spearman correlation ρ = 0.88, mean bias = 0.65). Misclassifications of panicles during the semantic segmentation step and mosaicking errors in the field orthomosaic contributed mainly to panicle detection errors. Overall, the approach based on deep learning semantic segmentation showed good promise and with a larger labeled dataset and extensive hyper-parameter tuning, should provide even more robust and effective characterization of sorghum panicle counts.

Tailoring the surface characteristics of electrophoretically deposited chitosan-based bioactive glass composite coatings on titanium implants via grit blasting

Chitosan, chitosan/45S5 bioactive glass (mBG), chitosan/nanobioglass (nBG), and chitosan/mBG/nBG coatings were deposited on grit-blasted and mirror-polished Ti6Al4 V alloy substrates by electrophoretic deposition (EPD). This study discusses how the surface characteristics of the substrates influence the 3D surface topography, areal surface roughness, morphology, and wettability of chitosan-based thin composite coatings while focusing on tuning the surface characteristics by grit-blasting. The surfaces of coated and uncoated specimens were characterised by SEM analysis with EDX, image analysis, contact angle measurements, 3D surface topography analysis, FTIR analysis, and in-vitro bioactivity studies. The coatings deposited on grit-blasted substrates presented rougher surface topography, higher areal surface roughness values and wettability compared to the mirror-polished surfaces. A deposition mechanism was suggested to clarify the deposition of chitosan/mBG coatings on grit-blasted substrates. Apatite crystals were formed on chitosan/mBG coatings after immersion in SBF for three days. A robust image analysis method was applied to reveal the distribution and area coverage percent of deposited mBG, which demonstrated the increase of mBG coverage with the increases in voltage. The present study shows that grit blasting could be a favourable surface treatment to tailor the surface characteristics of organic-inorganic composite coatings by influencing their surface topography, roughness, and wettability.

Long-range interactions keep bacterial cells from liquid-solid interfaces: Evidence of a bacteria exclusion zone near Nafion surfaces and possible implications for bacterial attachment

Hydrophilic surfaces of both abiotic and biological origin have been shown to bear particle-exclusion zones as large as hundreds of micrometers at liquid-solid interfaces. Here we present the first systematic investigation and evidence for bacteria-free exclusion zones for several bacterial strains, including pathogens associated with hospital infections and/or foodborne outbreaks: Staphylococcus aureus, Escherichia coli O157:H7, and Listeria monocytogenes. Tests were carried out both in a phosphate buffer, as well as triptic soy broth (TSB) of high ionic strength. Bacterial cell density distribution at the Nafion-liquid interface was visualized using confocal laser scanning microscopy. A robust image analysis method was developed to generate a profile of cell concentration near the interface and quantify EZ size. Results revealed an exclusion zone (EZ) of 40–60μm and a transition zone (TZ) of 40–80μm for bacterial cells suspended in tryptic soy broth. There were no statistical differences in the size of EZ and TZ for the bacterial strains tested with the same substrate, but differences existed for different substrates tested, implying a physicochemical underpinning for EZ. In a test conducted with E. coli, cells progressively penetrated EZ over 2days. Furthermore, EZ-bearing Nafion had 80% less biomass accumulation of E. coli over 2days compared to an EZ-less, hydrophilic, smooth aluminum oxide surface. This suggests that EZ may represent the first line of defense, spatially and temporally, against bacteria approaching certain hydrophilic surfaces. These findings could have important implications in developing biofouling-resistant material surfaces for applications sensitive to bacterial attachment and biofilm formation.

Optimization of Immunostaining for Prospective Image Analysis.

Biomarker discovery is a crucial part of the fast developing field of personalized medicine. Antibody-based techniques including immunostaining of tissue samples are widely used for biomarker evaluation in preclinical and clinical studies. When used in conjunction with robust image analysis methods, it provides a powerful means to assess biomarker modulation, toxicity, and patient response to targeted agents. Here, we describe the optimization of immunofluorescent (IF) staining protocols and a sample IF multiplex protocol suitable for colocalization image analysis.

Spatial reorganization of putaminal dopamine D2-like receptors in cranial and hand dystonia.

The putamen has a somatotopic organization of neurons identified by correspondence of firing rates with selected body part movements, as well as by complex, but organized, differential cortical projections onto putamen. In isolated focal dystonia, whole putaminal binding of dopamine D2-like receptor radioligands is quantitatively decreased, but it has not been known whether selected parts of the putamen are differentially affected depending upon the body part affected by dystonia. The radioligand [18F]spiperone binds predominantly to D2-like receptors in striatum. We hypothesized that the spatial location of [18F]spiperone binding within the putamen would differ in patients with dystonia limited to the hand versus the face, and we tested that hypothesis using positron emission tomography and magnetic resonance imaging. To address statistical and methodological concerns, we chose a straightforward but robust image analysis method. An automated algorithm located the peak location of [18F]spiperone binding within the striatum, relative to a brain atlas, in each of 14 patients with cranial dystonia and 8 patients with hand dystonia. The mean (left and right) |x|, y, and z coordinates of peak striatal binding for each patient were compared between groups by t test. The location of peak [18F]spiperone binding within the putamen differed significantly between groups (cranial dystonia z<hand dystonia z, p = 0.016). We conclude that in isolated focal dystonia, dopamine D2-like receptors are distributed differently in the putamen depending on the body part manifesting dystonia.

A novel 3D3C particle tracking method suitable for microfluidic flow measurements

This article presents a novel method for determining the three-dimensional location of fluorescent particles that is suitable for three-dimensional particle tracking velocimetry measurements in microfluidic flows. This method determines the depth of a particle by inserting a convex lens and axicon into the optical path between a microscope and camera. For particles close to the focal plane, this converts the wavefront from a particle into a Bessel beam, the frequency, and center of which can be directly related to the three-dimensional position of the particle. A robust image analysis method is presented that can determine the properties of the Bessel beam necessary to calculate the particle position. The theory and data analysis method are verified by comparing the calculated position of 1-μm particles to the known position of the particles which scanned through a depth of 100 μm. The average error in the calculated position was 4 μm. Finally, the method is applied to 3D3C particle tracking velocimetry of Poiseuille flow in a 200-μm-deep channel. Uniquely, this method requires no calibration procedure and is insensitive to variations in particle size and brightness.

Improving information quality of MR brain images by fully automatic and robust image analysis methods

Abstract— The medical experts are expected to make diagnosis on ever‐increasing amount of data. Although the intention with the technological advances in imaging devices, such as MRI (magnetic resonance imaging), is to improve the medical practice, the increased amount of data may make the medical expert's task even harder and could arguably be counter‐productive. A more desirable situation for the medical expert is the improvement in the clinical relevance of the image content and a faster, preferably automated, access to the most relevant part of the existing large amount of data. To this effect, the concept of information quality is introduced, and fully automatic, fast, and robust image analysis methods that improve it are proposed. As a guideline for a general framework, these methods are classified by the type of information they can improve: geometric, structural (sub‐organ), and organizational. The algorithmic contributions are in the domain of MR brain image analysis. These methods improve the information quality not only for the visualization and the diagnosis by the expert, but also for the quantification by other automated methods. The robustness of these algorithms has been extensively tested at the Leiden University Medical Center over a large database of 550 patients.

Robust Spatial Autoregressive Modeling for Hardwood Log Inspection

We explore the application of a stochastic texture modeling method toward a machine vision system for log inspection in the forest products industry. This machine vision system uses computerized tomography (CT) imaging to locate and identify internal defects in hardwood logs. The application of CT to such industrial vision problems requires efficient and robust image analysis methods. This paper addresses one particular aspect of the problem of creating such a computer vision system, namely, the use of image texture modeling for wood defect recognition. In particular, we contribute the first application of spatial autoregressive (SAR) modeling to wood-grain texture analysis of CT images of hardwood logs. Thereto a circularly shifted correlation approach is developed to discriminate the circular texture patterns on the cross-sectional CT images of logs. A robust algorithm for parameter estimation is applied to obtain model parameters associated with individual defects occurring inside a log. Based on the estimated model features, a simple minimum distance correlation-classifier is constructed which classifies an unknown defect into one of the prototypical defects. Experimental results from the proposed method, applied to CT images from different red oak wood, are given and show the efficacy of our approach.