Abstract
Spectral imaging has the potential to become a key technique in interventional medicine as it unveils much richer optical information compared to conventional RBG (red, green, and blue)-based imaging. Thus allowing for high-resolution functional tissue analysis in real time. Its higher information density particularly shows promise for the development of powerful perfusion monitoring methods for clinical use. However, even though in vivo validation of such methods is crucial for their clinical translation, the biomedical field suffers from a lack of publicly available datasets for this purpose. Closing this gap, we generated the SPECTRAL Perfusion Arm Clamping dAtaset (SPECTRALPACA). It comprises ten spectral videos (∼20 Hz, approx. 20,000 frames each) systematically recorded of the hands of ten healthy human participants in different functional states. We paired each spectral video with concisely tracked regions of interest, and corresponding diffuse reflectance measurements recorded with a spectrometer. Providing the first openly accessible in human spectral video dataset for perfusion monitoring, our work facilitates the development and validation of new functional imaging methods.
Introduction: The Need for Functional Vision
In modern surgical environments, standard RGB (Red, Green, Blue) cameras provide high-spatial-resolution anatomical views but remain "functionally blind" to physiological states like tissue oxygenation. Spectral Imaging (SI) addresses this by capturing light across multiple narrow wavelength bands, revealing the unique spectral signatures of molecules such as hemoglobin. Despite its potential, the clinical translation of SI is hindered by a scarcity of high-quality, publicly available human datasets. Most existing spectral data are either static or derived from satellite imagery, leaving a gap for real-time, video-rate functional monitoring benchmarks.
Methods and Experimental Design
To bridge this gap, we developed the SPECTRALPACA dataset using a snapshot multispectral camera (XIMEA MQ022HG-IM-SM4x4-VIS) capable of recording 16 spectral channels in the visible range at approximately 20 Hz.
Arm-Clamping Protocol
Controlled changes in blood flow were induced through arm-clamping experiments. A pressure cuff on the upper arm was incrementally adjusted from 0 mmHg to 160 mmHg. This gold-standard method predictably alters tissue oxygenation and blood volume in the hand, providing a ground truth for functional state transitions.
Hardware and Data Acquisition
The setup utilized a 0° laparoscope to ensure standardized illumination geometry, connected to a Xenon light source. This configuration mimics clinical environments while maintaining consistent lighting for spectral accuracy.
Data Processing
Four Regions of Interest (ROIs) on the palm were manually annotated and tracked across 20,000 frames using a Discriminative Correlation Filter (CSR-DCF). A critical preprocessing step involved $\ell_1$-normalization of the spectral data to eliminate multiplicative noise from varying camera distances or integration times, focusing the analysis on spectral shape rather than intensity.
The SPECTRALPACA Dataset
The resulting dataset consists of 10 spectral videos (one per subject), each approximately 18 minutes long. To ensure laboratory-grade validation, the camera data is paired with diffuse reflectance measurements from a high-precision spectrometer. The dataset is publicly hosted on Synapse (SynID: syn51625685) under a Creative Commons license.
Key Insights and Results
Functional Sensitivity and Accuracy
Validation using a Random Forest regression model confirmed that the dataset successfully captures oxygenation dips across all pressure levels above 20 mmHg. Furthermore, the spectral camera demonstrated high agreement with the spectrometer, with Euclidean distance deviations as low as 0.0007 for reference tiles.
The Role of Personalization
Principal Component Analysis (PCA) revealed that data points clustered more strongly by individual subject than by pressure state. This finding highlights a significant challenge for "one-size-fits-all" algorithms and suggests that future functional imaging tools may benefit from personalized baseline models.
Conclusion and Future Directions
SPECTRALPACA provides a massive leap in scale for annotated human spectral data, offering roughly 200,000 total frames for algorithm development. While currently limited to 10 healthy subjects, it serves as a foundational "playground" for validating synthetic tissue models and developing the next generation of real-time surgical assistants. Future work will focus on expanding cohort diversity and transitioning these methods to clinical pathologies.
Summary updated