arXiv subjects

Yoshiaki Nakamura

Publications and source records attributed to Yoshiaki Nakamura.

3 recordsLinked to original sources

arXiv2026

Enabling clinical use of foundation models for computational pathology

Foundation models for computational pathology are expected to facilitate the development of high-performing, generalisable deep learning systems. However, in addition to biologically relevant features, current foundation models also capture pre-analytic and scanner-specific variation that bias the predictions made by downstream task-specific models trained on these features. Here we show that introducing novel robustness losses during downstream model training reduces sensitivity to technical variability. A purpose-designed comprehensive experimentation setup with 27,042 whole-slide images from 6,155 patients is used to train thousands of models from the features of eight well-known foundation models for computational pathology. In addition to a substantial improvement in robustness, our approach improves classification accuracy by focusing on biologically relevant features. It mitigates robustness limitations of foundation models for computational pathology without retraining the foundation models themselves, enabling development of models that are more suitable in real-world clinical use.

cs.CV↗

arXiv2022

Phonon Interference at the Atomic Scale

Phonons diffraction and interference patterns are observed at the atomic scale, using molecular dynamics simulations in systems containing crystalline silicon and nanometric obstacles as voids or amorphous-inclusions. The diffraction patterns caused by these nano-architectured systems of the same order as the phonon wavelengths are similar to the ones predicted by a simple Fresnel-Kirchhoff integral, with a few differences due to the nature of the obstacle and the anisotropy of crystalline silicon. These findings give evidence of the wave nature of phonons, can help to a better comprehension of the interaction of phonons with nanoobjects and at long term can be useful for intelligent thermal management and phonon frequency filtering at the nanoscale.

cond-mat.mes-hall↗

arXiv2018

Ultimate confinement of phonon propagation in silicon nano-crystalline structure

Temperature-dependent thermal conductivity of epitaxial silicon nano-crystalline (SiNC) structures composed of nanometer-sized grains separated by ultra-thin silicon-oxide (SiO2) films is measured by the time domain thermoreflectance technique in the range from 50 to 300 K. Thermal conductivity of SiNC structures with grain size of 3 nm and 5 nm is anomalously low at the entire temperature range, significantly below the values of bulk amorphous Si and SiO2. Phonon gas kinetics model, with intrinsic transport properties obtained by first-principles-based anharmonic lattice dynamics and phonon transmittance across ultra-thin SiO2 films obtained by atomistic Green's function, reproduces the measured thermal conductivity without any fitting parameters. The analysis reveals that mean free paths of acoustic phonons in the SiNC structures are equivalent or even below half the phonon wavelength, i.e. the minimum thermal conductivity scenario. The result demonstrates that the nanostructures with extremely small length scales and controlled interface can give rise to ultimate classical confinement of thermal phonon propagation.

cond-mat.mes-hall↗