Automating Hair Loss Labels for Universally Scoring Alopecia From Images: Rethinking Alopecia Scores.

Importance: Clinical estimation of hair density has an important role in assessing and tracking the severity and progression of alopecia, yet to the authors' knowledge, no automation currently exists for this process. While some algorithms have been developed to assess alopecia presence on a binary level, their scope has been limited by focusing on a re-creation of the Severity of Alopecia Tool (SALT) score for alopecia areata (AA). Yet hair density loss is common to all alopecia forms, and an evaluation of that loss is used in established scoring systems for androgenetic alopecia (AGA), central centrifugal cicatricial alopecia (CCCA), and many more. Objective: To develop and validate a new model, HairComb, to automatically compute the percentage hair loss from images regardless of alopecia subtype. Design, Setting, and Participants: In this research study to create a new algorithmic quantification system for all hair loss, computational imaging analysis and algorithm design using retrospective image data collection were performed. This was a multicenter study, where images were collected at the Children's Hospital of Philadelphia, University of Pennsylvania (Penn), and via a Penn Dermatology web interface. Images were collected from 2015 to 2021, and they were analyzed from 2019 to 2021. Main Outcomes and Measures: Scoring systems correlation analysis was measured by linear and logarithmic regressions. Algorithm performance was evaluated using image segmentation accuracy, density probability regression error, and average percentage hair loss error for labeled images, and Pearson correlation for manual scores. Results: There were 404 participants aged 2 years and older that were used for designing and validating HairComb. Scoring systems correlation analysis was performed for 250 participants (70.4% female; mean age, 35.3 years): 75 AGA, 66 AA, 50 CCCA, 27 other alopecia diagnoses (frontal fibrosing alopecia, lichen planopilaris, telogen effluvium, etc), and 32 unaffected scalps without alopecia. Scoring systems showed strong correlations with underlying percentage hair loss, with coefficient of determination R2 values of 0.793 and 0.804 with respect to log of percentage hair loss. Using HairComb, 92% accuracy, 5% regression error, 7% hair loss difference, and predicted scores with errors comparable to annotators were achieved. Conclusions and Relevance: In this research study,it is shown that an algorithm quantitating percentage hair loss may be applied to all forms of alopecia. A generalizable automated assessment of hair loss would provide a way to standardize measurements of hair loss across a range of conditions.

Photographing Alopecia: How Many Pixels Are Needed for Clinical Evaluation?

Determining the minimum image resolution needed for clinical assessment is crucial for computational efficiency, image standardization, and storage needs alleviation. In this paper, we explore the image resolution requirements for the assessment of alopecia by analyzing how clinicians detect the presence of characteristics needed to quantify the disorder in the clinic. By setting the image resolution as a function of width of the patient's head, we mimicked experiments conducted in the computer vision field to understand human perception in the context of scene recognition and object detection and asked 6 clinicians to identify the regions of interest on a set of retrospectively collected de-identified images at different resolutions. The experts were able to detect the presence of alopecia at very low resolutions, while significantly higher resolution was required to identify the presence of vellus-like hair. Furthermore, the accuracy with which alopecia was detected as a function of resolution followed the same trend as the one obtained when we classified normal versus abnormal hair density using a standard neural network architecture, hinting that the resolution needed by an expert human observer may also provide an upper bound for future image processing algorithms.

Development and Initial Validation of a Multidimensional Acne Global Grading System Integrating Primary Lesions and Secondary Changes.

Importance: The qualitative grading of acne is important for routine clinical care and clinical trials, and although many useful systems exist, no single acne global grading system has had universal acceptance. In addition, many current instruments focus primarily on evaluating primary lesions (eg, comedones, papules, and nodules) or exclusively on signs of secondary change (eg, postinflammatory hyperpigmentation, scarring). Objectives: To develop and validate an acne global grading system that provides a comprehensive evaluation of primary lesions and secondary changes due to acne. Design, Setting, and Participants: This diagnostic study created a multidimensional acne severity feature space by analyzing decision patterns of pediatric dermatologists evaluating acne. Modeling acne severity patterns based on visual image features was then performed to reduce dimensionality of the feature space to a novel 2-dimensional grading system, in which severity levels are functions of multidimensional acne cues. The system was validated by 6 clinicians on a new set of images. All images used in this study were taken from a retrospective, longitudinal data set of 150 patients diagnosed with acne, ranging across the entire pediatric population (aged 0-21 years), excluding images with any disagreement on their diagnosis, and selected to adequately span the range of acne types encountered in the clinic. Data were collected from July 1, 2001, through June 30, 2013, and analyzed from March 1, 2015, through December 31, 2016. Main Outcomes and Measures: Prediction performance was evaluated as the mean square error (MSE) with the clinicians' scores. Results: The scale was constructed using acne visual features and treatment decisions of 6 pediatric dermatologists evaluating 145 images of patients with acne ranging in age from 0 to 21 years. Using the proposed scale to predict the severity scores on a new set of 40 images achieved an overall MSE of 0.821, which is smaller than the mean within-clinician differences (MSE of 0.998). Conclusions and Relevance: By integrating primary lesions and secondary changes, this novel acne global grading scale provides a more clinically relevant evaluation of acne that may be used for routine clinical care and clinical trials. Because the severity scores are based on actual clinical practice, this scoring system is also highly correlated with appropriate treatment choices.

Machine Learning Applications in the Evaluation and Management of Psoriasis: A Systematic Review.

BACKGROUND: Machine learning (ML), a subset of artificial intelligence (AI) that aims to teach machines to automatically learn tasks by inferring patterns from data, holds significant promise to aid psoriasis care. Applications include evaluation of skin images for screening and diagnosis as well as clinical management including treatment and complication prediction. OBJECTIVE: To summarize literature on ML applications to psoriasis evaluation and management and to discuss challenges and opportunities for future advances. METHODS: We searched MEDLINE, Google Scholar, ACM Digital Library, and IEEE Xplore for peer-reviewed publications published in English through December 1, 2019. Our search queries identified publications with any of the 10 computing-related keywords and "psoriasis" in the title and/or abstract. RESULTS: Thirty-three studies were identified. Articles were organized by topic and synthesized as evaluation- or management-focused articles covering 5 content categories: (A) Evaluation using skin images: (1) identification and differential diagnosis of psoriasis lesions, (2) lesion segmentation, and (3) lesion severity and area scoring; (B) clinical management: (1) prediction of complications and (2) treatment. CONCLUSION: Machine learning has significant potential to aid psoriasis evaluation and management. Current topics popular in ML research on psoriasis are the evaluation of medical images, prediction of complications, and treatment discovery. For patients to derive the greatest benefit from ML advancements, it is helpful for dermatologists to have an understanding of ML and how it can effectively aid their assessments and decision-making.

Quantifying Alopecia Areata via Texture Analysis to Automate the SALT Score Computation.

Quantifying alopecia areata in real time has been a challenge for clinicians and investigators. Although several scoring systems exist, they can be cumbersome. Because there are more clinical trials in alopecia areata, there is an urgent need for a quantitative system that is reproducible, standardized, and simple. In this article, a computer imaging algorithm to recreate the Severity of Alopecia Tool scoring system in an automated way is presented. A pediatric alopecia areata image set of four view-standardized photographs was created, and texture analysis was used to distinguish between normal hair and bald scalp. By exploiting local image statistics and the similarity of hair appearance variations across the pediatric alopecia examples, we then used a reference set of hair textures, derived from intensity distributions over very small image patches, to provide global context and improve partitioning of each individual image into areas of different hair densities. This algorithm can mimic a Severity of Alopecia Tool (score) and may also provide more information about the continuum of changes in density of hair seen in alopecia areata.

eCurves: A Temporal Shape Encoding.

OBJECTIVE: This paper presents a framework for temporal shape analysis to capture the shape and changes of anatomical structures from three-dimensional+t(ime) medical scans. METHOD: We first encode the shape of a structure at each time point with the spectral signature, i.e., the eigenvalues and eigenfunctions of the Laplace operator. We then expand it to capture morphing shapes by tracking the eigenmodes across time according to the similarity of their eigenfunctions. The similarity metric is motivated by the fact that small-shaped deformations lead to minor changes in the eigenfunctions. Following each eigenmode from the beginning to end results in a set of eigenmode curves representing the shape and its changes over time. RESULTS: We apply our encoding to a cardiac dataset consisting of series of segmentations outlining the right and left ventricles over time. We measure the accuracy of our encoding by training classifiers on discriminating healthy adults from patients that received reconstructive surgery for Tetralogy of Fallot (TOF). The classifiers based on our encoding significantly surpass deformation-based encodings of the right ventricle, the structure most impacted by TOF. CONCLUSION: The strength of our framework lies in its simplicity: It only assumes pose invariance within a time series but does not assume point-to-point correspondence across time series or a (statistical or physical) model. In addition, it is easy to implement and only depends on a single parameter, i.e., the number of curves.

Pediatric severity of alopecia tool.

The Severity of Alopecia Tool serves as a tool for alopecia research and a clinical guideline for following progression of disease. The original Severity of Alopecia Tool score does not take into account pediatric age groups. As new clinical trials for alopecia areata include more children, a more accurate tool should be available for this population. By collecting images from patients 2-21 years of age and aligning the hair-bearing regions of the scalp, we created an adaptation of the Severity of Alopecia Tool for scoring hair loss percentage of the top, parietal, and occipital scalp in individuals 2-5, 6-11, and 12-21 years of age.

The patient-centered acne severity scale study.

Acne is one of the most common skin conditions seen by dermatologists. As with many other cutaneous diseases, due to its visibility, acne often produces a large psychosocial impact on patients who suffer from the disease. Such psychosocial burdens are exacerbated by the variation in acne presentation that can lead to the usage of multiple different treatments before visible improvements are appreciated. Although many scales have been established to determine severity from the clinician standpoint, patient-oriented scales are lacking. Clinicians use these severity tools to guide management and judge patient improvement from visit to visit. Creation of such a severity scale from a patient's perspective would allow patients to not only assess their perception of their acne independent of a physician but could also be used to determine patient satisfaction with treatment that would then help to more effectively guide management. Therefore the goal of this study is to create and validate a patient-centered acne severity scale using a visual analogue scale format.

Stretch-activated ion channel Piezo1 directs lineage choice in human neural stem cells.

Neural stem cells are multipotent cells with the ability to differentiate into neurons, astrocytes, and oligodendrocytes. Lineage specification is strongly sensitive to the mechanical properties of the cellular environment. However, molecular pathways transducing matrix mechanical cues to intracellular signaling pathways linked to lineage specification remain unclear. We found that the mechanically gated ion channel Piezo1 is expressed by brain-derived human neural stem/progenitor cells and is responsible for a mechanically induced ionic current. Piezo1 activity triggered by traction forces elicited influx of Ca(2+), a known modulator of differentiation, in a substrate-stiffness-dependent manner. Inhibition of channel activity by the pharmacological inhibitor GsMTx-4 or by siRNA-mediated Piezo1 knockdown suppressed neurogenesis and enhanced astrogenesis. Piezo1 knockdown also reduced the nuclear localization of the mechanoreactive transcriptional coactivator Yes-associated protein. We propose that the mechanically gated ion channel Piezo1 is an important determinant of mechanosensitive lineage choice in neural stem cells and may play similar roles in other multipotent stem cells.

Extracting evolving pathologies via spectral clustering.

A bottleneck in the analysis of longitudinal MR scans with white matter brain lesions is the temporally consistent segmentation of the pathology. We identify pathologies in 3D+t(ime) within a spectral graph clustering framework. Our clustering approach simultaneously segments and tracks the evolving lesions by identifying characteristic image patterns at each time-point and voxel correspondences across time-points. For each 3D image, our method constructs a graph where weights between nodes capture the likeliness of two voxels belonging to the same region. Based on these weights, we then establish rough correspondences between graph nodes at different time-points along estimated pathology evolution directions. We combine the graphs by aligning the weights to a reference time-point, thus integrating temporal information across the 3D images, and formulate the 3D+t segmentation problem as a binary partitioning of this graph. The resulting segmentation is very robust to local intensity fluctuations and yields better results than segmentations generated for each time-point.

Temporal shape analysis via the spectral signature.

In this paper, we adapt spectral signatures for capturing morphological changes over time. Advanced techniques for capturing temporal shape changes frequently rely on first registering the sequence of shapes and then analyzing the corresponding set of high dimensional deformation maps. Instead, we propose a simple encoding motivated by the observation that small shape deformations lead to minor refinements in the spectral signature composed of the eigenvalues of the Laplace operator. The proposed encoding does not require registration, since spectral signatures are invariant to pose changes. We apply our representation to the shapes of the ventricles extracted from 22 cine MR scans of healthy controls and Tetralogy of Fallot patients. We then measure the accuracy score of our encoding by training a linear classifier, which outperforms the same classifier based on volumetric measurements.

Structural-flow trajectories for unravelling 3D tubular bundles.

We cast segmentation of 3D tubular structures in a bundle as partitioning of structural-flow trajectories. Traditional 3D segmentation algorithms aggregate local pixel correlations incrementally along a 3D stack. In contrast, structural-flow trajectories establish long range pixel correspondences and their affinities propagate grouping cues across the entire volume simultaneously, from informative to non-informative places. Segmentation by trajectory clustring recovers from persistent ambiguities caused by faint boundaries or low contrast, common in medical images. Trajectories are computed by linking successive registration fields, each one registering pairs of consecutive slices of the 3D stack. We show our method effectively unravels densely packed tubular structures, without any supervision or 3D shape priors, outperforming previous 2D and 3D segmentation algorithms.

Pop out many small structures from a very large microscopic image.

In medical research, many applications require counting and measuring small regions in a large image. Extracting these regions poses a dilemma in terms of segmentation granularity due to fine structures and segmentation complexity due to large image sizes. We propose a constrained spectral graph partitioning framework to address the former while also reducing the segmentation complexity associated with the latter. The final segmentation is obtained from a set of patch segmentations, each independently derived subject to stitching constraints between neighboring patches. Individual segmentation is based on local pairwise cues designed to pop out all cells simultaneously from their common background, while the constraints are derived from mutual agreement analysis on patch segmentations from a previous round of segmentation. Our results demonstrate that the constrained segmentation not only stitches solutions seamlessly along overlapping patch borders but also refines the segmentation in the patch interiors.

Segmentation subject to stitching constraints: finding many small structures in a large image.

Extracting numerous cells in a large microscopic image is often required in medical research. The challenge is to reduce the segmentation complexity on a large image without losing the fine segmentation granularity of small structures. We propose a constrained spectral graph partitioning approach where the segmentation of the entire image is obtained from a set of patch segmentations, independently derived but subject to stitching constraints between neighboring patches. The constraints come from mutual agreement analysis on patch segmentations from a previous round. Our experimental results demonstrate that the constrained segmentation not only stitches solutions seamlessly along overlapping patch borders but also refines the segmentation in the patch interiors.