The relevance and accuracy of an AI algorithm-based descriptor on 23 facial attributes in a diverse female US population.

BACKGROUND: The response of AI in situations that mimic real life scenarios is poorly explored in populations of high diversity. OBJECTIVE: To assess the accuracy and validate the relevance of an automated, algorithm-based analysis geared toward facial attributes devoted to the adornment routines of women. METHODS: In a cross-sectional study, two diversified groups presenting similar distributions such as age, ancestry, skin phototype, and geographical location was created from the selfie images of 1041 female in a US population. 521 images were analyzed as part of a new training dataset aimed to improve the original algorithm and 520 were aimed to validate the performance of the AI. From a total 23 facial attributes (16 continuous and 7 categorical), all images were analyzed by 24 make-up experts and by the automated descriptor tool. RESULTS: For all facial attributes, the new and the original automated tool both surpassed the grading of the experts on a diverse population of women. For the 16 continuous attributes, the gradings obtained by the new system strongly correlated with the assessment made by make-up experts (r ≥ 0.80; p < 0.0001) and supported by a low error rate. For the seven categorical attributes, the overall accuracy of the AI-facial descriptor was improved via enrichment of the training dataset. However, some weaker performance in spotting specific facial attributes were noted. CONCLUSION: In conclusion, the AI-automatic facial descriptor tool was deemed accurate for analysis of facial attributes for diverse women although some skin complexion, eye color, and hair features required some further finetuning.

Accuracy and clinical relevance of an automated, algorithm-based analysis of facial signs from selfie images of women in the United States of various ages, ancestries and phototypes: A cross-sectional observational study.

BACKGROUND: Real-life validation is necessary to ensure our artificial intelligence (AI) skin diagnostic tool is inclusive across a diverse and representative US population of various ages, ancestries and skin phototypes. OBJECTIVES: To explore the relevance and accuracy of an automated, algorithm-based analysis of facial signs in representative women of different ancestries, ages and phototypes, living in the same country. METHODS: In a cross-sectional study of selfie images of 1041 US women, algorithm-based analyses of seven facial signs were automatically graded by an AI-based algorithm and by 50 US dermatologists of various profiles (age, gender, ancestry, geographical location). For automated analysis and dermatologist assessment, the same referential skin atlas was used to standardize the grading scales. The average values and their variability were compared with respect to age, ancestry and phototype. RESULTS: For five signs, the grading obtained by the automated system were strongly correlated with dermatologists' assessments (r ≥ 0.75); cheek skin pores were moderately correlated (r = 0.63) and pigmentation signs, especially for the darkest skin tones, were weakly correlated (r = 0.40) to the dermatologist assessments. Age and ancestry had no effect on the correlations. In many cases, the automated system performed better than the dermatologist-assessed clinical grading due to 0.3-0.5 grading unit differences among the dermatologist panel that were not related to any individual characteristic (e.g. gender, age, ancestry, location). The use of phototypes, as discontinuous categorical variables, is likely a limiting factor in the assessments of grading, whether obtained by automated analysis or clinical assessment of the images. CONCLUSIONS: The AI-based automatic procedure is accurate and clinically relevant for analysing facial signs in a diverse and inclusive population of US women, as confirmed by a diverse panel of dermatologists, although skin tone requires further improvement.

[A brief history of artificial intelligence]. / Une brève introduction à l'intelligence artificielle.

For more than a decade, we have witnessed an acceleration in the development and the adoption of artificial intelligence (AI) technologies. In medicine, it impacts clinical and fundamental research, hospital practices, medical examinations, hospital care or logistics. These in turn contribute to improvements in diagnostics and prognostics, and to improvements in personalised and targeted medicine, advanced observation and analysis technologies, or surgery and other assistance robots. Many challenges in AI and medicine, such as data digitalisation, medical data privacy, algorithm explicability, inclusive AI system development or their reproducibility, have to be tackled in order to build the confidence of medical practitioners in these technologies. This will be possible by mastering the key concepts via a brief history of artificial intelligence.

TITLE: Une brève introduction à l'intelligence artificielle. ABSTRACT: Depuis plus d'une décennie, l'intelligence artificielle (IA) vit une accélération dans son développement et son adoption. En médecine, elle intervient dans la recherche fondamentale et clinique, la pratique hospitalière, les examens médicaux, les soins ou encore la logistique. Ce qui contribue à l'affinement des diagnostics et des pronostics, à une médecine encore plus personnalisée et ciblée, à des avancées dans les technologies d'observations et d'analyses ou encore dans les outils d'interventions chirurgicales et autres robots d'assistance. De nombreux enjeux propres à l'IA et à la médecine, tels que la dématérialisation des données, le respect de la vie privée, l'explicabilité1 des algorithmes, la conception de systèmes d'IA inclusifs ou leur reproductibilité, sont à surmonter pour construire une confiance du corps hospitalier dans ces outils. Cela passe par une maîtrise des concepts fondamentaux que nous présentons ici.

An animal-to-human scaling law for blast-induced traumatic brain injury risk assessment.

Despite recent efforts to understand blast effects on the human brain, there are still no widely accepted injury criteria for humans. Recent animal studies have resulted in important advances in the understanding of brain injury due to intense dynamic loads. However, the applicability of animal brain injury results to humans remains uncertain. Here, we use advanced computational models to derive a scaling law relating blast wave intensity to the mechanical response of brain tissue across species. Detailed simulations of blast effects on the brain are conducted for different mammals using image-based biofidelic models. The intensity of the stress waves computed for different external blast conditions is compared across species. It is found that mass scaling, which successfully estimates blast tolerance of the thorax, fails to capture the brain mechanical response to blast across mammals. Instead, we show that an appropriate scaling variable must account for the mass of protective tissues relative to the brain, as well as their acoustic impedance. Peak stresses transmitted to the brain tissue by the blast are then shown to be a power function of the scaling parameter for a range of blast conditions relevant to TBI. In particular, it is found that human brain vulnerability to blast is higher than for any other mammalian species, which is in distinct contrast to previously proposed scaling laws based on body or brain mass. An application of the scaling law to recent experiments on rabbits furnishes the first physics-based injury estimate for blast-induced TBI in humans.

Laser microfabricated poly(glycerol sebacate) scaffolds for heart valve tissue engineering.

Microfabricated poly(glycerol sebacate) (PGS) scaffolds may be applicable to tissue engineering heart valve leaflets by virtue of their controllable microstructure, stiffness, and elasticity. In this study, PGS scaffolds were computationally designed and microfabricated by laser ablation to match the anisotropy and peak tangent moduli of native bovine aortic heart valve leaflets. Finite element simulations predicted PGS curing conditions, scaffold pore shape, and strut width capable of matching the scaffold effective stiffnesses to the leaflet peak tangent moduli. On the basis of simulation predicted effective stiffnesses of 1.041 and 0.208 MPa for the scaffold preferred (PD) and orthogonal, cross-preferred (XD) material directions, scaffolds with diamond-shaped pores were microfabricated by laser ablation of PGS cured 12 h at 160°C. Effective stiffnesses measured for the scaffold PD (0.83 ± 0.13 MPa) and XD (0.21 ± 0.03 MPa) were similar to both predicted values and peak tangent moduli measured for bovine aortic valve leaflets in the circumferential (1.00 ± 0.16 MPa) and radial (0.26 ± 0.03 MPa) directions. Scaffolds cultivated with fibroblasts for 3 weeks accumulated collagen (736 ± 193 µg/g wet weight) and DNA (17 ± 4 µg/g wet weight). This study provides a basis for the computational design of biomimetic microfabricated PGS scaffolds for tissue-engineered heart valves.

Three-dimensional elastomeric scaffolds designed with cardiac-mimetic structural and mechanical features.

Tissue-engineered constructs, at the interface of material science, biology, engineering, and medicine, have the capacity to improve outcomes for cardiac patients by providing living cells and degradable biomaterials that can regenerate the native myocardium. With an ultimate goal of both delivering cells and providing mechanical support to the healing heart, we designed three-dimensional (3D) elastomeric scaffolds with (1) stiffnesses and anisotropy mimicking explanted myocardial specimens as predicted by finite-element (FE) modeling, (2) systematically varied combinations of rectangular pore pattern, pore aspect ratio, and strut width, and (3) structural features approaching tissue scale. Based on predicted mechanical properties, three scaffold designs were selected from eight candidates for fabrication from poly(glycerol sebacate) by micromolding from silicon wafers. Large 20×20 mm scaffolds with high aspect ratio features (5:1 strut height:strut width) were reproducibly cast, cured, and demolded at a relatively high throughput. Empirically measured mechanical properties demonstrated that scaffolds were cardiac mimetic and validated FE model predictions. Two-layered scaffolds providing fully interconnected pore networks were fabricated by layer-by-layer assembly. C2C12 myoblasts cultured on one-layered scaffolds exhibited specific patterns of cell elongation and interconnectivity that appeared to be guided by the scaffold pore pattern. Neonatal rat heart cells cultured on two-layered scaffolds for 1 week were contractile, both spontaneously and in response to electrical stimulation, and expressed sarcomeric α-actinin, a cardiac biomarker. This work not only demonstrated several scaffold designs that promoted functional assembly of rat heart cells, but also provided the foundation for further computational and empirical investigations of 3D elastomeric scaffolds for cardiac tissue engineering.

Anisotropic collagen fibrillogenesis within microfabricated scaffolds: implications for biomimetic tissue engineering.

Anisotropic collagen fibrillogenesis is demonstrated within the pores of an accordion-like honeycomb poly(glycerol sebacate) tissue engineering scaffold. Confocal reflectance microscopy and image analysis demonstrate increased fibril distribution order, fibril density, and alignment in accordion-like honeycomb pores compared with collagen gelled unconstrained. Finite element modeling predicts how collagen gel and scaffold mechanics couple in matching native heart muscle stiffness and anisotropy.

Finite element analysis of an accordion-like honeycomb scaffold for cardiac tissue engineering.

Optimizing the function of tissue engineered cardiac muscle is becoming more feasible with the development of microfabricated scaffolds amenable to mathematical modeling. In the current study, the elastic behavior of a recently developed poly(glycerol sebacate) (PGS) accordion-like honeycomb (ALH) scaffold [Engelmayr et al., 2008. Nature Materials 7 (12), 1003-1010] was analyzed. Specifically, 2D finite element (FE) models of the ALH unit cell (periodic boundary conditions) and tessellations (kinematic uniform boundary conditions) were utilized to determine a representative volume element (RVE) and to retrospectively predict the elastic effective stiffnesses. An RVE of 90 ALH unit cells (≃3.18×4.03mm) was found, indicating that previous experimental uni-axial test samples were mechanically representative. For ALH scaffolds microfabricated from PGS cured 7.5h at 160°C, FE predicted effective stiffnesses in the two orthogonal material directions (0.081±0.012 and 0.033±0.005MPa) matched published experimental data (0.083±0.004 and 0.031±0.002MPa) within 2.4% and 6.4%. Of potential use as a design criterion, model predicted global strain amplifications were lower in ALH (0.54 and 0.34) versus rectangular honeycomb (1.19 and 0.74) scaffolds, appearing to be inversely correlated with previously measured strains-to-failure. Important in matching the anisotropic mechanical properties of native cardiac muscle, FE predicted ALH scaffolds with 50µm wide PGS struts to be maximally anisotropic. The FE model will thus be useful in designing future variants of the ALH pore geometry that simultaneously provide proper cardiac anisotropy and reduced stiffness to enhance heart cell-mediated contractility.