In recent years, researchers designed several artificial intelligence solutions for healthcare applications, which usually evolved into functional solutions for clinical practice. Furthermore, deep learning (DL) methods are well-suited to process the broad amounts of data acquired by wearable devices, smartphones, and other sensors employed in different medical domains. Conceived to serve the role of diagnostic tool and surgical guidance, hyperspectral images emerged as a non-contact, non-ionizing, and label-free technology. However, the lack of large datasets to efficiently train the models limits DL applications in the medical field. Hence, its usage with hyperspectral images is still at an early stage. We propose a deep convolutional generative adversarial network to generate synthetic hyperspectral images of epidermal lesions, targeting skin cancer diagnosis, and overcome small-sized datasets challenges to train DL architectures. Experimental results show the effectiveness of the proposed framework, capable of generating synthetic data to train DL classifiers.

Currently, one of the most common causes of death worldwide is cancer. The development of innovative methods to support the early and accurate detection of cancers is required to increase the recovery rate of patients. Several studies have shown that medical Hyperspectral Imaging (HSI) combined with artificial intelligence algorithms is a powerful tool for cancer detection. Various preprocessing methods are commonly applied to hyperspectral data to improve the performance of the algorithms. However, there is currently no standard for these methods, and no studies have compared them so far in the medical field. In this work, we evaluated different combinations of preprocessing steps, including spatial and spectral smoothing, Min-Max scaling, Standard Normal Variate normalization, and a median spatial smoothing technique, with the goal of improving tumor detection in three different HSI databases concerning colorectal, esophagogastric, and brain cancers. Two machine learning and deep learning models were used to perform the pixel-wise classification. The results showed that the choice of preprocessing method affects the performance of tumor identification. The method that showed slightly better results with respect to identifing colorectal tumors was Median Filter preprocessing (0.94 of area under the curve). On the other hand, esophagogastric and brain tumors were more accurately identified using Min-Max scaling preprocessing (0.93 and 0.92 of area under the curve, respectively). However, it is observed that the Median Filter method smooths sharp spectral features, resulting in high variability in the classification performance. Therefore, based on these results, obtained with different databases acquired by different HSI instrumentation, the most relevant preprocessing technique identified in this work is Min-Max scaling.

The early detection of skin cancer is of crucial importance to plan an effective therapy to treat the lesion. In routine medical practice, the diagnosis is based on the visual inspection of the lesion and it relies on the dermatologists’ expertise. After a first examination, the dermatologist may require a biopsy to confirm if the lesion is malignant or not. This methodology suffers from false positives and negatives issues, leading to unnecessary surgical procedures. Hyperspectral imaging is gaining relevance in this medical field since it is a non-invasive and non-ionizing technique, capable of providing higher accuracy than traditional imaging methods. Therefore, the development of an automatic classification system based on hyperspectral images could improve the medical practice to distinguish pigmented skin lesions from malignant, benign, and atypical lesions. Additionally, the system can assist general practitioners in first aid care to prevent noncritical lesions from reaching dermatologists, thereby alleviating the workload of medical specialists. In this paper is presented a parallel pipeline for skin cancer detection that exploits hyperspectral imaging. The computational times of the serial processing have been reduced by adopting multicore and many-core technologies, such as OpenMP and CUDA paradigms. Different parallel approaches have been combined, leading to the development of fifteen classification pipeline versions. Experimental results using in-vivo hyperspectral images show that a hybrid parallel approach is capable of classifying an image of 50 × 50 pixels with 125 bands in less than 1 s.

Cancer originates from the uncontrolled growth of healthy cells into a mass. Chromophores, such as hemoglobin and melanin, characterize skin spectral properties, allowing the classification of lesions into different etiologies. Hyperspectral imaging systems gather skin-reflected and transmitted light into several wavelength ranges of the electromagnetic spectrum, enabling potential skin-lesion differentiation through machine learning algorithms. Challenged by data availability and tiny inter and intra-tumoral variability, here we introduce a pipeline based on deep neural networks to diagnose hyperspectral skin cancer images, targeting a handheld device equipped with a low-power graphical processing unit for routine clinical testing. Enhanced by data augmentation, transfer learning, and hyperparameter tuning, the proposed architectures aim to meet and improve the well-known dermatologist-level detection performances concerning both benign-malignant and multiclass classification tasks, being able to diagnose hyperspectral data considering real-time constraints. Experiments show 87% sensitivity and 88% specificity for benign-malignant classification and specificity above 80% for the multiclass scenario. AUC measurements suggest classification performance improvement above 90% with adequate thresholding. Concerning binary segmentation, we measured skin DICE and IOU higher than 90%. We estimated 1.21 s, at most, consuming 5 Watts to segment the epidermal lesions with the U-Net++ architecture, meeting the imposed time limit. Hence, we can diagnose hyperspectral epidermal data assuming real-time constraints.

Background: COVID-19 is an emerging infectious disease, that is heavily challenging health systems worldwide. Admission Arterial Blood Gas (ABG) and Lung Ultrasound (LUS) can be of great help in clinical decision making, especially during the current pandemic and the consequent overcrowding of the Emergency Department (ED). The aim of the study was to demonstrate the capability of alveolar-to-arterial oxygen difference (AaDO₂) in predicting the need for subsequent oxygen support and survival in patients with COVID-19 infection, especially in the presence of baseline normal PaO₂/FiO₂ ratio (P/F) values. Methods: A cohort of 223 swab-confirmed COVID-19 patients underwent clinical evaluation, blood tests, ABG and LUS in the ED. LUS score was derived from 12 ultrasound lung windows. AaDO₂ was derived as AaDO₂ = ((FiO₂) (Atmospheric pressure − H₂O pressure) − (PaCO₂/R)) − PaO₂. Endpoints were subsequent oxygen support need and survival. Results: A close relationship between AaDO₂ and P/F and between AaDO₂ and LUS score was observed (R² = 0.88 and R² = 0.67, respectively; p < 0.001 for both). In the subgroup of patients with P/F between 300 and 400, 94.7% (n = 107) had high AaDO₂ values, and 51.4% (n = 55) received oxygen support, with 2 ICU admissions and 10 deaths. According to ROC analysis, AaDO₂ > 39.4 had 83.6% sensitivity and 90.5% specificity (AUC 0.936; p < 0.001) in predicting subsequent oxygen support, whereas a LUS score > 6 showed 89.7% sensitivity and 75.0% specificity (AUC 0.896; p < 0.001). Kaplan–Meier curves showed different mortality in the AaDO₂ subgroups (p = 0.0025). Conclusions: LUS and AaDO₂ are easy and effective tools, which allow bedside risk stratification in patients with COVID-19, especially when P/F values, signs, and symptoms are not indicative of severe lung dysfunction.