Explainable classifier for improving the accountability in decision-making for colorectal cancer diagnosis from histopathological images.

Pathologists are responsible for cancer type diagnoses from histopathological cancer tissues. However, it is known that microscopic examination is tedious and time-consuming. In recent years, a long list of machine learning approaches to image classification and whole-slide segmentation has been developed to support pathologists. Although many showed exceptional performances, the majority of them are not able to rationalize their decisions. In this study, we developed an explainable classifier to support decision making for medical diagnoses. The proposed model does not provide an explanation about the causality between the input and the decisions, but offers a human-friendly explanation about the plausibility of the decision. Cumulative Fuzzy Class Membership Criterion (CFCMC) explains its decisions in three ways: through a semantical explanation about the possibilities of misclassification, showing the training sample responsible for a certain prediction and showing training samples from conflicting classes. In this paper, we explain about the mathematical structure of the classifier, which is not designed to be used as a fully automated diagnosis tool but as a support system for medical experts. We also report on the accuracy of the classifier against real world histopathological data for colorectal cancer. We also tested the acceptability of the system through clinical trials by 14 pathologists. We show that the proposed classifier is comparable to state of the art neural networks in accuracy, but more importantly it is more acceptable to be used by human experts as a diagnosis tool in the medical domain.

Self-ordering of chemisorbed PTCDA molecules on Ge(001) driven by repulsive forces.

Realization of future hybrid electronic devices combining organic and inorganic semiconductors requires a well-defined interface between both components. Such an interface can be formed generally by self-ordering of organic molecules on inorganic substrates, which is usually hindered by strong covalent bonds to the semiconductor surface. In this paper, the 3,4,9,10-perylenetetracarboxylic dianhydride (PTCDA) molecules were unexpectedly found to form a locally self-ordered monolayer on a strongly interacting semiconductor surface of the Ge(001). Molecular arrangements with preferential separations between the molecules were observed by the scanning tunneling microscopy at various coverages of the molecules and substrate temperatures, suggesting strong inter-molecular interaction. Atomic structures of two paired molecules and their inter-molecular interaction energies in five different configurations were calculated by density functional theory. Simple Monte Carlo simulations show that mobility of molecules activated only by the inter-molecular interactions is sufficient to reproduce the local self-ordering. A dominant inter-molecular interaction between neighboring chemisorbed molecules has mostly positive energy (destabilizing) except for a single configuration, which leads to the formation of one-dimensional chains of the molecules and finally a periodic two-dimensional array by increasing the coverage.

Adsorption of ethylene on Sn and In terminated Si(001) surface studied by photoelectron spectroscopy and scanning tunneling microscopy.

Interaction of ethylene (C2H4) with Si(001)-Sn-2 × 2 and Si(001)-In-2 × 2 at room temperature has been studied using core level (C 1s) X-ray photoelectron spectroscopy with synchrotron radiation and scanning tunneling microscopy. Sn and In form similar dimer chains on Si(001)2 × 1, but exhibit different interaction with ethylene. While ethylene adsorbs on top of Sn dimers of the Si(001)-Sn-2 × 2 surface, the Si(001)-In-2 × 2 surface turned out to be inert. Furthermore, the reactivity of the Sn terminated surface is found to be considerably decreased in comparison with Si(001)2 × 1. According to the proposed adsorption model ethylene bonds to Sn dimers via [2 + 2] cycloaddition by interacting with their π dimer bonds. In contrast, indium dimers do not contain π bonds, which renders the In terminated Si(001) surface inert for ethylene adsorption.

Resolving the intrinsic short-range ordering of K+ ions on cleaved muscovite mica.

Muscovite mica, KAl2(Si3Al)O10(OH)2, is a common layered phyllosilicate with perfect cleavage planes. The atomically flat surfaces obtained through cleaving lend themselves to scanning probe techniques with atomic resolution and are ideal to model minerals and clays. Despite the importance of the cleaved mica surfaces, several questions remain unresolved. It is established that K+ ions decorate the cleaved surface, but their intrinsic ordering - unaffected by the interaction with the environment - is not known. This work presents clear images of the K+ distribution of cleaved mica obtained with low-temperature non-contact atomic force microscopy (AFM) under ultra-high vacuum (UHV) conditions. The data unveil the presence of short-range ordering, contrasting previous assumptions of random or fully ordered distributions. Density functional theory (DFT) calculations and Monte Carlo simulations show that the substitutional subsurface Al3+ ions have an important role for the surface K+ ion arrangement.

Growth-rate model of epitaxial layer-by-layer growth by pulsed-laser deposition.

We present a numerical model of epitaxial thin-film growth applicable for pulsed-laser deposition on a single crystalline substrate. The model is based on rate equations describing the time development of monolayer coverages and of densities of movable particles on atomically flat terraces. Numerical solution of the equations showed that the time dependence of surface roughness obeys a scaling law, the exponent of which depends on probabilities of various atomistic processes included in the simulation model. From the time dependence of monolayer coverages we calculated x-ray diffracted intensity in a quasiforbidden anti-Bragg reflection and showed that its oscillatory behavior is affected by these probabilities as well. The results show the possibility to study atomistic processes during the deposition from the time dependence of the anti-Bragg intensity measured during deposition.

Effect of pulse laser frequency on PLD growth of LuFeO3 explained by kinetic simulations of in-situ diffracted intensities.

Atomistic processes during pulsed-laser deposition (PLD) growth influence the physical properties of the resulting films. We investigated the PLD of epitaxial layers of hexagonal LuFeO[Formula: see text] by measuring the X-ray diffraction intensity in the quasiforbidden reflection 0003 in situ during deposition. From measured X-ray diffraction intensities we determined coverages of each layer and studied their time evolution which is described by scaling exponent [Formula: see text] directly connected to the surface roughness. Subsequently we modelled the growth using kinetic Monte Carlo simulations. While the experimentally obtained scaling exponent [Formula: see text] decreases with the laser frequency, the simulations provided the opposite behaviour. We demonstrate that the increase of the surface temperature caused by impinging ablated particles satisfactorily explains the recorded decrease in the scaling exponent with the laser frequency. This phenomena is often overlooked during the PLD growth.

Effective algorithm for simulations of layer-by-layer growth during pulsed-laser deposition.

The atomistic simulation of materials growing in the layer-by-layer mode by the pulsed-laser deposition is a significant challenge mainly due to the short timescales in which the fastest processes on the surface occur together with long periods between pulses. We present a kinetic Monte Carlo algorithm which overcomes the scaling problem by approximation of fast diffusion and by neglecting complex chemical processes. The atomic diffusion is modeled as a two-dimensional gas of material units on each layer. The model is based on a few elementary processes-the condensation of units on the surface, their dissolution back to the gas, and interlayer transport, which can be influenced by the Ehrlich-Schwoebel barrier. With these simplifications, the computational time of the algorithm scales only linearly with the size of the substrate while describing physically relevant growth kinetics. We demonstrate that the simplified model is suitable for simulations of layered growth of thin films in the range from quasicontinuous deposition to low-frequency cases. The model is successfully implemented to provide an alternative explanation of the time evolution of layer coverages by interlayer transport after pulses of deposition experimentally observed during perovskite growth [G. Eres et al., Phys. Rev. B 84, 195467 (2011)PRBMDO1098-012110.1103/PhysRevB.84.195467].

Self-assembly of a two-dimensional molecular layer in a nonhomogeneous electric field: Kinetic Monte Carlo simulations.

Behavior of mobile adsorbed species can be affected by the presence of a strong non-homogeneous electric field. Such a field exists in the proximity of a biased tip of the scanning tunneling microscope. Depending on the electronic properties of the adsorbate and the polarity of the electric field, self-assembly of ordered structures on the surface can be facilitated or prevented. We use a kinetic Monte Carlo model to investigate the effect of the electric field on the assembly of planar molecules on weakly interacting surfaces. Using phthalocyanine molecules as a model system, we study mechanisms behind the results of our previous experimental study of electric-field controlled switching of molecular arrays. The complex interplay between intermolecular and field-molecule interactions is illustrated by detailed phase diagrams at various surface coverages.

Local electronic structure of doping defects on Tl/Si(111)1x1.

The Tl/Si(111)1 × 1 surface is a representative of a 2D layer with Rashba-type spin-split electronic bands. To utilize the spin polarization, doping of the system should be understood on atomic level. We present a study of two types of atomic defects predicted to dope the considered electronic system - Si-induced vacancies and defects associated with the presence of extra Tl atoms. Structural calculations based on density functional theory (DFT) confirm the stability of the proposed defect structure consisting of an extra Si atom and missing seven Tl atoms as proposed in an earlier experimental study. The calculated spatial charge distributions indicate an enhancement of the charge around the extra Si atom, which correctly reproduces topographies of the corresponding scanning tunneling microscopy images while the calculated local densities of states of this system explain obtained scanning tunneling spectra. The DFT structural calculations let us determine the atomic structure of the defect caused by the presence of an extra Tl atom. The calculated spatial charge distributions show a ring-like feature around the extra Tl atom. The obtained results indicate a charge transfer from the central extra Tl atom to its vicinity in the agreement with earlier photoemission measurements.

Electric-field-controlled phase transition in a 2D molecular layer.

Self-assembly of organic molecules is a mechanism crucial for design of molecular nanodevices. We demonstrate unprecedented control over the self-assembly, which could allow switching and patterning at scales accessible by lithography techniques. We use the scanning tunneling microscope (STM) to induce a reversible 2D-gas-solid phase transition of copper phthalocyanine molecules on technologically important silicon surface functionalized by a metal monolayer. By means of ab-initio calculations we show that the charge transfer in the system results in a dipole moment carried by the molecules. The dipole moment interacts with a non-uniform electric field of the STM tip and the interaction changes the local density of molecules. To model the transition, we perform kinetic Monte Carlo simulations which reveal that the ordered molecular structures can form even without any attractive intermolecular interaction.

Desorption-induced structural changes of metal/Si(111) surfaces: kinetic Monte Carlo simulations.

We used a configuration-based kinetic Monte Carlo model to explain important features related to formation of the (√3×√3)R30° mosaic of metal and semiconductor atoms on the Si(111) surface. Using first-order desorption processes, we simulate the surprising zero-order desorption spectra, reported in some cases of metal desorption from the Si(111) surface. We show that the mechanism responsible for the zerolike order of desorption is the enhanced desorption from disordered areas. Formation of the √3×√3 mosaic with properties of a strongly frustrated antiferromagnetic Ising model is simulated by a configuration-sensitive desorption. For substitution of desorbed metal atoms by Si adatoms, fast diffusion of the adatoms on top of a 1×1 layer is proposed as the most probable. Simulations of desorption-induced structural transitions provide us a link between underlying atomistic processes and the observed evolving morphologies with resultant macroscopic desorption fluxes. An effect of the desorption sensitivity on a configuration of neighboring atoms is emphasized.

Modeling growth of one-dimensional islands: influence of reactive defects.

Influence of reactive defects on size distribution of one-dimensional islands is studied by means of kinetic Monte Carlo simulations in combination with an analytical approach. Two different models are examined: a model with anisotropically diffusing atoms irreversibly aggregating to islands, and a reversible model close to thermal equilibrium which allows atom detachment from islands during the growth. The models can be used to simulate island growth of group III metals deposited on the Si(100)2 x 1 surface at room temperature: Al, Ga (irreversible model), and In (equilibrium model). We demonstrate that concentration of the reactive defects 0.0025 per site may change the island size distribution from monomodal to monotonically decreasing in the case of the irreversible model. At concentration >or=0.005 defects per site, a difference between results of the studied models is suppressed by the influence of the defects and similar island size distributions are obtained.

Direct observation of long-range assisted formation of Ag clusters on Si(111)7 x 7.

Formation of Ag clusters on reconstructed surface Si(111)7 x 7 was for the first time observed in real time during deposition by means of scanning tunneling microscopy. The sequences of images taken at room temperature show mechanisms controlling the growth and behavior of individual Ag adatoms. Obtained data reveal new details of attractive interaction between adsorbates occupying adjacent half-unit cells of the 7 x 7 reconstruction. Time evolution of growth characteristics was simulated by means of a simple model. The growth scenario observed in vivo is discussed with respect to previously reported models based on data obtained after finishing the deposition--post-mortem.

Ga-rich limit of surface reconstructions on GaAs(001): atomic structure of the (4 x 6) phase.

The Ga-rich reconstruction of the GaAs(001) surface has been studied. Using scanning tunneling microscopy (STM), we have found the existence of a well-ordered (4 x 6) reconstruction under extreme Ga-rich conditions. A structure model, consisting of subsurface Ga-Ga dimers and surface Ga-As dimers, is proposed for the (4 x 6) surface. This model is found to be energetically favorable at the Ga-rich limit and agrees well with our experimental data from STM and reflection high-energy electron diffraction.

Kinetics in surface reconstructions on GaAs(001).

We have successfully controlled the surface structures of GaAs(001) by changing incident As-molecular species. Under As4 fluxes, the c(4 x 4) reconstruction with Ga-As dimers [c(4 x 4)alpha structure] is obtained, but the formation of three As-As dimer structures [c(4 x 4)beta structure] is kinetically limited. On the other hand, the structure change from the (2 x 4), through c(4 x 4)alpha, to c(4 x 4)beta phases is observed under As2 fluxes. We found that the c(4 x 4)alpha structure is energetically metastable and provides a kinetic pathway for the structure change between the (2 x 4) and c(4 x 4)beta phases under As2 fluxes.