ABSTRACTThe purpose of this study is to evaluate the ability of a machine learning algorithm to classify in vivo magnetic resonance images (MRI) of human articular cartilage for development of osteoarthritis (OA). Sixty‐eight subjects were selected from the osteoarthritis initiative (OAI) control and incidence cohorts. Progression to clinical OA was defined by the development of symptoms as quantified by the Western Ontario and McMaster Universities Arthritis (WOMAC) questionnaire 3 years after baseline evaluation. Multi‐slice T2‐weighted knee images, obtained through the OAI, of these subjects were registered using a nonlinear image registration algorithm. T2 maps of cartilage from the central weight bearing slices of the medial femoral condyle were derived from the registered images using the multiple available echo times and were classified for “progression to symptomatic OA” using the machine learning tool, weighted neighbor distance using compound hierarchy of algorithms representing morphology (WND‐CHRM). WND‐CHRM classified the isolated T2 maps for the progression to symptomatic OA with 75% accuracy. Clinical significance: Machine learning algorithms applied to T2 maps have the potential to provide important prognostic information for the development of OA. © 2017 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 35:2243–2250, 2017.

In this study, we describe a morphological biomarker that detects multiple discrete subpopulations (or "age-states") at several chronological ages in a population of nematodes (Caenorhabditis elegans). We determined the frequencies of three healthy adult states and the timing of the transitions between them across the lifespan. We used short-lived and long-lived strains to confirm the general applicability of the state classifier and to monitor state progression. This exploration revealed healthy and unhealthy states, the former being favored in long-lived strains and the latter showing delayed onset. Short-lived strains rapidly transitioned through the putative healthy state. We previously found that age-matched animals in different age-states have distinct transcriptome profiles. We isolated animals at the beginning and end of each identified state and performed microarray analysis (principal component analysis, relative sample to sample distance measurements, and gene set enrichment analysis). In some comparisons, chronologically identical individuals were farther apart than morphologically identical individuals isolated on different days. The age-state biomarker allowed assessment of aging in a novel manner, complementary to chronological age progression. We found hsp70 and some small heat shock protein genes are expressed later in adulthood, consistent with the proteostasis collapse model.

Bio-image informatics is the systematic application of image analysis algorithms to large image datasets to provide an objective method for accurately and consistently scoring image data. Within this field, pattern recognition (PR) is a form of supervised machine learning where the computer identifies relevant patterns in groups (classes) of images after being trained on examples. Rather than segmentation, image-specific algorithms or adjustable parameter sets, PR relies on extracting a common set of image descriptors (features) from the entire image to determine similarities and differences between image classes.Gross morphology can be the only available description of biological systems prior to their molecular characterization, but these descriptions can be subjective and qualitative. In principle, generalized PR can provide an objective and quantitative characterization of gross morphology, thus providing a means of computationally defining morphological biomarkers. In this study, we investigated the potential of a pattern recognition approach to a problem traditionally addressed using genetic or biochemical biomarkers. Often these molecular biomarkers are unavailable for investigating biological processes that are not well characterized, such as the initial steps of stem cell differentiation.Here we use a general contrast technique combined with generalized PR software to detect subtle differences in cellular morphology present in early differentiation events in murine embryonic stem cells (mESC) induced to differentiate by the overexpression of selected transcription factors. Without the use of reporters, or a priori knowledge of the relevant morphological characteristics, we identified the earliest differentiation event (3 days), reproducibly distinguished eight morphological trajectories, and correlated morphological trajectories of 40 mESC clones with previous micro-array data. Interestingly, the six transcription factors that caused the greatest morphological divergence from an ESC-like state were previously shown by expression profiling to have the greatest influence on the expression of downstream genes.

ObjectiveThe purpose of this study is to evaluate the ability of machine learning to discriminate between magnetic resonance images (MRI) of normal and pathological human articular cartilage obtained under standard clinical conditions.MethodAn approach to MRI classification of cartilage degradation is proposed using pattern recognition and multivariable regression in which image features from MRIs of histologically scored human articular cartilage plugs were computed using weighted neighbor distance using compound hierarchy of algorithms representing morphology (WND-CHRM). The WND-CHRM method was first applied to several clinically available MRI scan types to perform binary classification of normal and osteoarthritic osteochondral plugs based on the Osteoarthritis Research Society International (OARSI) histological system. In addition, the image features computed from WND-CHRM were used to develop a multiple linear least-squares regression model for classification and prediction of an OARSI score for each cartilage plug.ResultsThe binary classification of normal and osteoarthritic plugs yielded results of limited quality with accuracies between 36% and 70%. However, multiple linear least-squares regression successfully predicted OARSI scores and classified plugs with accuracies as high as 86%. The present results improve upon the previously-reported accuracy of classification using average MRI signal intensities and parameter values.ConclusionMRI features detected by WND-CHRM reflect cartilage degradation status as assessed by OARSI histologic grading. WND-CHRM is therefore of potential use in the clinical detection and grading of osteoarthritis.

Normal cognitive aging is associated with deficits in memory processes dependent on the hippocampus, along with large-scale changes in the hippocampal expression of many genes. Histone acetylation can broadly influence gene expression and has been recently linked to learning and memory. We hypothesized that CREB-binding protein (CBP), a key histone acetyltransferase, may contribute to memory decline in normal aging. Here, we quantified CBP protein levels in the hippocampus of young, aged unimpaired, and aged impaired rats, classified on the basis of spatial memory capacity documented in the Morris water maze. First, CBP-immunofluorescence was quantified across the principal cell layers of the hippocampus using both low and high resolution laser scanning imaging approaches. Second, digital images of CBP immunostaining were analyzed by a multipurpose classifier algorithm with validated sensitivity across many types of input materials. Finally, CBP protein levels in the principal subfields of the hippocampus were quantified by quantitative Western blotting. CBP levels were equivalent as a function of age and cognitive status in all analyses. The sensitivity of the techniques used was substantial, sufficient to reveal differences across the principal cell fields of the hippocampus, and to correctly classify images from young and aged animals independent of CBP immunoreactivity. The results are discussed in the context of recent evidence suggesting that CBP decreases may be most relevant in conditions of aging that, unlike normal cognitive aging, involve significant neuron loss.

We present an initial molecular characterization of a morphological transition between two early aging states. In previous work, an age score reflecting physiological age was developed using a machine classifier trained on images of worm populations at fixed chronological ages throughout their lifespan. The distribution of age scores identified three stable post-developmental states and transitions. The first transition occurs at day 5 post-hatching, where a significant percentage of the population exists in both state I and state II. The temperature dependence of the timing of this transition (Q 10 ~ 1.17) is too low to be explained by a stepwise process with an enzymatic or chemical rate-limiting step, potentially implicating a more complex mechanism. Individual animals at day 5 were sorted into state I and state II groups using the machine classifier and analyzed by microarray expression profiling. Despite being isogenic, grown for the same amount of time, and indistinguishable by eye, these two morphological states were confirmed to be molecularly distinct by hierarchical clustering and principal component analysis of the microarray results. These molecular differences suggest that pharynx morphology reflects the aging state of the whole organism. Our expression profiling yielded a gene set that showed significant overlap with those from three previous age-related studies and identified several genes not previously implicated in aging. A highly represented group of genes unique to this study is involved in targeted ubiquitin-mediated proteolysis, including Skp1-related (SKR), F-box-containing, and BTB motif adaptors.

We present results from machine classification of melanoma biopsies sectioned and stained with hematoxylin/eosin (H&E) on tissue microarrays (TMA). The four stages of melanoma progression were represented by seven tissue types, including benign nevus, primary tumors with radial and vertical growth patterns (stage I) and four secondary metastatic tumors: subcutaneous (stage II), lymph node (stage III), gastrointestinal and soft tissue (stage IV). Our experiment setup comprised 14,208 image samples based on 164 TMA cores. In our experiments, we constructed an HE color space by digitally deconvolving the RGB images into separate H (hematoxylin) and E (eosin) channels. We also compared three different classifiers: Weighted Neighbor Distance (WND), Radial Basis Functions (RBF), and k‐Nearest Neighbors (kNN). We found that the HE color space consistently outperformed other color spaces with all three classifiers, while the different classifiers did not have as large of an effect on accuracy. This showed that a more physiologically relevant representation of color can have a larger effect on correct image interpretation than downstream processing steps. We were able to correctly classify individual fields of view with an average of 96% accuracy when randomly splitting the dataset into training and test fields. We also obtained a classification accuracy of 100% when testing entire cores that were not previously used in training (four random trials with one test core for each of 7 classes, 28 tests total). Because each core corresponded to a different patient, this test more closely mimics a clinically relevant setting where new patients are evaluated based on training with previous cases. The analysis method used in this study contains no parameters or adjustments that are specific to melanoma morphology, suggesting it can be used for analyzing other tissues and phenotypes, as well as potentially different image modalities and contrast techniques. Published 2012 Wiley Periodicals, Inc.