Abstract:Hyperspectral (HS) imaging presents itself as a non-contact, non-ionizing and non-invasive technique, proven to be suitable for medical diagnosis. However, the volume of information contained in these images makes difficult providing the surgeon with information about the boundaries in real-time. To that end, High-Performance-Computing (HPC) platforms become necessary. This paper presents a comparison between the performances provided by five different HPC platforms while processing a spatial-spectral approach to classify HS images, assessing their main benefits and drawbacks. To provide a complete study, two different medical applications, with two different requirements, have been analyzed. The first application consists of HS images taken from neurosurgical operations; the second one presents HS images taken from dermatological interventions. While the main constraint for neurosurgical applications is the processing time, in other environments, as the dermatological one, other requirements can be considered. In that sense, energy efficiency is becoming a major challenge, since this kind of applications are usually developed as hand-held devices, thus depending on the battery capacity. These requirements have been considered to choose the target platforms: on the one hand, three of the most powerful Graphic Processing Units (GPUs) available in the market; and, on the other hand, a low-power GPU and a manycore architecture, both specifically thought for being used in battery-dependent environments.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to optimize the real - time processing capacity in medical diagnosis by evaluating the performance of five different high - performance computing (HPC) platforms when processing the spatial - spectral classification of hyperspectral images. Specifically, the paper focuses on two different medical application scenarios: 1. **Neurosurgical application**: Use hyperspectral imaging (HSI) technology to identify and delineate living brain tumors in real - time during surgery to assist neurosurgeons in tumor resection. In this application, the main goal is to obtain classification results within the shortest computing time, providing near - real - time results without being affected by the power limitations in the operating room. 2. **Dermatological application**: Use a handheld hyperspectral acquisition and processing system for the early detection of skin cancer. This application scenario is not intended to be used during the surgical process, so the requirement for real - time performance is relatively loose. However, since this is an independent and portable system that needs to work without an external power connection, there is a relatively high requirement for energy efficiency. To achieve these goals, the paper proposes a spatial - spectral (SS) classification framework that combines the spectral values and spatial coordinates of pixels to improve classification accuracy. This framework consists of three main algorithms: - **Support vector machine (SVM) classifier**: Used to classify based on the spectral information of each pixel and calculate the probability that each pixel belongs to each category. - **Principal component analysis (PCA) dimensionality - reduction algorithm**: Used to reduce the redundancy between spectral bands and extract the main features of the image. - **K - nearest neighbor (KNN) filtering algorithm**: Used to refine the pixel - level classification probability map generated by the SVM by matching and averaging non - local neighborhoods. The paper finds the optimal balance between performance and energy consumption by implementing these three algorithms on different HPC platforms and evaluating their performance and power consumption. These platforms include three high - performance GPUs, one low - power GPU, and one low - power multi - core architecture. Through this evaluation, the paper aims to provide guidance for selecting appropriate computing platforms for future medical applications.

Parallel Implementations Assessment of a Spatial-Spectral Classifier for Hyperspectral Clinical Applications

Acceleration of Hyperspectral Skin Cancer Image Classification through Parallel Machine-Learning Methods

Parallel Implementation Of Hyperspectral Image Processing Algorithms

A New GPU Implementation of Support Vector Machines for Fast Hyperspectral Image Classification

Parallel Spatial–Spectral Hyperspectral Image Classification with Sparse Representation and Markov Random Fields on GPUs

Parallel Computation for Blood Cell Classification in Medical Hyperspectral Imagery

High Performance Hyperspectral Image Classification using Graphics Processing Units

High Performance Computing for Hyperspectral Remote Sensing

GPU Parallel Implementation of Spatially Adaptive Hyperspectral Image Classification

Parallel Morphological/neural Processing of Hyperspectral Images Using Heterogeneous and Homogeneous Platforms.

An Experimental Comparison of Parallel Algorithms for Hyperspectral Analysis Using Heterogeneous and Homogeneous Networks of Workstations

Parallel Implementation of a Hyperspectral Unmixing Chain: Graphic Processing Units Versus Multi-Core Processors

Performance Portability Study of an Automatic Target Detection and Classification Algorithm for Hyperspectral Image Analysis Using Opencl

Real-Time Identification of Hyperspectral Subspaces

Non-Invasive Skin Cancer Diagnosis Using Hyperspectral Imaging for In-Situ Clinical Support

Parallel Implementation of Sparse Representation Classifiers for Hyperspectral Imagery on Gpus

Multi-GPU Implementation of Nearest-Regularized Subspace Classifier for Hyperspectral Image Classification

Performance Versus Energy Consumption of Hyperspectral Unmixing Algorithms on Multi-Core Platforms

Supervised Machine Learning Methods and Hyperspectral Imaging Techniques Jointly Applied for Brain Cancer Classification

Edge and cloud computing approaches in the early diagnosis of skin cancer with attention-based vision transformer through hyperspectral imaging

Parallel Implementation of Spatial–Spectral Endmember Extraction on Graphic Processing Units