Gabriela Oana Cula

Assessing facial wrinkles: automatic detection and quantification

As people mature, their skin gradually presents lines, wrinkles, and folds that become more pronounced with time. Skin wrinkles are perceived as important cues in communicating information about the age of the person. Nowadays, documenting the facial appearance through imaging is prevalent in skin research, therefore detection and quantitative assessment of the degree of facial wrinkling can be a useful tool for establishing an objective baseline and for assessing benefits to facial appearance due to various dermatological treatments. However, few image‐based algorithms for computationally assessing facial wrinkles are present in the literature, and those that exist have limited reliability.

Fluorescence spectroscopy for endogenous porphyrins in human facial skin

The activity of certain bacteria in skin is known to correlate to the presence of porphyrins. In particular the presence of coproporphyrin produced by P.acnes inside plugged pores has been correlated to acne vulgaris. Another porphyrin encountered in skin is protoporphyrin IX, which is produced by the body in the pathway for production of heme. In the present work, a fluorescence spectroscopy system was developed to measure the characteristic spectrum and quantify the two types of porphyrins commonly present in human facial skin. The system is comprised of a Xe lamp both for fluorescence excitation and broadband light source for diffuse reflectance measurements. A computer-controlled filter wheel enables acquisition of sequential spectra, first excited by blue light at 405 nm then followed by the broadband light source, at the same location. The diffuse reflectance spectrum was used to correct the fluorescence spectrum due to the presence of skin chromophores, such as blood and melanin. The resulting fluorescence spectra were employed for the quantification of porphyrin concentration in a population of healthy subjects. The results show great variability on the concentration of these porphyrins and further studies are being conducted to correlate them with skin conditions such as inflammation and acne vulgaris.

Automatic Localization of Skin Layers in Reflectance Confocal Microscopy

Reflectance Confocal Microscopy (RCM) is a noninvasive imaging tool used in clinical dermatology and skin research, allowing real time visualization of skin structural features at different depths at a resolution comparable to that of conventional histology [1]. Currently, RCM is used to generate a rich skin image stack (about 60 to 100 images per scan) which is visually inspected by experts, a process that is tedious, time consuming and exclusively qualitative. Based on the observation that each of the skin images in the stack can be characterized as a texture, we propose a quantitative approach for automatically classifying the images in the RCM stack, as belonging to the different skin layers: stratum corneum, stratum granulosum, stratum spinosum, stratum basale, and the papillary dermis. A reduced set of images in the stack are used to generate a library of representative texture features named textons. This library is employed to characterize all the images in the stack with a corresponding texton histogram. The stack is ultimately separated into 5 different sets of images, each corresponding to different skin layers, exhibiting good correlation with expert grading. The performance of the method is tested against three RCM stacks and we generate promising classification results. The proposed method is especially valuable considering the currently scarce landscape of quantitative solutions for RCM imaging.

Multimodal and time‐lapse skin registration

Computational skin analysis is revolutionizing modern dermatology. Patterns extracted from image sequences enable algorithmic evaluation. Stacking multiple images to analyze pattern variation implicitly assumes that the images are aligned per‐pixel. However, breathing and involuntary motion of the patient causes significant misalignment. Alignment algorithms designed for multimodal and time‐lapse skin images can solve this problem. Sequences from multi‐modal imaging capture unique appearance features in each modality. Time‐lapse image sequences capture skin appearance change over time.

Imaging Inflammatory Acne: Lesion Detection and Tracking

It is known that effectiveness of acne treatment increases when the lesions are detected earlier, before they could progress into mature wound-like lesions, which lead to scarring and discoloration. However, little is known about the evolution of acne from early signs until after the lesion heals. In this work we computationally characterize the evolution of inflammatory acne lesions, based on analyzing cross-polarized images that document acne-prone facial skin over time. Taking skin images over time, and being able to follow skin features in these images present serious challenges, due to change in the appearance of skin, difficulty in repositioning the subject, involuntary movement such as breathing. A computational technique for automatic detection of lesions by separating the background normal skin from the acne lesions, based on fitting Gaussian distributions to the intensity histograms, is presented. In order to track and quantify the evolution of lesions, in terms of the degree of progress or regress, we designed a study to capture facial skin images from an acne-prone young individual, followed over the course of 3 different time points. Based on the behavior of the lesions between two consecutive time points, the automatically detected lesions are classified in four categories: new lesions, resolved lesions (i.e. lesions that disappear completely), lesions that are progressing, and lesions that are regressing (i.e. lesions in the process of healing). The classification our methods achieve correlates well with visual inspection of a trained human grader.

Assessing Facial Wrinkles: Automatic Detection and Quantification

Nowadays, documenting the face appearance through imaging is prevalent in skin research, therefore detection and quantitative assessment of the degree of facial wrinkling is a useful tool for establishing an objective baseline and for communicating benefits to facial appearance due to cosmetic procedures or product applications. In this work, an algorithm for automatic detection of facial wrinkles is developed, based on estimating the orientation and the frequency of elongated features apparent on faces. By over-filtering the skin texture image with finely tuned oriented Gabor filters, an enhanced skin image is created. The wrinkles are detected by adaptively thresholding the enhanced image, and the degree of wrinkling is estimated based on the magnitude of the filter responses. The algorithm is tested against a clinically scored set of images of periorbital lines of different severity and we find that the proposed computational assessment correlates well with the corresponding clinical scores.

Hybrid deep learning for Reflectance Confocal Microscopy skin images

Reflectance Confocal Microscopy (RCM) is used for evaluation of human skin disorders and the effects of skin treatments by imaging the skin layers at different depths. Traditionally, clinical experts manually categorize the images captured into different skin layers. This time-consuming labeling task impedes the convenient analysis of skin image datasets. In recent automated image recognition tasks, deep learning with convolutional neural nets (CNN) has achieved remarkable results. However in many clinical settings, training data is often limited and insufficient for CNN training. For recognition of RCM skin images, we demonstrate that a CNN trained on a moderate size dataset leads to low accuracy. We introduce a hybrid deep learning approach which uses traditional texton-based feature vectors as input to train a deep neural network. This hybrid method uses fixed filters in the input layer instead of tuned filters, yet superior performance is achieved. Our dataset consists of 1500 images from 15 RCM stacks belonging to six different categories of skin layers. We show that our hybrid deep learning approach performs with a test accuracy of 82% compared with 51% for CNN. We also compare the results with additional proposed methods for RCM image recognition and show improved accuracy.

From photography to microbiology: Eigenbiome models for skin appearance

Skin appearance modeling using high resolution photography has led to advances in recognition, rendering and analysis. Computational appearance provides an exciting new opportunity for integrating macroscopic imaging and microscopic biology. Recent studies indicate that skin appearance is dependent on the unseen distribution of microbes on the skin surface, i.e. the skin microbiome. While modern sequencing methods can be used to identify microbes, these methods are costly and time-consuming. We develop a computational skin texture model to characterize image-based patterns and link them to underlying microbiome clusters. The pattern analysis uses ultraviolet and blue fluorescence multimodal skin photography. The intersection of appearance and microbiome clusters reveals a pattern of microbiome that is predictable with high accuracy based on skin appearance. Furthermore, the use of non-negative matrix factorization allows a representation of the microbiome eigenvector as a physically plausible positive distribution of bacterial components. In this paper, we present the first results in this area of predicting microbiome clusters based on computational skin texture.