-
Learning a Clinically-Relevant Concept Bottleneck for Lesion Detection in Breast Ultrasound
Authors:
Arianna Bunnell,
Yannik Glaser,
Dustin Valdez,
Thomas Wolfgruber,
Aleen Altamirano,
Carol Zamora González,
Brenda Y. Hernandez,
Peter Sadowski,
John A. Shepherd
Abstract:
Detecting and classifying lesions in breast ultrasound images is a promising application of artificial intelligence (AI) for reducing the burden of cancer in regions with limited access to mammography. Such AI systems are more likely to be useful in a clinical setting if their predictions can be explained to a radiologist. This work proposes an explainable AI model that provides interpretable pred…
▽ More
Detecting and classifying lesions in breast ultrasound images is a promising application of artificial intelligence (AI) for reducing the burden of cancer in regions with limited access to mammography. Such AI systems are more likely to be useful in a clinical setting if their predictions can be explained to a radiologist. This work proposes an explainable AI model that provides interpretable predictions using a standard lexicon from the American College of Radiology's Breast Imaging and Reporting Data System (BI-RADS). The model is a deep neural network featuring a concept bottleneck layer in which known BI-RADS features are predicted before making a final cancer classification. This enables radiologists to easily review the predictions of the AI system and potentially fix errors in real time by modifying the concept predictions. In experiments, a model is developed on 8,854 images from 994 women with expert annotations and histological cancer labels. The model outperforms state-of-the-art lesion detection frameworks with 48.9 average precision on the held-out testing set, and for cancer classification, concept intervention is shown to increase performance from 0.876 to 0.885 area under the receiver operating characteristic curve. Training and evaluation code is available at https://github.com/hawaii-ai/bus-cbm.
△ Less
Submitted 28 June, 2024;
originally announced July 2024.
-
Toward Zero-Shot Sim-to-Real Transfer Learning for Pneumatic Soft Robot 3D Proprioceptive Sensing
Authors:
Uksang Yoo,
Hanwen Zhao,
Alvaro Altamirano,
Wenzhen Yuan,
Chen Feng
Abstract:
Pneumatic soft robots present many advantages in manipulation tasks. Notably, their inherent compliance makes them safe and reliable in unstructured and fragile environments. However, full-body shape sensing for pneumatic soft robots is challenging because of their high degrees of freedom and complex deformation behaviors. Vision-based proprioception sensing methods relying on embedded cameras and…
▽ More
Pneumatic soft robots present many advantages in manipulation tasks. Notably, their inherent compliance makes them safe and reliable in unstructured and fragile environments. However, full-body shape sensing for pneumatic soft robots is challenging because of their high degrees of freedom and complex deformation behaviors. Vision-based proprioception sensing methods relying on embedded cameras and deep learning provide a good solution to proprioception sensing by extracting the full-body shape information from the high-dimensional sensing data. But the current training data collection process makes it difficult for many applications. To address this challenge, we propose and demonstrate a robust sim-to-real pipeline that allows the collection of the soft robot's shape information in high-fidelity point cloud representation. The model trained on simulated data was evaluated with real internal camera images. The results show that the model performed with averaged Chamfer distance of 8.85 mm and tip position error of 10.12 mm even with external perturbation for a pneumatic soft robot with a length of 100.0 mm. We also demonstrated the sim-to-real pipeline's potential for exploring different configurations of visual patterns to improve vision-based reconstruction results. The code and dataset are available at https://github.com/DeepSoRo/DeepSoRoSim2Real.
△ Less
Submitted 7 March, 2023;
originally announced March 2023.
-
On finite size effects, ensemble choice and force influence in dissipative particle dynamics simulations
Authors:
Miguel Ángel Balderas Altamirano,
Elías Pérez,
Armando Gama Goicochea
Abstract:
The influence of finite size effects, choice of statistical ensemble and contribution of the forces in numerical simulations using the dissipative particle dynamics (DPD) model are revisited here. Finite size effects in stress anisotropy, interfacial tension and dynamic viscosity are computed and found to be minimal with respect to other models. Additionally, the choice of ensemble is found to be…
▽ More
The influence of finite size effects, choice of statistical ensemble and contribution of the forces in numerical simulations using the dissipative particle dynamics (DPD) model are revisited here. Finite size effects in stress anisotropy, interfacial tension and dynamic viscosity are computed and found to be minimal with respect to other models. Additionally, the choice of ensemble is found to be of fundamental importance for the accurate calculation of properties such as the solvation pressure, especially for relatively small systems. Lastly, the contribution of the random, dissipative and conservative forces that make up the DPD model in the prediction of properties of simple liquids such as the pressure is studied as well. Some tricks of the trade are provided, which may be useful for those carrying out high-performance numerical simulations using the DPD model.
△ Less
Submitted 6 December, 2016;
originally announced December 2016.
-
Adsorption of polyelectrolytes on silica and gold surfaces
Authors:
M. A. Balderas Altamirano,
R. Camacho,
E. Pérez,
A. Gama Goicochea
Abstract:
The results of a study that helps understand the mechanisms of adsorption of polyelectrolytes on particles, using numerical simulation methods, specifically the one known as dissipative particle dynamics are reported here. The adsorption of cationic polyelectrolytes of two different polymerization degrees interacting with two types of surfaces, one made of gold and the other of silica is predicted…
▽ More
The results of a study that helps understand the mechanisms of adsorption of polyelectrolytes on particles, using numerical simulation methods, specifically the one known as dissipative particle dynamics are reported here. The adsorption of cationic polyelectrolytes of two different polymerization degrees interacting with two types of surfaces, one made of gold and the other of silica is predicted and compared. We find that a more negatively charged wall does not necessarily adsorb more cationic polyelectrolytes because the electrostatic repulsion between the wall and the polyelectrolytes is stronger. Additionally, intra chain repulsion plays an important role, because the largest polyelectrolyte chains have larger excluded volume than the shorter ones. In regard to the adsorption dependence on the polyelectrolyte polymerization degree we find that the excluded volume drives the adsorption throughout the intrachain electrostatic repulsion, because the SiO2 surface is strongly negative. These results are expected to be useful for several nanotechnological applications of current interest, such as in gene therapy and in the improvement of drug delivering mechanisms.
△ Less
Submitted 18 May, 2016;
originally announced May 2016.
-
Friction coefficient and viscosity of polymer brushes with and without free polymers as slip agents
Authors:
A. Gama Goicochea,
R. López-Esparza,
M. A. Balderas Altamirano,
E. Rivera-Paz,
M. A. Waldo-Mendoza,
E. Pérez
Abstract:
There is ample evidence that polymer brushes reduce friction between surfaces. Several industrial applications take advantage of this fact, such as those in plastic bag production, where the brushes act as slip agents; however, the complex mechanisms that give rise to such reduction of friction are not yet fully understood. In this work we report coarse grained, dissipative particle dynamics simul…
▽ More
There is ample evidence that polymer brushes reduce friction between surfaces. Several industrial applications take advantage of this fact, such as those in plastic bag production, where the brushes act as slip agents; however, the complex mechanisms that give rise to such reduction of friction are not yet fully understood. In this work we report coarse grained, dissipative particle dynamics simulations carried out for surfaces functionalized with erukamide brushes, a polymer commonly used in the plastics industry as a slip agent between surfaces. We calculate their rheological properties, such as the coefficient of friction (COF) and the viscosity, η, as functions of the number of chains grafted on the surfaces under the influence of stationary, Couette flow. Moreover, we consider also the case when a fraction of the erukamide chains is not adsorbed and moves freely between the surfaces. We show that the COF reaches an equilibrium value of about 0.29 in these two cases, in agreement with experimental results. On other hand, the viscosity grows monotonically, as a result of the increasing collisions when the erukamide content is increased. The force between brushes is found to be in agreement with predictions from scaling theories. We find that the addition to free chains helps stabilize the film formed by the brushes and the solvent, as others have found experimentally. The mechanisms that give rise to these phenomena are studied in detail.
△ Less
Submitted 14 March, 2016;
originally announced March 2016.
-
Importance of molecular interactions in colloidal dispersions
Authors:
R. Lopez-Esparza,
M. A. Balderas Altamirano,
E. Pérez,
A. Gama Goicochea
Abstract:
We review briefly the concept of colloidal dispersions, their general properties and some of their most important applications, as well as the basic molecular interactions that give rise to their properties in equilibrium. Similarly, we revisit Brownian motion and hydrodynamic interactions associated with the concept of viscosity of colloidal dispersion. It is argued that the use of modern researc…
▽ More
We review briefly the concept of colloidal dispersions, their general properties and some of their most important applications, as well as the basic molecular interactions that give rise to their properties in equilibrium. Similarly, we revisit Brownian motion and hydrodynamic interactions associated with the concept of viscosity of colloidal dispersion. It is argued that the use of modern research tools, such as computer simulations, allows one to predict accurately some macroscopically measurable properties by solving relatively simple models of molecular interactions for a large number of particles. Lastly, as a case study, we report the prediction of rheological properties of polymer brushes using state of the art, coarse grained computer simulations, which are in excellent agreement with experiments.
△ Less
Submitted 24 September, 2015; v1 submitted 6 August, 2015;
originally announced August 2015.
-
Importance of pore length and geometry in the adsorption/desorption process: a molecular simulation study
Authors:
M. A. Balderas Altamirano,
S. Cordero,
R. López-Esparza,
E. Pérez,
A. Gama Goicochea
Abstract:
Discrete potentials can describe properly the liquid vapor boundary that is necessary to model the adsorption of gas molecules in mesoporous systems with computer simulations. Although there are some works in this subject, the simulations are still highly time - consuming. Here we show that an efficient alternative is to use the three - dimensional Ising model, which allows one to model large syst…
▽ More
Discrete potentials can describe properly the liquid vapor boundary that is necessary to model the adsorption of gas molecules in mesoporous systems with computer simulations. Although there are some works in this subject, the simulations are still highly time - consuming. Here we show that an efficient alternative is to use the three - dimensional Ising model, which allows one to model large systems, with geometries as complex as required that accurately represent the liquid vapor boundary. In particular, we report molecular simulations of cylindrical pores of two different geometry, using a discrete potential. The effect of the length of the pore in the hysteresis loop for a finite pore and infinite one is studied in detail. Lastly, we compare our predictions with experimental results and find excellent agreement between the area of the hysteresis loop predicted for the finite pore and that found in adsorption/desorption experiments.
△ Less
Submitted 5 July, 2015;
originally announced July 2015.
-
On the computational modeling of the viscosity of colloidal dispersions and its relation with basic molecular interactions
Authors:
A. Gama Goicochea,
M. A. Balderas Altamirano,
R. Lopez-Esparza,
M. A. Waldo,
E. Perez
Abstract:
The connection between fundamental interactions acting in molecules in a fluid and macroscopically measured properties, such as the viscosity between colloidal particles coated with polymers, is studied here. The role that hydrodynamic and Brownian forces play in colloidal dispersions is also discussed. It is argued that many body systems in which all these interactions take place can be accuratel…
▽ More
The connection between fundamental interactions acting in molecules in a fluid and macroscopically measured properties, such as the viscosity between colloidal particles coated with polymers, is studied here. The role that hydrodynamic and Brownian forces play in colloidal dispersions is also discussed. It is argued that many body systems in which all these interactions take place can be accurately solved using computational simulation tools. One of those modern tools is the technique known as dissipative particle dynamics, which incorporates Brownian and hydrodynamic forces, as well as basic conservative interactions. A case study is reported, as an example of the applications of this technique, which consists of the prediction of the viscosity and friction between two opposing parallel surfaces covered with polymer chains, under the influence of a steady flow. This work is intended to serve as an introduction to the subject of colloidal dispersions and computer simulations, for last year undergraduate students and beginning graduate students who are interested in beginning research in soft matter systems. To that end, a computational code is included that students can use right away to study complex fluids in equilibrium.
△ Less
Submitted 20 June, 2015;
originally announced June 2015.
-
Adsorption desorption processes on mesoscopic pores conected to microscopic pores of complex geometry using the Ising model
Authors:
M. A. Balderas Altamirano,
S. Cordero,
G. Roman,
A. Gama Goicochea
Abstract:
In this work we report studies of nitrogen adsorption and desorption onto solid surfaces using computer simulations of the three dimensional Ising model, for systems with complex porous structures at the mesoscopic and microscopic levels. A hysteresis cycle between the adsorption and desorption processes appears and we find that its characteristics are dependent on the geometry of the pore and on…
▽ More
In this work we report studies of nitrogen adsorption and desorption onto solid surfaces using computer simulations of the three dimensional Ising model, for systems with complex porous structures at the mesoscopic and microscopic levels. A hysteresis cycle between the adsorption and desorption processes appears and we find that its characteristics are dependent on the geometry of the pore and on the strength of the surface fluid interaction. We obtained also an average adsorption isotherm, which represents a combination of differently shaped pores, and shows robust jumps at certain values of the chemical potential as a consequence of the structures of the pores. Lastly, we compare our results with experimental data and also report the filling process of microscopic pores connected with mesopores. It is argued that these predictions are useful for researchers working on the enhanced recovery of oil and for the design of new nanomaterials, among others.
△ Less
Submitted 15 March, 2015;
originally announced March 2015.
-
The role of the dissipative and random forces in the calculation of the pressure of simple fluids with dissipative particle dynamics
Authors:
A. Gama Goicochea,
M. A. Balderas Altamirano,
J. D. Hernández,
E. Pérez
Abstract:
The role of viscous forces coupled with Brownian forces in momentum conserving computer simulations is studied here in the context of their contribution to the total average pressure of a simple fluid as derived from the virial theorem, in comparison with the contribution of the conservative force to the total pressure. The specific mesoscopic model used is the one known as dissipative particle dy…
▽ More
The role of viscous forces coupled with Brownian forces in momentum conserving computer simulations is studied here in the context of their contribution to the total average pressure of a simple fluid as derived from the virial theorem, in comparison with the contribution of the conservative force to the total pressure. The specific mesoscopic model used is the one known as dissipative particle dynamics, although our conclusions apply to similar models that obey the fluctuation dissipation theorem for short range interactions and have velocity dependent viscous forces. We find that the average contribution of the random and dissipative forces to the pressure is negligible for long simulations, provided these forces are appropriately coupled and when the finite time step used in the integration of the equation of motion is not too small. Finally, we study the properties of the fluid when the random force is made equal to zero and find that the system freezes as a result of the competition of the dissipative and conservative forces.
△ Less
Submitted 6 November, 2014;
originally announced November 2014.
-
Folding of the apolipoprotein A1 driven by the salt concentration as a possible mechanism to improve cholesterol trap**
Authors:
M. A. Balderas Altamirano,
A. Gama Goicochea,
E. Pérez
Abstract:
The folding of the cholesterol trap** apolipoprotein A1 in aqueous solution at increasing ionic strength is studied using atomically detailed molecular dynamics simulations. We calculate various structural properties to characterize the conformation of the protein, such as the radius of gyration, the radial distribution function and the end to end distance. Additionally we report information usi…
▽ More
The folding of the cholesterol trap** apolipoprotein A1 in aqueous solution at increasing ionic strength is studied using atomically detailed molecular dynamics simulations. We calculate various structural properties to characterize the conformation of the protein, such as the radius of gyration, the radial distribution function and the end to end distance. Additionally we report information using tools specifically tailored for the characterization of proteins, such as the mean smallest distance matrix and the Ramachandran plot. We find that two qualitatively different configurations of this protein are preferred, one where the protein is extended, and one where it forms loops or closed structures. It is argued that the latter promote the association of the protein with cholesterol and other fatty acids.
△ Less
Submitted 11 June, 2014;
originally announced June 2014.
