Hepatocellular carcinoma (HCC), the predominant form of primary liver cancer, significantly threatens to global health. Despite considerable advances in diagnostic and therapeutic approaches in recent years, the prognosis for patients with HCC remains unsatisfactory. The emergence of artificial intelligence (AI), particularly deep learning technologies, offers new hope for improving the diagnosis and treatment of HCC. Researchers have extensively explored ways to integrate deep learning models into the clinical management of HCC patients, which provides a valuable foundation for developing more personalized treatment strategies. Compared with other detection methods, computed tomography (CT) has attracted significant research interest because of its comprehensive advantages, including wide availability and high resolution, making it well suited for AI-powered analysis. This review systematically integrates deep learning technologies for HCC based on CT imaging, while focusing primarily on tumor diagnosis, segmentation, treatment response prediction, and patient prognosis prediction. Moreover, we review popular deep learning networks in various fields and describe the advantages of these prevalent deep learning models for different applications. Furthermore, we discuss the outstanding challenges in applying deep learning to extract information from CT images for the diagnosis and treatment of HCC patients. These insights could provide guidance for subsequent studies.

As one of the ten major categories on the UNESCO's Representative List of Intangible Cultural Heritage of Humanity, traditional opera is a crystallization of human civilization. ‌ The Chinese Traditional Opera Video Super-Resolution (CTOVSR) was developed to protect these precious and irreplaceable aged Chinese opera videos. We analyzed the entire degradation process throughout video lifecycle in this paper. By utilizing high-resolution (HR) videos from professionally restored films and their corresponding low-resolution (LR) versions distributed online, we proposed a novel construction method for LR-HR video sequence pairs, named "Real-world+". This method ensures that the pairs accurately reflect the real-world degradation process and are strictly aligned both spatially and temporally. We further augmented CTOVSR with synthetically degraded data, resulting in 900 LR-HR video sequence pairs, each pair containing 100 consecutive frames, featuring various unique elements of Chinese traditional opera. While the primary focus of this work is the dataset itself, our proposed dataset construction methodology also offers a valuable practical approach for the preservation of other types of precious historical heritage.

Blood T1 is a key parameter for hemodynamic quantification in both non-contrast- and contrast-enhanced imaging. Individual vessel T1 has been measured using a modified Look-Locker scheme with multi-shot EPI or FLASH in high spatial resolution, requiring ∼1 min. Here, by exploiting the temporal sparsity from the excessive number of inversion delays, we apply Golden Angle rotated Spiral k-t Sparse Parallel imaging (GASSP) to enable blood T1 measurement in a single shot of 10 s. The pulse sequence with single-shot GASSP reconstruction was developed for T1 measurement from the internal jugular vein (IJV) with 1 × 1 mm2 in-plane resolution. On nine healthy volunteers, the single-shot GASSP was compared to the segmented EPI readout and was repeated to assess its intra-scan reproducibility. Another experiment was performed on three patients, during which the 10 s GASSP was obtained at different time points prior to and following the Gadolinium (Gd) administration to assess dynamic changes in blood T1. The blood T1 values measured with the highly undersampled GASSP method were strongly correlated (r = 0.83) with those using the multi-shot EPI readout and exhibited high reproducibility (r = 0.88) within the session. The baseline IJV T1 values measured were 1700-2000 ms. Following the Gd injection, the T1 values of IJVs gradually recovered from ∼300-400 to ∼500 ms within 10-15 min. The feasibility of an ultrafast blood T1 measurement was demonstrated with high spatial resolution in a single shot of 10 s, applicable to both pre- and post-contrast conditions.

Ferritin is a ubiquitous and evolutionarily conserved iron-storage protein that plays a fundamental role in cellular iron homeostasis. By catalyzing the oxidation of ferrous iron and sequestering it as a ferric mineral within a protein nanocage, ferritin prevents toxic accumulation of labile iron and reactive oxygen species that damage proteins, lipids, and DNA. In humans, ferritin assembles into a 24-subunit nearly spherical shell enclosing a central cavity that safely stores thousands of iron atoms. This organized architecture enables ferritin to act as both an efficient iron detoxification system and a dynamic intracellular iron reservoir. Recent advances in cryo-electron microscopy (cryo-EM) have transformed ferritin research by revealing its structural organization, molecular interactions, and functional states at high resolution. Additionally, beyond protein-protein interactions, cryo-EM now enables direct visualization of ferritin-mediated biomineralization, allowing in situ observation of iron nucleation, mineral growth, and core organization within intact nanocages. Together, these advances establish cryo-EM as a transformative tool for elucidating ferritin structure, dynamics, and function - reshaping our understanding of iron metabolism and guiding the rational design of ferritin-based nanomaterials for biomedical applications.

Abstract Hyperspectral imaging acquires spatially resolved spectral signatures, enabling a wide range of applications from scientific research to industrial processes. Traditional microelectron-mechanical systems (MEMS) Fabry–Pérot (FP) spectrometers offer a compact and simple design but are limited by single free spectral range (FSR) operation. This limitation introduces a fundamental trade-off: achieving high spectral resolution necessitates narrowing the operational bandwidth. Furthermore, maintaining such high resolution demands a larger number of sampling channels, which increases the acquisition time for a single hyperspectral image and thereby limits the frame rate. Here, we present a computational hyperspectral imaging framework that achieves broadband spectral coverage and high frame rate without sacrificing spectral resolution. By dynamically modulating the MEMS-FP cavity to span multiple FSRs, we generate a set of low-correlation spectral sampling patterns as spectral encoders. When combined with a tailored reconstruction algorithm, the system accurately decodes spectral information from a significantly reduced number of sampling channels. We experimentally validate the effectiveness of our system through LED array inspection, demonstrating its potential for high-throughput defect detection in LEDs or screen manufacturing lines. Our work presents a strategy that leverages rapidly advancing computational techniques to overcome the limitations of conventional hardware architectures in hyperspectral imaging. This compact and integrable solution is particularly well-suited for deployment in resource-constrained environments.

Abstract. There is increasing interest in global dynamic soil information with changes in soil properties mapped over time and at high spatial resolution. Thanks to long-term, multi-temporal, and fine- and medium-resolution satellite missions such as Landsat, MODIS, Copernicus Sentinel and similar, it is possible to produce globally consistent predictions of key soil variables that match other 10–30 m spatial resolution global data sets. This paper describes data preparation, modeling, and production of OpenLandMap-soildb: global dynamic predictions of soil organic carbon content, soil organic carbon density, bulk density, soil pH in H2O, soil texture fractions (clay, sand and silt) and USDA subgroup soil types (USDA soil taxonomy subgroups) at 30 m spatial resolution based on spatiotemporal Machine Learning (Quantile Regression Random Forest with output predictions showing the mean plus the 68 % probability lower and upper prediction intervals). To train the models, a large compilation of soil samples imported from legacy soil projects was used: 216 000 soil samples with soil carbon density (kg m−3), 408 000 soil samples with soil carbon content (g kg−1), 272 000 soil samples with soil pH in H2O, 363 000 soil samples with clay, silt and sand content (%) and 134 000 samples with bulk density oven dry (t m−3). Soil carbon and soil pH were mapped with 5-year time-intervals; soil texture fractions, bulk density, and soil types were mapped for recent years only. The cross-validation results indicate Root Mean Square Error (RMSE) of 17.7 (kg m−3; 0.486 in log-scale) and Concordance Correlation Coefficient (CCC) of 0.88 for SOC density, RMSE of 51.3 (g kg−1; 0.574 in log-scale) and CCC of 0.87 for SOC content, RMSE of 0.15 (t m−3) and CCC of 0.92 for bulk density of fine-earth, RMSE of 0.51 and CCC of 0.91 for soil pH, RMSE of 8.4 % and CCC of 0.87 for soil clay content, and RMSE of 12.6 % and CCC of 0.84 for soil sand content respectively. The most important variables for predicting soil organic carbon density (kg m−3) were: soil depth, Landsat-based uncalibrated Gross Primary Productivity (GPP), Normalized Difference Vegetation Index (NDVI) and CHELSA bioclimatic indices. The global distribution of soil pH can be primarily explained by the CHELSA Aridity Index (long-term), annual precipitation, and salinity grade. The global stocks for 2020–2022+ period for 0–30 cm depth interval are estimated at 461 Pg (Peta grams); the results further indicate that, in the last 25 years, the world has lost at least 11 Pg of SOC in the top soil. Suggestions are made on how to set up global permanent monitoring stations to accurately track land degradation and enable land restoration projects. The training data set is available at https://doi.org/10.5281/zenodo.4748499 (Hengl and Gupta, 2025), while the resulting data products can be accessed at https://doi.org/10.5281/zenodo.15470431 (Consoli et al., 2025) and https://world.soils.app (OpenGeoHub Foundation, 2026). Both datasets are released under a CC-BY license.

Advanced nanopore measurements allow structural probing of molecules with high spatial and temporal resolution. We report high signal-to-noise, 1-10 MHz bandwidth, translocation measurements of the multistate folding of heme protein cytochrome c in KCl solution through optimally designed silicon nitride pores of 2.3 - 3.3 nm diameter and 3.6-3.8 nm effective thickness, and an optimal concentration of a denaturant (Gdm-Cl). The pore diameter is slightly smaller than the protein's size, forcing the protein to squeeze through the pore. The sufficiently large pore thickness allows enough time for protein probing at an applied field of ∼ 250 kV/cm. Through Bayesian Information Criterion score analysis, current blockades reveal six distinct levels, attributed to specific protein states. We calculate the transition probabilities between the states and the conditional probabilities of the protein leaving the pore from each state. We validate the model by simulating events and comparing them with experimental data.

Current reference standards for measuring gastric emptying and motility are not considered optimal due to the time required, ionizing radiation, invasiveness, and spatial resolution. To assess gastric motility using novel real-time dynamic magnetic resonance imaging in combination with static measurements for gastric emptying and training of an automated deep-learning-based segmentation pipeline. Prospective. The study included 36 healthy volunteers (20 female, mean 24 ± 3 years) and three patients with diagnosed Crohn's disease. Participants ingested water to assess fasting motility and pineapple juice for the postprandial state. 3 T, 3D spoiled gradient echo (GRE) sequence and real-time spoiled GRE. Gastric emptying was measured by using the gastric volume, while motility was analyzed by tracking changes in the antrum's cross-sectional area and applying Fast Fourier Transformation. Segmentations were performed using a trained semantic segmentation model. Linear Mixed Model with continuous dependent variables and fixed effects. Models included a random intercept for participants. Statistical significance was defined as p = 0.05. The method enabled volumetric analysis of gastric content from 3D breath-hold static acquisition and time-resolved quantification of peristaltic parameters from real-time FLASH2 imaging at high temporal resolution (here 6.24 fps). Water emptied rapidly and exponentially (t1/2 = 14.77 ± 10.55 min), while juice showed slower emptying (t1/2 = 64.24 ± 11.87 min). Contraction frequencies (fasted: 2.76 ± 0.43 cpm, fed: 2.89 ± 0.43 cpm) and velocities (fasted: 1.67 ± 0.38 mm/s, fed: 1.72 ± 0.37 mm/s) were within physiological ranges, with fasting conditions characterized by stronger occlusion compared to the fed. Measurements taken from three patients proved that the workflow could be used in a clinical context. Real-time MRI with AI-based analysis enabled quantitative assessment of gastric emptying and motility, revealing physiological peristaltic parameters and state-dependent differences in occlusion. 2. Stage 1.

Oral Squamous Cell Carcinoma (OSCC) is a prevalent and aggressive malignancy where deep learning-based computer-aided diagnosis and prognosis can enhance clinical assessments. However, existing publicly available OSCC datasets often suffer from limited patient cohorts and a restricted focus on either diagnostic or prognostic tasks, limiting the development of comprehensive and generalizable models. To bridge this gap, we introduce Multi-OSCC, a new histopathology image dataset comprising 1,325 OSCC patients, integrating both diagnostic and prognostic information to expand existing public resources. Each patient is represented by six high resolution histopathology images captured at ×200, ×400, and ×1000-two per magnification-covering both the core and edge tumor regions. The Multi-OSCC dataset is richly annotated for six critical clinical tasks: recurrence prediction (REC), lymph node metastasis (LNM), tumor differentiation (TD), tumor invasion (TI), cancer embolus (CE), and perineural invasion (PI). We systematically evaluate the impact of different visual encoders, multi-image fusion techniques, stain normalization, and multi-task learning frameworks to benchmark this dataset. To accelerate future research, we publicly release the Multi-OSCC dataset at: https://github.com/guanjinquan/OSCC-PathologyImageDataset.

Abstract. On-road transportation is a major contributor to CO2 emissions in cities, and high-resolution CO2 traffic emission maps are essential for analyzing emission patterns and characteristics. In this study, we developed new hourly on-road CO2 emission maps with a 100 × 100 m resolution for 20 major cities in France, Germany, and the Netherlands in 2023. We used commercial Floating Car Data (FCD) based on anonymized GPS signals periodically reported by individual vehicles, providing hourly information on mean speed and the number of GPS sample counts per street. Machine learning models were developed to fill FCD data gaps and convert sample counts into actual traffic volumes, and the COPERT model was used to estimate speed- and vehicle-type-dependent emission factors. These models were calibrated using independent traffic observations available for Paris and Berlin, and subsequently applied to the remaining 18 cities in an extrapolated manner due to data availability constraints. Hourly emissions, initially estimated at the street level, were aggregated to 100 × 100 m grid cells. Annual on-road CO2 emissions across the 20 European cities in 2023 ranged from 0.4 to 7.9 Mt CO2, with emissions strongly correlated with urban area (R2= 0.98) and, to a lesser extent, population size (R2= 0.74). Spatially, emissions are either highly concentrated along major highways in cities such as Paris and Amsterdam or more evenly distributed in cities such as Berlin and Bordeaux, highlighting the need for context-specific mitigation strategies. Temporally, this study shows the CO2 emission fluctuations due to holiday periods, weekly activity cycles, and distinct usage profiles of different vehicle types. Due to the low latency of FCD, this approach could support near-real-time traffic emission mapping in the future. Our approach enhances the spatial and temporal characterization of CO2 emissions in on-road transportation compared to the conventional method used in gridded inventories, indicating the potential of FCD data for near-real-time urban emission monitoring and timely policy-making. The datasets generated by this study are available on Zenodo https://doi.org/10.5281/zenodo.16600210 (Shi et al., 2025).

This paper presents an open-access telemetry dataset designed to support research and training in intelligent fixed-wing unmanned aerial systems. The dataset contains 240 fully annotated autonomous missions flown outdoors over repeatable, waypoint-based trajectories using two onboard architectures: a compact SpeedyBee F405 flight controller running INAV, and a Holybro Pixhawk 6X paired with a Jetson Orin NX companion computer running PX4. The missions cover key phases, including take-off, cruise, dynamic manoeuvres, and autonomous landing. Each log provides synchronised multi-sensor telemetry (IMU, GNSS, barometric altitude, actuator states, flight modes, and power metrics) at high temporal resolution, enabling realistic modelling of flight dynamics, estimator behaviour, and sensor noise. The dataset supports benchmarking for trajectory tracking under degraded GNSS, anomaly detection, wind-aware navigation, and energy-optimised mission planning. The paper documents hardware integration, communication architecture, mission procedures, and the dataset file structure, and includes representative analyses to illustrate reuse for contested, safety-critical, and complex operational environments in field. No neural network is trained or evaluated; deep learning is cited only as a motivating application domain.

Abstract A thorough understanding of soot particle formation and evolution during combustion is essential for improving energy efficiency and reducing pollutant emissions. In this study, we constructed an integrated diagnostic method combinding the two separate pulses with traditional two-color LII and particle size analysis models, enabling simultaneous acquisition of multiple parameters (including initial temperature T0, peak temperature TM, absorption function Em, volume fraction fv, and primary particle size Dp). This system has high spatial and temporal resolution, allowing for detailed characterization of soot dynamics within diffusion flames. The results demonstrate LII can detect particles in the middle and top regions of flame, the parameters of which exhibit significant spatial non-uniformities and complex interdependencies. It is found that the particles get more and more mature during the particles evolving upwards due to carbonization, and therefore show increasing absorption. At the same time, the volume fraction and size of particles decrease due to oxidation. In addition, the volume fraction and size of particles are larger at the edges of flame, which is attributed to the more easily happening nucleation and aggregation processes An overall physical picture about the generation and evolution of the soot particles is finally provided based on the distributions of those multiple parameters. This study validates the effectiveness of SP-2D-2C-LII for multi-parameter coupling analysis and spatially resolved diagnosis, offering valuable insights into soot particles evolution mechanisms and supporting the optimization of combustion and emission control strategies.

Background Mapping surface and ground fuels is key to predicting fire behavior and effects. Remote sensing effectively describes tree canopy attributes, but it is less effective for surface and ground fuels because they are partially obscured by vegetation. Aims We mapped duff loads by leveraging the coupled relationship between canopy, surface and ground fuels: canopy fuels provide the source material for surface fuels, which serve as the material for duff formation. Methods We mapped annual surface fuel production from tree components that contribute to duff formation using airborne laser scanning data. We adjusted a forest soil organic carbon model to simulate duff development under environmental factors constraining decomposition, and included consumption rates accounting for prescribed burning fuel removal. We tested the workflow in southeastern US pine flatwoods where a long-term prescribed burning experiment has been running since 1958. Key results Comparison between modeled duff estimates and field observations provided good agreement, showing low bias and high coefficient of determination. Conclusions This approach allows surface and ground fuels mapping at high spatial resolution (≤5 m) using airborne laser scanning. Implications This integrated modeling workflow improves on current methods to describe heterogeneous duff layers and shows that surface and ground fuels can be mapped using remote sensing and ecological modeling.

Abstract Planetary upper atmospheres couple the deep atmosphere to the space environment. The dynamics and energetics of this rarefied, partially ionized region govern atmospheric evolution. At Jupiter, decades of past plasma measurements have revealed a variable and enigmatic ionosphere inconsistent with photochemical predictions and unusual global structures imprinted by the planet’s powerful magnetic field. Upper-atmospheric temperatures have been measured sporadically and thus are unable to fully characterize the energy transfer mechanism responsible for its unexpectedly hot thermosphere. Observations to date have been too limited spatially, or too insensitive, to uncover the driving mechanisms behind the strong variability and magnetically organized features found in Jupiter’s upper atmosphere. Here, we present high spatial resolution global maps of ion densities and temperatures constructed from >175,000 Keck/NIRSPEC spectra collected over 4 yr. Irregular ionospheric emission features, first seen more than 25 yr ago, are shown to be persistent and due to local ion density modifications. Global temperatures decrease steadily from auroral to equatorial latitudes and are remarkably stable, with equatorial deviations of <10% night to night. Thus, despite appearing stochastic in previous observations, Jupiter’s upper atmosphere exhibits predominantly spatial rather than temporal variability, yielding a steady global structure generated by local plasma dynamics. These results illustrate how neutral winds and magnetic fields can create globally persistent plasma structures on hydrogen-dominated worlds and provide an explanation for decades of puzzling Jovian upper-atmospheric observations.

Abstract Flares are usually observed during the afterglow phases of Gamma-ray bursts (GRBs) in the soft-X-ray, optical, and radio bands—but rarely in the gamma-ray band. Despite its extraordinary brightness, GECAM-C has accurately measured both the bright prompt emission and flare emission of GRB 221009A without instrumental effects, offering a good opportunity to study the relation between them. In this work, we present a comprehensive analysis of the flare emission of GRB 221009A, which is composed of a series of flares. Among them, we identify an exceptionally bright flare with a record-breaking isotropic energy E iso = 1.82 × 10 53 erg for GRB flares. It exhibits the highest peak energy ever detected in GRB flares, E peak ∼ 300 keV, making it a genuine gamma-ray flare. It also shows rapid rise and decay timescales, significantly shorter than those of typical X-ray flares observed in the soft-X-ray or optical bands but comparable to those observed in prompt emissions. Despite these exceptional properties, the flare shares several common properties with typical GRB flares. We note that this is the first observation of a GRB flare in the keV–MeV band with sufficiently high temporal resolution and high statistics, bridging the last gap between the prompt emission and flare.

Abstract Long-term (1998–2024) satellite-derived precipitation and spaceborne precipitation radar data with high spatiotemporal resolution indicate that the region from the eastern East China Sea (ECS) to Kyushu during the Meiyu/Baiu season (June–July) is one of the areas in East Asia where the diurnal precipitation cycle is most apparent. This area experiences higher rainfall intensity and higher rainfall frequency with exceptionally high convective-type rainfall within the Meiyu/Baiu rainband, leading to the highest rainfall totals over East Asia during the season. The maximum (minimum) rainfall in the rainband north of ~ 30°N, extending from the ECS/Kyushu to the Pacific Ocean, occurs during dawn to noon (evening–midnight). The diurnal cycle and convergence of water vapor flux are the main factors driving the diurnal cycle of precipitation. From midnight to dawn, the anomalous southwesterly water vapor flux with clockwise rotation enhances the total southwesterly water vapor flux, which reaches its maximum around the ECS/Kyushu. Subsequently, water vapor flux convergence increases, resulting in the rainfall peak during the morning. We discussed possible mechanisms for the diurnal cycle in water vapor flux around the ECS/Kyushu, focusing on inertial oscillation driven by terrestrial atmospheric boundary layer processes and land–sea thermal contrast.

Occurrence and spatial distribution of emerging organic contaminants in the coastal and deep Red Sea sediments utilizing liquid chromatography tandem high resolution mass spectrometry.

Understanding ocean transport is critical for applications ranging from fisheries to chemical plume tracking and carbon dioxide removal modeling. However, available hydrodynamic data often lack the spatial resolution needed for effective transport simulations. We apply statistical downscaling to coarse-resolution ocean reanalysis and atmospheric wind data, reconstructing fine-scale fields validated against high-resolution dynamic models in the Bering Sea. This enables the prediction of transport patterns without the need to run high resolution physics simulations, saving computational costs and time. We examined five years of high-resolution, statistically downscaled ocean currents and surface winds and found that the correlation of ocean current and wind components with GLORYS and ERA5 reanalysis models were r = 0.87 and r = 0.98. The Liu-mean skill score was 0.75 for ocean current velocity. Okubo-Weiss analyses showed comparable vorticity and shear between downscaled and dynamical models. The Finite-time Layupanov Exponent analysis showed consistent Lagrangian Cohesive Structures across datasets. Multi-year particle tracking using both downscaled and reanalysis forcing showed consistent relative separation distances with mean Bhattacharyya coefficient of 0.720 ± 0.133. The demonstrated parity in dispersal patterns indicates statistically downscaled approaches can substitute dynamical models for large-scale applications. Future work should validate these results across diverse oceanographic regimes and incorporate biogeochemical feedback mechanisms.

A major weakness in the field of neurogastroenterology research has been a lack of technology to determine the spatial and temporal coordination of electrical activity along the gastrointestinal (GI) tract in-vivo, without requiring a surgical procedure. To overcome this weakness, we developed a miniaturized endoscope consisting of 128 iridium oxide recording sensors that allowed us to make high resolution intraluminal electrophysiological recordings in-vivo from the mucosal surface of the terminal large intestine of anesthetized mice. Recordings revealed discharges of smooth muscle action potentials organized into complex spatiotemporal patterns. The patterns were modified by pharmacological agents donepezil and atropine that stimulated or suppressed cholinergic neurotransmission, respectively. The patterns were also ablated by benzalkonium chloride, known to disrupt the function of the enteric nervous system. The endoscope was further validated under ex-vivo recording conditions, where blocking enteric neural activity with tetrodotoxin (TTX) again altered spontaneously occurring action potential patterns. This approach offers a unique opportunity to easily characterize normal and dysfunctional patterns of GI electrical activity in genetically modified and/or diseased mouse models, including drug discovery and high-throughput studies.

Objective.To design and validate a single, reconfigurable gamma probe that overcomes the static compromise between spatial resolution and sensitivity in radio-guided surgery, enabling both rapid lesion detection and precise margin delineation.Approach.A dual-layer lead collimator was designed for a LaBr₃(Ce)-SiPM detector. A validated analytical model coupled with a multi-objective genetic algorithm (NSGA-II) was used to explore the theoretical performance limits and identify optimal geometries. A two-phase computational search identified a single, universal geometry that can be switched intraoperatively between a high-sensitivity (HS) mode and a high-resolution (HR) mode by adjusting collimator positions.Main results.The universal design, at a 30 mm distance, achieves a spatial resolution of 6.41 mm full width at half maximum (FWHM) in HR mode and a sensitivity of 1483 counts per second (cps)/MBq in HS mode. The optimization framework identified specialized, distance-specific theoretical designs with resolutions as fine as 3.26 mm FWHM. The underlying detector's energy resolution is sufficient to distinguish between ⁹⁹mTc (140.5 keV) and123I (159 keV).Significance.This work presents a practical, single-instrument solution that offers surgeons the intraoperative flexibility to prioritize either rapid detection or precise delineation. The developed design methodology provides a robust framework for creating next-generation, application-specific surgical guidance tools.