Automated detection of ulcers and erosions in capsule endoscopy images using a convolutional neural network.

Capsule endoscopy (CE) is an important tool in the management of patients with known or suspected inflammatory bowel disease. Ulcers and erosions of the enteric mucosa are prevalent findings in these patients. They frequently occur together, and their identification in CE is crucial for an accurate evaluation of disease severity. Nevertheless, reviewing CE images is a time-consuming task, and the risk of overlooking lesions is significant.Over the last decade, artificial intelligence (AI) has emerged as a means for overcoming these pitfalls. Of all AI methods, convolutional neural networks (CNN), due to their complex multilayer architecture present the best results in medical image analysis, particularly capsule endoscopy. Therefore, we aimed to develop a CNN for the automatic identification of ulcers and erosions in the small bowel mucosa. A total of 1483 CE exams (PillCam SB3®) performed at a single center between 2015 and 2020 were analysed. From these exams, a total of 6130 frames of the enteric mucosa were obtained, 4233 containing enteric ulcers and erosions, and the remaining containing normal mucosa or other findings. Ulcers and erosions were stratified according to Saurin's classification for bleeding potential: P1E-erosions with intermediate bleeding risk; P1U-ulcers with intermediate bleeding risk; P2U-ulcers with high bleeding risk. For automatic identification of these lesions, these images were inserted into a CNN model with transfer learning. The pool of images was divided for constitution of training and validation datasets, comprising 80% and 20% of the total number of images, respectively. The output provided by the CNN was compared to the classification provided by a consensus of specialists. After optimizing the neural architecture of the algorithm, our model was able to automatically detect and distinguish ulcers and erosions (any bleeding potential) in the small intestine mucosa with an accuracy of 95.6%, sensitivity of 90.8%, and a specificity of 97.1%. We believe that our study lays the foundation for the development and application of effective AI tools to CE. These techniques should improve diagnostic accuracy and reading efficiency. Schematic representation of the workflow and summary of the results.

Artificial intelligence and colon capsule endoscopy: development of an automated diagnostic system of protruding lesions in colon capsule endoscopy.

BACKGROUND: Colon capsule endoscopy (CCE) is a minimally invasive alternative for patients unwilling to undergo conventional colonoscopy, or for whom the latter exam is contraindicated. This is particularly important in the setting of colorectal cancer screening. Nevertheless, these exams produce large numbers of images, and reading them is a monotonous and time-consuming task, with the risk of overlooking important lesions. The development of automated tools based on artificial intelligence (AI) technology may improve some of the drawbacks of this diagnostic instrument. METHODS: A database of CCE images was used for development of a Convolutional Neural Network (CNN) model. This database included anonymized images of patients with protruding lesions in the colon or patients with normal colonic mucosa or with other pathologic findings. A total of 3,387,259 frames from 24 CCE exams were retrospectively reviewed. For CNN development, 3640 images (860 protruding lesions and 2780 with normal mucosa or other findings) were ultimately extracted. Training and validation datasets were constructed for the development and testing of the CNN. RESULTS: The CNN detected protruding lesions with a sensitivity, specificity, positive and negative predictive values of 90.7, 92.6, 79.2 and 96.9%, respectively. The area under the receiver operating characteristic curve for detection of protruding lesions was 0.97. CONCLUSIONS: The deep learning algorithm we developed is capable of accurately detecting protruding lesions. The application of AI technology to CCE may increase its diagnostic accuracy and acceptance for screening of colorectal neoplasia.

Altered mechanics of vaginal smooth muscle cells due to the lysyl oxidase-like1 knockout.

The remodeling mechanisms that cause connective tissue of the vaginal wall, consisting mostly of smooth muscle, to weaken after vaginal delivery are not fully understood. Abnormal remodeling after delivery can contribute to development of pelvic organ prolapse and other pelvic floor disorders. The present study used vaginal smooth muscle cells (vSMCs) isolated from knockout mice lacking the expression of the lysyl oxidase-like1 (LOXL1) enzyme, a well-characterized animal model for pelvic organ prolapse. We tested if vaginal smooth muscle cells from LOXL1 knockout mice have altered mechanics including stiffness and surface adhesion. Using atomic force microscopy, we performed nanoindentations on both isolated and confluent cells to evaluate the effect of LOXL1 knockout on in vitro cultures of vSMCs cells from nulliparous mice. The results show that LOXL1 knockout vSMCs have increased stiffness in pre-confluent but decreased stiffness in confluent cultures (p* < 0.05) and significant decreased surface adhesion in pre-confluent cultures (p* < 0.05). This study provides evidence that the weakening of vaginal connective tissue in the absense of LOXL1 changes the mechanical properties of the vSMCs. STATEMENT OF SIGNIFICANCE: Pelvic organ prolapse is a common condition affecting millions of women worldwide, which significantly impacts their quality of life. Alterations in vaginal and pelvic floor mechanical properties can change their ability to support the pelvic organs. This study provides evidence of altered stiffness of vaginal smooth muscle cells from mice resembling pelvic organ prolapse. The results from this study set a foundation to develop pathophysiology-driven therapies focused on the interplay between smooth muscle mechanics and extracellular matrix remodeling.

Characterization of hyperelastic and damage behavior of tendons.

In this paper, we characterized the hyperelastic and damage behavior of the Extensor Digitorum Longus (EDL) human tendon under loading conditions. The study was conducted in both categories of models, phenomenological and physically motivated, to allow the prediction and the macroscopic response of the tendon under specific loading conditions, assuming that its response follows a hyperelastic anisotropic model in conjunction with damage law. We benchmarked multiple hyperelastic and damage models to fit the response of the tendons in uniaxial tensile loading conditions, and by employing a genetic algorithm, we obtained the material parameters for both elastic and damage models. The objective of this study was to explore different mathematical models to determine which would be the best option to predict the behavior of tendons and ligaments in complex biological systems using Finite Elements (FE) models. Therefore, we took into account accuracy as well as computational features. We considered the model proposed by Shearer and coupled it with a sigmoid function, which governs the evolution of damage in tendons, as the most appropriate for the fitting of the experimental data. The achieved solution shows to be of high interest attributable to the simplicity of the damage law function and its low computational cost.

Simulation of the uterine contractions and foetus expulsion using a chemo-mechanical constitutive model.

During vaginal delivery women sustain stretching of their pelvic floor, risking tissue injury and adverse outcomes. Since studies in pregnant women are limited with ethical constraints, computational models have become an interesting alternative to elucidate the pregnancy mechanisms. This research investigates the uterine contractions during foetus expulsion without an imposed trajectory. Such physical process is captured by means of a chemo-mechanical constitutive model, where the uterine contractions are triggered by chemical stimuli. The foetus descent, which includes both pushing and resting stages, has a descent rate within the physiological range. Moreover, the behaviour of the foetus and the uterus stretch agree well with clinical data presented in the literature. The follow-up of this study will be to obtain a complete childbirth simulation, considering also the pelvic floor muscles and its supporting structures. The simulation of a realistic rate of descent, including the pushing and resting stages, is of significant importance to study the pelvic floor muscles due to their viscoelastic nature.

On the effect of labour durations using an anisotropic visco-hyperelastic-damage approach to simulate vaginal deliveries.

Injuries sustained by the pelvic floor muscles during childbirth are one of the major risk factors for the development of pelvic floor dysfunctions. The ability to predict the loss of the tissue integrity and the most affected regions prior to the childbirth would represent a compelling difference in choosing the appropriate management of labour. Previous biomechanical studies, using the finite element method, were able to simulate a vaginal delivery and analyse the mechanical effects on the pelvic floor muscles during the passage of the foetus. Complementing these studies, the aim of this work is to improve the characterization of the pelvic floor muscles, by using an anisotropic visco-hyperelastic constitutive model, including a continuum mechanics damage model. Viscoelasticity is a key feature to obtain more realistic results since biological tissues present relaxation effects that allow larger deformations without damage. This work analyses the reaction forces and the loss of tissue integrity sustained by the pelvic floor and evaluates the effects of different durations of labour. A delaying pushing technique of rest and descend is also studied in this work. The results obtained showed that the reaction forces vary with the duration of labour, with higher force levels associated with higher stretch rates. The pubovisceral muscle is the most affected of the levator ani, presenting an affected region of approximately 30%. The relaxation properties of the tissue contribute to diminish the damage levels, supporting the theory of delayed pushing applied in the second stage of labour.

Viscous effects in pelvic floor muscles during childbirth: A numerical study.

During vaginal delivery, women sustain stretching of their pelvic floor, risking tissue injury and adverse outcomes. Realistic numerical simulations of childbirth can help in the understanding of the pelvic floor mechanics and on the prevention of related disorders. In previous studies, biomechanical finite element simulations of a vaginal delivery have been performed disregarding the viscous effects present on all biological soft tissues. The inclusion of the viscoelastic behaviour is fundamental, since it allows to investigate rate-dependent responses. The present work uses a viscohyperelastic constitutive model to evaluate how the childbirth duration affects the efforts sustained by the pelvic floor during delivery. It was concluded that viscoelasticity adds a stiffness component that leads to higher forces comparing with the elastic response. Viscous solutions are rate dependent, and precipitous labours could be associated to higher efforts, while lower reaction forces were denoted for normal and prolonged labours, respectively. The existence of resting stages during labour demonstrated the capability of the tissue to relax and recover some of the initial properties, which helped to lower the forces and stresses involved. The present work represents a step further in achieving a robust non-invasive procedure, allowing to estimate how obstetrical factors influence labour and its outcomes.

A general framework for the numerical implementation of anisotropic hyperelastic material models including non-local damage.

The highly nonlinear mechanical behaviour of soft tissues solicited within the physiological range usually involves degradation of the material properties. Mechanically, having these biostructures undergoing such stretch patterns may bring about pathological conditions related to the steady deterioration of both collagen fibres and material's ground substance. Tissue and subject variability observed in the phenomenological mechanical characterisation of soft tissues often hinder the choice of the computational constitutive model. Therefore, this contribution brings forth a detailed overview of the constitutive implementation in a computational framework of anisotropic hyperelastic materials with damage. Surmounting the challenge posed by the mesh dependency pathology requires the incorporation of an integral-type non-local averaging, which seeks to include the effects of the microstructure in order to limit the localisation phenomena of the damage variables. By adopting this approach, one can make use of multiple developed material models available in the literature, a combination of those, or even propose new models within the same numerical framework. The numerical examples of three-dimensional displacement and force-driven boundary value problems highlight the possibility of using multiple material models within the same numerical framework. Particularities concerning the considered material models and the damage effect implications to represent the Mullins effect, induced anisotropy, hysteresis, and mesh dependency are discussed.