Deep Learning Classification Algorithms Research Articles

Classification of Parkinson's disease by deep learning on midbrain MRI.

Susceptibility map weighted imaging (SMWI), based on quantitative susceptibility mapping (QSM), allows accurate nigrosome-1 (N1) evaluation and has been used to develop Parkinson's disease (PD) deep learning (DL) classification algorithms. Neuromelanin-sensitive (NMS) MRI could improve automated quantitative N1 analysis by revealing neuromelanin content. This study aimed to compare classification performance of four approaches to PD diagnosis: (1) N1 quantitative "QSM-NMS" composite marker, (2) DL model for N1 morphological abnormality using SMWI ("Heuron IPD"), (3) DL model for N1 volume using SMWI ("Heuron NI"), and (4) N1 SMWI neuroradiological evaluation. PD patients (n = 82; aged 65 ± 9 years; 68% male) and healthy-controls (n = 107; 66 ± 7 years; 48% male) underwent 3 T midbrain MRI with T2*-SWI multi-echo-GRE (for QSM and SMWI), and NMS-MRI. AUC was used to compare diagnostic performance. We tested for correlation of each imaging measure with clinical parameters (severity, duration and levodopa dosing) by Spearman-Rho or Kendall-Tao-Beta correlation. Classification performance was excellent for the QSM-NMS composite marker (AUC = 0.94), N1 SMWI abnormality (AUC = 0.92), N1 SMWI volume (AUC = 0.90), and neuroradiologist (AUC = 0.98). Reasons for misclassification were right-left asymmetry, through-plane re-slicing, pulsation artefacts, and thin N1. In the two DL models, all 18/189 (9.5%) cases misclassified by Heuron IPD were controls with normal N1 volumes. We found significant correlation of the SN QSM-NMS composite measure with levodopa dosing (rho = -0.303, p = 0.006). Our data demonstrate excellent performance of a quantitative QSM-NMS marker and automated DL PD classification algorithms based on midbrain MRI, while suggesting potential further improvements. Clinical utility is supported but requires validation in earlier stage PD cohorts.

An Analysis of Traditional Methods and Deep Learning Methods in SSVEP-Based BCI: A Survey

The brain–computer interface (BCI) is a direct communication channel between humans and machines that relies on the central nervous system. Neuroelectric signals are collected by placing electrodes, and after feature sampling and classification, they are converted into control signals to control external mechanical devices. BCIs based on steady-state visual evoked potential (SSVEP) have the advantages of high classification accuracy, fast information conduction rate, and relatively strong anti-interference ability, so they have been widely noticed and discussed. From k-nearest neighbor (KNN), multilayer perceptron (MLP), and support vector machine (SVM) classification algorithms to the current deep learning classification algorithms based on neural networks, a wide variety of discussions and analyses have been conducted by numerous researchers. This article summarizes more than 60 SSVEP- and BCI-related articles published between 2015 and 2023, and provides an in-depth research and analysis of SSVEP-BCI. The survey in this article can save a lot of time for scholars in understanding the progress of SSVEP-BCI research and deep learning, and it is an important guide for designing and selecting SSVEP-BCI classification algorithms.

A comparative study of deep learning models and classification algorithms for chemical compound identification and Tox21 prediction

Chemical compound classification, toxicity prediction, and environmental risk assessments are critically important in various applications within the field of chemistry. Deep learning models provide highly effective tools for extracting features from complex large datasets and performing classification tasks. Four different deep learning models, namely ResNet50V2, VGG19, InceptionV3, and MobileNetV2, have been compared with the random forest (RF) and k-nearest neighbors (KNN) algorithms. The results obtained from experiments conducted using QRCODE images of the Tox21SMILES dataset demonstrate the effectiveness of deep learning models for classifying chemical compounds and showcase the performance of different classification algorithms. The findings of the study thoroughly evaluate the performance of deep learning models and classification algorithms in the task of chemical classification. While ResNet50V2 and VGG19 models achieve high accuracy and precision, InceptionV3 and MobileNetV2 models provide more balanced results. Additionally, in terms of classification algorithms, the k-nearest neighbors (KNN) algorithm generally outperforms the Random Forest (RF) algorithm. Although the RF algorithm achieves good accuracy, the KNN algorithm proves to be more effective in terms of sensitivity and F1 score. These results emphasize the factors to consider when choosing which deep learning model or classification algorithm to use in chemical classification tasks. In conclusion, this study presents a comprehensive analysis comparing the performance of deep learning models and classification algorithms in chemical classification tasks. The selection of the most suitable model and algorithm for a specific task supports achieving better results in the classification of chemical compounds and related applications.

Deep Learning Classification Algorithms Applications: A Review

This paper examines the recent articles on classification tasks, particularly focusing on deep learning Algorithms. The process of categorizing data into distinct classes based on specific features is essential for tasks such as image recognition, sentiment analysis, disease diagnosis, and more. This article the fundamental concepts of deep learning, including neural network architectures like Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and their variants. It explores the significance of feature selection techniques in improving classification model performance. Additionally, this article provides a detailed literature review, aiming to foster the development of more effective and efficient classification algorithms and methodologies and highlighting their applications in fields such as healthcare, agriculture, disaster response, and beyond. Through this review, this article underscores the transformative impact of deep learning approaches in enabling automated decision-making, pattern recognition, and data-driven insights, offering valuable insights for researchers, practitioners, and policymakers involved in classification tasks, this article aims to facilitate the development of more effective and efficient classification algorithms and methodologies.

Deep learning of mammogram images to reduce unnecessary breast biopsies: a preliminary study

BackgroundPatients with a Breast Imaging Reporting and Data System (BI-RADS) 4 mammogram are currently recommended for biopsy. However, 70–80% of the biopsies are negative/benign. In this study, we developed a deep learning classification algorithm on mammogram images to classify BI-RADS 4 suspicious lesions aiming to reduce unnecessary breast biopsies.Materials and methodsThis retrospective study included 847 patients with a BI-RADS 4 breast lesion that underwent biopsy at a single institution and included 200 invasive breast cancers, 200 ductal carcinoma in-situ (DCIS), 198 pure atypias, 194 benign, and 55 atypias upstaged to malignancy after excisional biopsy. We employed convolutional neural networks to perform 4 binary classification tasks: (I) benign vs. all atypia + invasive + DCIS, aiming to identify the benign cases for whom biopsy may be avoided; (II) benign + pure atypia vs. atypia-upstaged + invasive + DCIS, aiming to reduce excision of atypia that is not upgraded to cancer at surgery; (III) benign vs. each of the other 3 classes individually (atypia, DCIS, invasive), aiming for a precise diagnosis; and (IV) pure atypia vs. atypia-upstaged, aiming to reduce unnecessary excisional biopsies on atypia patients.ResultsA 95% sensitivity for the “higher stage disease” class was ensured for all tasks. The specificity value was 33% in Task I, and 25% in Task II, respectively. In Task III, the respective specificity value was 30% (vs. atypia), 30% (vs. DCIS), and 46% (vs. invasive tumor). In Task IV, the specificity was 35%. The AUC values for the 4 tasks were 0.72, 0.67, 0.70/0.73/0.72, and 0.67, respectively.ConclusionDeep learning of digital mammograms containing BI-RADS 4 findings can identify lesions that may not need breast biopsy, leading to potential reduction of unnecessary procedures and the attendant costs and stress.

Identification of Paddy Stages from Images using Deep Learning

Rice, a crucial global staple, is integral to food security. Precise identification of paddy growth stages, booting, heading, anthesis, grain filling, and grain maturity is vital for agricultural decisions. However, a gap exists in recognizing these stages using red-green-blue (RGB) images. This study uses state-of-the-art computer vision and deep learning classification (Convolutional Neural Networks) algorithms to address this gap. Among the studied algorithms, EfficientNet_B0 achieved an impressive 82.8% overall accuracy. Notably, increasing image size from 64X64 pixels to 128X128 pixels significantly enhanced accuracy. A detailed assessment of growth stages revealed varying accuracy levels, with boot leaf being the most accurately detected (95.1%) and anthesis being the most challenging (72.28%). This work significantly advances automated monitoring, empowering researchers in real-time decision-making.

Remote Disease Diagnosis through IoMT-Enhanced Blood Cell Classification with Deep Learning

Background For the purpose of diagnosing diseases and developing treatment plans, blood cell pictures must be accurately classified. This procedure can be greatly enhanced by automated systems that make use of deep learning and the Internet of Medical Things (IoMT). Objective In order to improve illness detection and increase healthcare accessibility, this work suggests an IoMT-based system for remote blood cell picture transmission and classification utilizing deep learning algorithms. Methods High-resolution pictures of blood cells are captured by an IoMT tiny camera and wirelessly sent to a cloud-based infrastructure. The blood cells are divided into groups according to a, deeplearning classification algorithm, including neutrophils, lymphocytes, monocytes, and eosinophils. Results The IoMT-enabled system excels in transmitting and analyzing blood cell images, achieving precise classification. Utilizing deep learning models with multi-scale feature extraction and attention mechanisms, the system demonstrates robust performance. Numerical results showcase a high accuracy of approximately 97.21%, along with noteworthy precision, recall, and F1 scores for individual blood cell classes. Eosinophil, Lymphocyte, Monocyte, and Neutrophil classes exhibit strong performance metrics, emphasizing the system's effectiveness in accurate blood cell classification. Conclusion By combining IoMT and deep learning with blood cell image analysis, diagnostic accessibility and efficiency are improved. The suggested approach has the potential to completely transform healthcare by facilitating prompt interventions, individualized treatment regimens, and better patient outcomes. It is essential to continuously enhance and validate the system in order to maximize its efficacy and dependability in a variety of healthcare settings.

Multiclass Sentiment Prediction of Airport Service Online Reviews Using Aspect-Based Sentimental Analysis and Machine Learning

Airport service quality ratings found on social media such as Airline Quality and Google Maps offer invaluable insights for airport management to improve their quality of services. However, there is currently a lack of research analysing these reviews by airport services using sentimental analysis approaches. This research applies multiclass models based on Aspect-Based Sentimental Analysis to conduct a comprehensive analysis of travellers’ reviews, in which the major airport services are tagged by positive, negative, and non-existent sentiments. Seven airport services commonly utilised in previous studies are also introduced. Subsequently, various Deep Learning architectures and Machine Learning classification algorithms are developed, tested, and compared using data collected from Twitter, Google Maps, and Airline Quality, encompassing travellers’ feedback on airport service quality. The results show that the traditional Machine Learning algorithms such as the Random Forest algorithm outperform Deep Learning models in the multiclass prediction of airport service quality using travellers’ feedback. The findings of this study offer concrete justifications for utilising multiclass Machine Learning models to understand the travellers’ sentiments and therefore identify airport services required for improvement.

Droplet-based logic gates simulation of viscoelastic fluids under electric field

Nano and microfluidic technologies have shown great promise in the development of controlled drug delivery systems and the creation of microfluidic devices with logic-like functionalities. Here, we focused on investigating a droplet-based logic gate that can be used for automating medical diagnostic assays. This logic gate uses viscoelastic fluids, which are particularly relevant since bio-fluids exhibit viscoelastic properties. The operation of the logic gate is determined by evaluating various parameters, including the Weissenberg number, the Capillary number, and geometric factors. To effectively classify the logic gates operational conditions, we employed a deep learning classification to develop a reduced-order model. This approach accelerates the prediction of operating conditions, eliminating the need for complex simulations. Moreover, the deep learning model allows for the combination of different AND/OR branches, further enhancing the versatility of the logic gate. We also found that non-operating regions, where the logic gate does not function properly, can be transformed into operational regions by applying an external force. By utilizing an electrical induction technique, we demonstrated that the application of an electric field can repel or attract droplets, thereby improving the performance of the logic gate. Overall, our research shows the potential of the droplet-based logic gates in the field of medical diagnostics. The integration of deep learning classification algorithms enables rapid evaluation of operational conditions and facilitates the design of complex logic circuits. Additionally, the introduction of external forces and electrical induction techniques opens up new possibilities for enhancing the functionality and reliability of these logic gates.

Durian Species Classification Using Deep Learning Method

Durian is a popular fruit in Southeast Asia, and the market offers various species of durians. Accurate species classification is crucial for quality control, grading, and marketing. However, the complexity of this task has led to the utilization of machine learning and deep learning methods. Traditional machine learning algorithms, such as K-Nearest Neighbors, Linear Discriminant Analysis, Support Vector Machines, and Random Forests, have demonstrated good accuracy, but they require extensive feature engineering. Deep learning algorithms, particularly Convolutional Neural Networks, can automatically extract features, making them less dependent on manual feature selection. This research aims to review deep learning classification algorithms, including Convolutional Neural Networks and Recurrent Neural Networks, to determine the most suitable algorithm for an efficient and accurate durian classification system. The objective is to enhance the precision and speed of durian species classification, presenting potential advantages for both durian producers and consumers. The literature review revealed that Convolutional Neural Networks outperformed other deep learning and traditional machine learning algorithms on datasets of varying sizes, achieving the highest accuracy of 98.96% through techniques like image resizing, color conversion, and additional parameters such as days harvested and dry weight. Deep learning emerges as a promising approach for robust and accurate durian species recognition, with future directions including developing models to classify durian species from different plant parts and even real-time video analysis. However, while Convolutional Neural Networks lead the way, a critical research gap exists in identifying optimal features, necessitating further investigation to refine durian species recognition accuracy.

Remote sensing Image processing, DEEP learning Classification algorithms survey and review

Remote sensing Image processing, DEEP learning Classification algorithms survey and review

Selection of an algorithm for classifying data quoted on the Day Ahead Market of TGE S.A. in MATLAB and Simulink using Deep Learning Toolbox

The article contains an analysis leading to the selection of an algorithm for classifying data listed on the Day-Ahead Market of TGE S.A. in MATLAB and Simulink using Deep Learning Toolbox. In this regard, an introduction to deep learning methods, classification methods, and classification algorithms is provided first. Particular attention was paid to the essence of three important deep learningmethods in the classification, i.e. the methods called: Stochastic Gradient Descent Momentum, Root Mean Square Prop and Adaptive Moment Estimation. Then, three architectures of artificial neural networks used in deep learning were characterized, i.e.: Deep Belief Network, Convolutional Neural Network and Recurrent Neural Network. Attention was paid to the selection parameters of algorithms for learning deep artificial neural networks that can be used in classification, such as: accuracy, information losses and learning time. Practical aspects of research experiments were also shown, including selected results of research conducted on volume and fixing 1 data quoted on the TGE S.A. Day-Ahead Market. After analyzing the obtained test results for the hourly system, it was noted that the least suitable algorithm for classification purposes was the Stochastic Gradient Descent Momentum algorithm, which in each case had worse results than the other two algorithms, i.e. the Adaptive Moment Estimation algorithm and the Root Mean algorithm Square Prop. However, the best algorithm turned out to be the Adaptive Moment Estimation algorithm, which obtained the highest accuracy, which was at a level comparable to the Root Mean Square Prop algorithm, with the latter algorithm having larger losses.

Advances in Quantum Machine Learning and Deep Learning for Image Classification: A Survey

Image classification, which is a fundamental element of Computer Vision (CV) and Artificial Intelligence (AI), has been researched intensively in numerous domains and embedded in many products. However, with the exponential increase in the number of images and the complexity of the required tasks, deep-learning classification algorithms demand more intensive resources and computational power to train the models and update the massive number of parameters. Quantum computing is a new research technology with a promising capability of exponential speed up and operational parallelization with its unique phenomena including superposition and entanglement. Researchers have already started utilizing Quantum Deep Learning (QDL) and Quantum Machine Learning (QML) in image classification. Yet, to our knowledge, there exists no comprehensive published literature review on quantum image classification. Therefore, this paper analyzes the advances in this field by dividing the studies based on a unique taxonomy, discussing the limitations, summarizing essential aspects of each research, and finally, emphasizing the gaps, challenges, and recommendations. One of the key challenges presented in the paper is that quantum computers are in the Noisy Intermediate-Scale Quantum (NISQ) era, where they contain a limited number of noisy qubits, therefore challenging complex quantum classifiers and complex images from advanced datasets. This research recommends constructing a novel quantum image encoding method that adapts to the available number of qubits and enables RGB images as a critical contribution to the existing research.

Disease detection in citrus crops using optical and thermal remote sensing: a literature review

Brazil stands out in the international citrus trade, especially due to its oranges, having produced around 16 million tons in 2021. However, productivity could be increased with greater control of diseases such as greening, which has spread around the world and leads to the total loss of affected trees. Given this scenario, it is necessary to perform fast and accurate detections in order to better manage actions and inputs. Since remote sensing is a pillar of digital agriculture, a literature review was carried out to analyze the use of optical and thermal sensors for the detection of diseases that affect citrus groves. For this purpose, the international databases Scopus and Web of Science were used to select references published between 2012 and 2022, resulting in twelve studies — most from China or the United States of America. The results showed a prevalence of methodologies that combine bands and spectral indices obtained through the use of multispectral and hyperspectral sensors, predominantly on board unmanned aircrafts (UAVs). Machine learning (ML) and deep learning (DL) classification algorithms produced good results in the detection of citrus groves affected by diseases, mainly greening. These results are affected by the stage of the infection, the presence or absence of symptoms, and the spectral and spatial resolutions of the sensors: the Red-Edge band and data with higher spatial detail result in more accurate classification models. However, the analyzed literature is still inconclusive regarding the early detection of infected plants.

Contribution of Geometric Feature Analysis for Deep Learning Classification Algorithms of Urban LiDAR Data.

The use of a Machine Learning (ML) classification algorithm to classify airborne urban Light Detection And Ranging (LiDAR) point clouds into main classes such as buildings, terrain, and vegetation has been widely accepted. This paper assesses two strategies to enhance the effectiveness of the Deep Learning (DL) classification algorithm. Two ML classification approaches are developed and compared in this context. These approaches utilize the DL Pipeline Network (DLPN), which is tailored to minimize classification errors and maximize accuracy. The geometric features calculated from a point and its neighborhood are analyzed to select the features that will be used in the input layer of the classification algorithm. To evaluate the contribution of the proposed approach, five point-clouds datasets with different urban typologies and ground topography are employed. These point clouds exhibit variations in point density, accuracy, and the type of aircraft used (drone and plane). This diversity in the tested point clouds enables the assessment of the algorithm's efficiency. The obtained high classification accuracy between 89% and 98% confirms the efficacy of the developed algorithm. Finally, the results of the adopted algorithm are compared with both rule-based and ML algorithms, providing insights into the positioning of DL classification algorithms among other strategies suggested in the literature.

Comparison of Multiple Machine Learning Algorithms for Music Genre Classification

With the fast advance of the Internet and the continuous improvement of computer technology, speech recognition has been applied in many fields, and speech recognition has broad prospects for development. Music and audio classification technology can add category labels to music based on music content, which is of great significance in the research and application of efficient organization, retrieval and recommendation of music resources. In order to efficiently classify audio from massive online music data and help users to obtain the most suitable music style, a deep learning classification algorithm based on convolutional neural network (CNN) is proposed. To examine its effectiveness, it is compared with traditional machine learning algorithm. First, the original music data set was preprocessed and then feature extraction was carried out to obtain music features and transform them into spectral maps. Traditional machine learning model and deep learning component model were used for simulation experiments. The testing accuracy of the deep learning model is up to 92%, verifying the model's superiority.

Intelligence system for sentiment classification with deep topic embedding using N-gram based topic modeling

Multi-modal information outbreak is consistently increasing in social media. Classification of tweet sentiments using various information modalities will help the recommender systems to achieve success in digital marketing. Moreover, aspect-level sentiment analysis categorizes a target’s sentiment polarity in a specific environment. Using topic modelling in aspect-level sentiment analysis enables the identification of more accurate aspect-based tweet sentiments. The existing sentiment classification techniques used for the development of recommendation systems do not focus on the aspect-based approach modelled using deep learning classifier with temporal analysis on the social media data. Hence, this paper proposes an efficient sentiment classification model that highlights the impact of topic modelling-based word feature embedding for improvising the classification of Twitter sentiments and product reviews based on temporal reasoning and analysis for performing predictive analysis. For tweets context analysis, Latent Dirichlet Allocation based topic modelling is used in this work which generates the topics. For each topic, the sentiment is calculated separately and the topic guided feature expansion is done using Senti-wordnet. Moreover, an extended deep learning classification algorithm called Long Short-Term Memory (LSTM) with word embedding and temporal reasoning(LSTMWTR) is proposed in this paper for improving the classification accuracy. Finally, the labelled data are classified using the existing machine learning algorithms namely Naïve Bayes, Support Vector Machines and also using the deep learning models such as Convolution Neural Network(CNN),LSTM, Recurrent Neural Networks (RNN) and the transformer model namelyBi-directional Encoder Representation from Transformers (BERT),Convolution Bi-directional Recurrent Neural Network (CBRNN) and the proposed deep learning algorithm namelyLSTMWTR. These sentiment classification algorithms have been evaluated with word embedding for tweet sentiment classification and product review classification. The results obtained from this work show that the proposed LSTMWTR algorithm emerges as the highly accurate model for tweet sentiment and product review classification.

Plasma Exosome Analysis for Protein Mutation Identification Using a Combination of Raman Spectroscopy and Deep Learning.

Protein mutation detection using liquid biopsy can be simply performed periodically, making it easy to detect the occurrence of newly emerging mutations rapidly. However, it has low diagnostic accuracy since there are more normal proteins than mutated proteins in body fluids. To increase the diagnostic accuracy, we analyzed plasma exosomes using nanoplasmonic spectra and deep learning. Exosomes, a promising biomarker, are abundant in plasma and stably carry intact proteins originating from mother cells. However, the mutated exosomal proteins cannot be detected sensitively because of the subtle changes in their structure. Therefore, we obtained Raman spectra that provide molecular information about structural changes in mutated proteins. To extract the unique features of the protein from complex Raman spectra, we developed a deep-learning classification algorithm with two deep-learning models. Consequently, controls with wild-type proteins and patients with mutated proteins were classified with high accuracy. As a proof of concept, we discriminated the lung cancer patients with mutations in the epidermal growth factor receptor (EGFR), L858R, E19del, L858R + T790M, and E19del + T790M, from controls with an accuracy of 0.93. Moreover, the protein mutation status of the patients with primary (E19del, L858R) and secondary (+T790M) mutations was clearly monitored. Overall, our technique is expected to be applied as a novel method for companion diagnostic and treatment monitoring.

Open‐source workflow approaches to passive acoustic monitoring of bats

AbstractThe affordability, storage and power capacity of compact modern recording hardware have evolved passive acoustic monitoring (PAM) of animals and soundscapes into a non‐invasive, cost‐effective tool for research and ecological management particularly useful for bats and toothed whales that orient and forage using ultrasonic echolocation. The use of PAM at large scales hinges on effective automated detectors and species classifiers which, combined with distance sampling approaches, have enabled species abundance estimation of toothed whales. But standardized, user‐friendly and open access automated detection and classification workflows are in demand for this key conservation metric to be realized for bats.We used the PAMGuard toolbox including its new deep learning classification module to test the performance of four open‐source workflows for automated analyses of acoustic datasets from bats. Each workflow used a different initial detection algorithm followed by the same deep learning classification algorithm and was evaluated against the performance of an expert manual analyst.Workflow performance depended strongly on the signal‐to‐noise ratio and detection algorithm used: the full deep learning workflow had the best classification accuracy (≤67%) but was computationally too slow for practical large‐scale bat PAM. Workflows using PAMGuard's detection module or triggers onboard an SM4BAT or AudioMoth accurately classified up to 47%, 59% and 34%, respectively, of calls to species. Not all workflows included noise sampling critical to estimating changes in detection probability over time, a vital parameter for abundance estimation. The workflow using PAMGuard's detection module was 40 times faster than the full deep learning workflow and missed as few calls (recall for both ~0.6), thus balancing computational speed and performance.We show that complete acoustic detection and classification workflows for bat PAM data can be efficiently automated using open‐source software such as PAMGuard and exemplify how detection choices, whether pre‐ or post‐deployment, hardware or software‐driven, affect the performance of deep learning classification and the downstream ecological information that can be extracted from acoustic recordings. In particular, understanding and quantifying detection/classification accuracy and the probability of detection are key to avoid introducing biases that may ultimately affect the quality of data for ecological management.

PAST-AI: Physical-Layer Authentication of Satellite Transmitters via Deep Learning

Physical-layer security is regaining traction in the research community, due to the performance boost introduced by deep learning classification algorithms. This is particularly true for sender authentication in wireless communications via radio fingerprinting. However, previous research mainly focused on terrestrial wireless devices while, to the best of our knowledge, none of the previous work considered satellite transmitters. The satellite scenario is generally challenging because, among others, satellite radio transducers feature non-standard electronics (usually aged and specifically designed for harsh conditions). Moreover, the fingerprinting task is specifically difficult for Low-Earth Orbit (LEO) satellites (like the ones we focus in this paper) since they feature a low bit-rate and orbit at about 800 Km from the Earth, at a speed of around 25,000 Km/h, thus making the receiver experiencing a down-link with unique attenuation and fading characteristics. In this paper, we investigate the effectiveness and main limitations of AI-based solutions to the physical-layer authentication of LEO satellites. Our study is performed on massive real data—more than <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$100M$ </tex-math></inline-formula> I-Q samples—collected from an extensive measurements campaign on the IRIDIUM LEO satellites constellation, lasting 589 hours. Our results show that Convolutional Neural Networks (CNN) and autoencoders (if properly calibrated) can be successfully adopted to authenticate the satellite transducers, with an accuracy spanning between 0.8 and 1, depending on prior assumptions. However, the relatively high number of I-Q samples required by the proposed methodology, coupled with the low bandwidth of satellite link, might prevent the detection of the spoofing attack under certain configuration parameters.