arXiv subjects

Sol Lee

Publications and source records attributed to Sol Lee.

13 recordsLinked to original sources

arXiv2026

RaLMPH: Reliability-aware Learning for Multi-Pathologist Harmonization in Whole-Slide Image Classification

Multiple Instance Learning (MIL) is a standard paradigm for Whole-Slide Image (WSI) analysis and has achieved strong results in computational pathology. However, most MIL pipelines assume a single "gold" label per slide, which conflicts with clinical practice where substantial inter-pathologist variability is common. Existing multi-annotator learning and label-refinement methods typically estimate global annotator reliability or rely on single-instance assumptions, making them poorly suited to MIL and to localized diagnostic contexts where experts disagree. We propose RaLMPH (Reliability-aware Learning for Multi-Pathologist Harmonization), a MIL-based label reconciliation framework for WSIs annotated by multiple pathologists. RaLMPH introduces a reliability field that jointly models (i) local neighborhood structure in WSI feature space and (ii) expert uncertainty (entropy), enabling per-sample identification of trustworthy reference neighborhoods. Leveraging this field, RaLMPH performs sample-wise local annotator ranking to select reliable opinions per slide and applies an adaptive gating mechanism to fuse labels conditioned on local reliability. Experiments on a clinical WSI dataset with labels from six pathologists, as well as controlled simulated benchmarks, show that RaLMPH consistently outperforms existing approaches. Further analyses clarify how our reliability-aware mechanism improves label reconciliation and downstream MIL performance.

cs.CV↗

arXiv2026

Every Error has Its Magnitude: Asymmetric Mistake Severity Training for Multiclass Multiple Instance Learning

Multiple Instance Learning (MIL) has emerged as a promising paradigm for Whole Slide Image (WSI) diagnosis, offering effective learning with limited annotations. However, existing MIL frameworks overlook diagnostic priorities and fail to differentiate the severity of misclassifications in multiclass, leaving clinically critical errors unaddressed. We propose a mistake-severity-aware training strategy that organizes diagnostic classes into a hierarchical structure, with each level optimized using a severity-weighted cross-entropy loss that penalizes high-severity misclassifications more strongly. Additionally, hierarchical consistency is enforced through probabilistic alignment, a semantic feature remix applied to the instance bag to robustly train class priority and accommodate clinical cases involving multiple symptoms. An asymmetric Mikel's Wheel-based metric is also introduced to quantify the severity of errors specific to medical fields. Experiments on challenging public and real-world in-house datasets demonstrate that our approach significantly mitigates critical errors in MIL diagnosis compared to existing methods. We present additional experimental results on natural domain data to demonstrate the generalizability of our proposed method beyond medical contexts.

cs.CV↗

arXiv2025

Diagnose Like A REAL Pathologist: An Uncertainty-Focused Approach for Trustworthy Multi-Resolution Multiple Instance Learning

With the increasing demand for histopathological specimen examination and diagnostic reporting, Multiple Instance Learning (MIL) has received heightened research focus as a viable solution for AI-centric diagnostic aid. Recently, to improve its performance and make it work more like a pathologist, several MIL approaches based on the use of multiple-resolution images have been proposed, delivering often higher performance than those that use single-resolution images. Despite impressive recent developments of multiple-resolution MIL, previous approaches only focus on improving performance, thereby lacking research on well-calibrated MIL that clinical experts can rely on for trustworthy diagnostic results. In this study, we propose Uncertainty-Focused Calibrated MIL (UFC-MIL), which more closely mimics the pathologists' examination behaviors while providing calibrated diagnostic predictions, using multiple images with different resolutions. UFC-MIL includes a novel patch-wise loss that learns the latent patterns of instances and expresses their uncertainty for classification. Also, the attention-based architecture with a neighbor patch aggregation module collects features for the classifier. In addition, aggregated predictions are calibrated through patch-level uncertainty without requiring multiple iterative inferences, which is a key practical advantage. Against challenging public datasets, UFC-MIL shows superior performance in model calibration while achieving classification accuracy comparable to that of state-of-the-art methods.

cs.CV↗

arXiv2025

Priority-Aware Clinical Pathology Hierarchy Training for Multiple Instance Learning

Multiple Instance Learning (MIL) is increasingly being used as a support tool within clinical settings for pathological diagnosis decisions, achieving high performance and removing the annotation burden. However, existing approaches for clinical MIL tasks have not adequately addressed the priority issues that exist in relation to pathological symptoms and diagnostic classes, causing MIL models to ignore priority among classes. To overcome this clinical limitation of MIL, we propose a new method that addresses priority issues using two hierarchies: vertical inter-hierarchy and horizontal intra-hierarchy. The proposed method aligns MIL predictions across each hierarchical level and employs an implicit feature re-usability during training to facilitate clinically more serious classes within the same level. Experiments with real-world patient data show that the proposed method effectively reduces misdiagnosis and prioritizes more important symptoms in multiclass scenarios. Further analysis verifies the efficacy of the proposed components and qualitatively confirms the MIL predictions against challenging cases with multiple symptoms.

cs.CV↗

arXiv2025

CP-Dilatation: A Copy-and-Paste Augmentation Method for Preserving the Boundary Context Information of Histopathology Images

Medical AI diagnosis including histopathology segmentation has derived benefits from the recent development of deep learning technology. However, deep learning itself requires a large amount of training data and the medical image segmentation masking, in particular, requires an extremely high cost due to the shortage of medical specialists. To mitigate this issue, we propose a new data augmentation method built upon the conventional Copy and Paste (CP) augmentation technique, called CP-Dilatation, and apply it to histopathology image segmentation. To the well-known traditional CP technique, the proposed method adds a dilation operation that can preserve the boundary context information of the malignancy, which is important in histopathological image diagnosis, as the boundary between the malignancy and its margin is mostly unclear and a significant context exists in the margin. In our experiments using histopathology benchmark datasets, the proposed method was found superior to the other state-of-the-art baselines chosen for comparison.

eess.IV↗

arXiv2025

Towards Classifying Histopathological Microscope Images as Time Series Data

As the frontline data for cancer diagnosis, microscopic pathology images are fundamental for providing patients with rapid and accurate treatment. However, despite their practical value, the deep learning community has largely overlooked their usage. This paper proposes a novel approach to classifying microscopy images as time series data, addressing the unique challenges posed by their manual acquisition and weakly labeled nature. The proposed method fits image sequences of varying lengths to a fixed-length target by leveraging Dynamic Time-series Warping (DTW). Attention-based pooling is employed to predict the class of the case simultaneously. We demonstrate the effectiveness of our approach by comparing performance with various baselines and showcasing the benefits of using various inference strategies in achieving stable and reliable results. Ablation studies further validate the contribution of each component. Our approach contributes to medical image analysis by not only embracing microscopic images but also lifting them to a trustworthy level of performance.

cs.CV↗

arXiv2023

Electrical transport properties driven by unique bonding configuration in gamma-GeSe

Group-IV monochalcogenides have recently shown great potential for their thermoelectric, ferroelectric, and other intriguing properties. The electrical properties of group-IV monochalcogenides exhibit a strong dependence on the chalcogen type. For example, GeTe exhibits high doping concentration, whereas S/Se-based chalcogenides are semiconductors with sizable bandgaps. Here, we investigate the electrical and thermoelectric properties of gamma-GeSe, a recently identified polymorph of GeSe. gamma-GeSe exhibits high electrical conductivity (~106 S/m) and a relatively low Seebeck coefficient (9.4 uV/K at room temperature) owing to its high p-doping level (5x1021 cm-3), which is in stark contrast to other known GeSe polymorphs. Elemental analysis and first-principles calculations confirm that the abundant formation of Ge vacancies leads to the high p-doping concentration. The magnetoresistance measurements also reveal weak-antilocalization because of spin-orbit coupling in the crystal. Our results demonstrate that gamma-GeSe is a unique polymorph in which the modified local bonding configuration leads to substantially different physical properties.

cond-mat.mtrl-sci↗

arXiv2022

In-Plane and Out-of-Plane Excitonic Coupling in 2D Molecular Crystals

Understanding the nature of molecular excitons in low-dimensional molecular solids is of paramount importance in fundamental photophysics and various applications such as energy harvesting, switching electronics and display devices. Despite this, the spatial evolution of molecular excitons and their transition dipoles have not been captured in the precision of molecular length scales. Here we show in-plane and out-of-plane excitonic evolution in quasilayered two-dimensional (2D) perylene-3, 4, 9, 10-tetracarboxylic dianhydride (PTCDA) crystals assembly-grown on hexagonal boron nitride (BN) crystals. Complete lattice constants with orientations of two herringbone-configured basis molecules are determined with polarization-resolved spectroscopy and electron diffraction methods. In the truly 2D limit of single layers, two Frenkel emissions Davydov-split by Kasha-type intralayer coupling exhibit energy inversion with decreasing temperature, which enhances excitonic coherence. As the thickness increases, the transition dipole moments of newly emerging charge transfer excitons are reoriented because of mixing with the Frenkel states. The current spatial anatomy of 2D molecular excitons will inspire a deeper understanding and groundbreaking applications of low-dimensional molecular systems.

cond-mat.mtrl-sci↗

arXiv2022

Atomically Sharp, Closed Bilayer Phosphorene Edges by Self-Passivation

Two-dimensional (2D) crystals' edge structures not only influence their overall properties but also dictate their formation due to edge-mediated synthesis and etching processes. Edges must be carefully examined because they often display complex, unexpected features at the atomic scale, such as reconstruction, functionalization, and uncontrolled contamination. Here, we examine atomic-scale edge structures and uncover reconstruction behavior in bilayer phosphorene. We use in situ transmission electron microscopy (TEM) of phosphorene/graphene specimens at elevated temperatures to minimize surface contamination and reduce e-beam damage, allowing us to observe intrinsic edge configurations. Bilayer zigzag (ZZ) edge was found the most stable edge configuration under e-beam irradiation. Through first-principles calculations and TEM image analysis under various tilting and defocus conditions, we find that bilayer ZZ edges undergo edge reconstruction and so acquire closed, self-passivated edge configurations. The extremely low formation energy of the closed bilayer ZZ edge and its high stability against e-beam irradiation are confirmed by first-principles calculations. Moreover, we fabricate bilayer phosphorene nanoribbons with atomically-sharp closed ZZ edges. The identified bilayer ZZ edges will aid in the fundamental understanding of the synthesis, degradation, reconstruction, and applications of phosphorene and related structures.

cond-mat.mes-hall↗

arXiv2022

STEM image analysis based on deep learning: identification of vacancy defects and polymorphs of ${MoS_2}$

Scanning transmission electron microscopy (STEM) is an indispensable tool for atomic-resolution structural analysis for a wide range of materials. The conventional analysis of STEM images is an extensive hands-on process, which limits efficient handling of high-throughput data. Here we apply a fully convolutional network (FCN) for identification of important structural features of two-dimensional crystals. ResUNet, a type of FCN, is utilized in identifying sulfur vacancies and polymorph types of ${MoS_2}$ from atomic resolution STEM images. Efficient models are achieved based on training with simulated images in the presence of different levels of noise, aberrations, and carbon contamination. The accuracy of the FCN models toward extensive experimental STEM images is comparable to that of careful hands-on analysis. Our work provides a guideline on best practices to train a deep learning model for STEM image analysis and demonstrates FCN's application for efficient processing of a large volume of STEM data.

cond-mat.mes-hall↗

arXiv2021

{\gamma}-GeSe:a new hexagonal polymorph from group IV-VI monochalcogenides

The family of group IV-VI monochalcogenides has an atomically puckered layered structure, and their atomic bond configuration suggests the possibility for the realization of various polymorphs. Here, we report the synthesis of the first hexagonal polymorph from the family of group IV-VI monochalcogenides, which is conventionally orthorhombic. Recently predicted four-atomic-thick hexagonal GeSe, so-called {\gamma}-GeSe, is synthesized and clearly identified by complementary structural characterizations, including elemental analysis, electron diffraction, high-resolution transmission electron microscopy imaging, and polarized Raman spectroscopy. The electrical and optical measurements indicate that synthesized {\gamma}-GeSe exhibits high electrical conductivity of 3x10^5 S/m, which is comparable to those of other two-dimensional layered semimetallic crystals. Moreover, {\gamma}-GeSe can be directly grown on h-BN substrates, demonstrating a bottom-up approach for constructing vertical van der Waals heterostructures incorporating {\gamma}-GeSe. The newly identified crystal symmetry of {\gamma}-GeSe warrants further studies on various physical properties of {\gamma}-GeSe.

cond-mat.mes-hall↗

arXiv2021

Single-Crystalline Metallic Films Induced by van der Waals Epitaxy on Black Phosphorus

The properties of metal-semiconductor junctions are often unpredictable because of non-ideal interfacial structures, such as interfacial defects or chemical reactions introduced at junctions. Black phosphorus (BP), an elemental two-dimensional (2D) semiconducting crystal, possesses the puckered atomic structure with high chemical reactivity, and the establishment of a realistic atomic-scale picture of BP's interface toward metallic contact has remained elusive. Here we examine the interfacial structures and properties of physically-deposited metals of various kinds on BP. We find that Au, Ag, and Bi form single-crystalline films with (110) orientation through guided van der Waals epitaxy. Transmission electron microscopy and X-ray photoelectron spectroscopy confirm that atomically sharp van der Waals metal-BP interfaces forms with exceptional rotational alignment. Under a weak metal-BP interaction regime, the BP's puckered structure play an essential role in the adatom assembly process and can lead to the formation of a single crystal, which is supported by our theoretical analysis and calculations. The experimental survey also demonstrates that the BP-metal junctions can exhibit various types of interfacial structures depending on metals, such as the formation of polycrystalline microstructure or metal phosphides. This study provides a guideline for obtaining a realistic view on metal-2D semiconductor interfacial structures, especially for atomically puckered 2D crystals.

cond-mat.mtrl-sci↗

arXiv2019

Fabrication and Imaging of Monolayer Phosphorene with Preferred Edge Configurations via Graphene-Assisted Layer-by-Layer Thinning

Phosphorene, a monolayer of black phosphorus (BP), is an elemental two-dimensional material with interesting physical properties, such as high charge carrier mobility and exotic anisotropic in-plane properties. To fundamentally understand these various physical properties, it is critically important to conduct an atomic-scale structural investigation of phosphorene, particularly regarding various defects and preferred edge configurations. However, it has been challenging to investigate mono- and few-layer phosphorene because of technical difficulties arising in the preparation of a high-quality sample and damages induced during the characterization process. Here, we successfully fabricate high-quality monolayer phosphorene using a controlled thinning process with transmission electron microscopy, and subsequently perform atomic-resolution imaging. Graphene protection suppresses the e-beam-induced damage to multi-layer BP and one-side graphene protection facilitates the layer-by-layer thinning of the samples, rendering high-quality monolayer and bilayer regions. We also observe the formation of atomic-scale crystalline edges predominantly aligned along the zigzag and (101) terminations, which is originated from edge kinetics under e-beam-induced sputtering process. Our study demonstrates a new method to image and precisely manipulate the thickness and edge configurations of air-sensitive two-dimensional materials.

cond-mat.mes-hall↗