RESUMO

The purpose of this study was to identify changes in the temporomandibular joint disc affected by effusion by using texture analysis of magnetic resonance images (MRIs). METHODS: A total of 223 images of the TMJ, 42 with joint effusion and 181 without, were analyzed. Three consecutive slices were then exported to MaZda software, in which two oval ROIs (one in the anterior band and another in the intermediate zone of the joint disc) were determined in each slice and eleven texture parameters were calculated by using a gray-level co-occurrence matrix. Spearman's correlation coefficient test was used to assess the correlation between texture variables and to select variables for analysis. The Mann-Whitney test was used to compare the groups. RESULTS: The significance level was set at 5%, with the results demonstrating that there was no high correlation between the parameter directions. It was possible to observe a trend between the average parameters, in which the group with effusion always had smaller values than the group without effusion, except for the parameter measuring the difference in entropy. CONCLUSION: The trend towards lower overall values for the texture parameters suggested a different behavior between TMJ discs affected by effusion and those not affected, indicating that there may be intrinsic changes.

RESUMO

PURPOSE: Computer-aided diagnosis (CAD) systems on breast ultrasound (BUS) aim to increase the efficiency and effectiveness of breast screening, helping specialists to detect and classify breast lesions. CAD system development requires a set of annotated images, including lesion segmentation, biopsy results to specify benign and malignant cases, and BI-RADS categories to indicate the likelihood of malignancy. Besides, standardized partitions of training, validation, and test sets promote reproducibility and fair comparisons between different approaches. Thus, we present a publicly available BUS dataset whose novelty is the substantial increment of cases with the above-mentioned annotations and the inclusion of standardized partitions to objectively assess and compare CAD systems. ACQUISITION AND VALIDATION METHODS: The BUS dataset comprises 1875 anonymized images from 1064 female patients acquired via four ultrasound scanners during systematic studies at the National Institute of Cancer (Rio de Janeiro, Brazil). The dataset includes biopsy-proven tumors divided into 722 benign and 342 malignant cases. Besides, a senior ultrasonographer performed a BI-RADS assessment in categories 2 to 5. Additionally, the ultrasonographer manually outlined the breast lesions to obtain ground truth segmentations. Furthermore, 5- and 10-fold cross-validation partitions are provided to standardize the training and test sets to evaluate and reproduce CAD systems. Finally, to validate the utility of the BUS dataset, an evaluation framework is implemented to assess the performance of deep neural networks for segmenting and classifying breast lesions. DATA FORMAT AND USAGE NOTES: The BUS dataset is publicly available for academic and research purposes through an open-access repository under the name BUS-BRA: A Breast Ultrasound Dataset for Assessing CAD Systems. BUS images and reference segmentations are saved in Portable Network Graphic (PNG) format files, and the dataset information is stored in separate Comma-Separated Value (CSV) files. POTENTIAL APPLICATIONS: The BUS-BRA dataset can be used to develop and assess artificial intelligence-based lesion detection and segmentation methods, and the classification of BUS images into pathological classes and BI-RADS categories. Other potential applications include developing image processing methods like despeckle filtering and contrast enhancement methods to improve image quality and feature engineering for image description.

Assuntos

RESUMO

ABSTRACT Objective: To explore the diagnostic value of the TUIAS (SW_TH01/II) computer-aided diagnosis (CAD) software system for the ultrasound Thyroid Imaging Reporting and Data System (TI-RADS) features in thyroid nodules. Materials and methods: This retrospective study enrolled patients with thyroid nodules in Shanghai East Hospital between January 2017 and October 2021. The novel CAD software (SW_TH01/II) and three sonographers performed a qualitative analysis of the ultrasound TI-RADS features in aspect ratio, margin irregularity, margin smoothness, calcification, and echogenicity of the thyroid nodules. Results: A total of 225 patients were enrolled. The accuracy, sensitivity, and specificity of the CAD software in "aspect ratio" were 95.6%, 96.2%, and 95.4%, in "margin irregularity" were 90.7%, 90.5%, and 90.9%, in "margin smoothness" were 85.8%, 88.5%, and 83.0%, in "calcification" were 83.6%, 81.7%, and 82.0%, in "homogeneity" were 88.9%, 90.6%, and 82.2%, in "major echo" were 85.3%, 88.0%, and 85.4%, and in "contains very hypoechoic echo" were 92.0%, 90.0%, and 92.4%. The analysis time of the CAD software was significantly shorter than for the sonographers (2.7 ± 1.6 vs. 29.7 ± 12.7 s, P < 0.001). Conclusion: The CAD system achieved high accuracy in describing thyroid nodule features. It might assist in clinical thyroid nodule analysis.

RESUMO

Breast cancer is a significant health concern for women, emphasizing the need for early detection. This research focuses on developing a computer system for asymmetry detection in mammographic images, employing two critical approaches: Dynamic Time Warping (DTW) for shape analysis and the Growing Seed Region (GSR) method for breast skin segmentation. The methodology involves processing mammograms in DICOM format. In the morphological study, a centroid-based mask is computed using extracted images from DICOM files. Distances between the centroid and the breast perimeter are then calculated to assess similarity through Dynamic Time Warping analysis. For skin thickness asymmetry identification, a seed is initially set on skin pixels and expanded based on intensity and depth similarities. The DTW analysis achieves an accuracy of 83%, correctly identifying 23 possible asymmetry cases out of 20 ground truth cases. The GRS method is validated using Average Symmetric Surface Distance and Relative Volumetric metrics, yielding similarities of 90.47% and 66.66%, respectively, for asymmetry cases compared to 182 ground truth segmented images, successfully identifying 35 patients with potential skin asymmetry. Additionally, a Graphical User Interface is designed to facilitate the insertion of DICOM files and provide visual representations of asymmetrical findings for validation and accessibility by physicians.

RESUMO

Objective: To explore the diagnostic value of the TUIAS (SW_TH01/II) computer-aided diagnosis (CAD) software system for the ultrasound Thyroid Imaging Reporting and Data System (TI-RADS) features in thyroid nodules. Materials and methods: This retrospective study enrolled patients with thyroid nodules in Shanghai East Hospital between January 2017 and October 2021. The novel CAD software (SW_TH01/II) and three sonographers performed a qualitative analysis of the ultrasound TI-RADS features in aspect ratio, margin irregularity, margin smoothness, calcification, and echogenicity of the thyroid nodules. Results: A total of 225 patients were enrolled. The accuracy, sensitivity, and specificity of the CAD software in "aspect ratio" were 95.6%, 96.2%, and 95.4%, in "margin irregularity" were 90.7%, 90.5%, and 90.9%, in "margin smoothness" were 85.8%, 88.5%, and 83.0%, in "calcification" were 83.6%, 81.7%, and 82.0%, in "homogeneity" were 88.9%, 90.6%, and 82.2%, in "major echo" were 85.3%, 88.0%, and 85.4%, and in "contains very hypoechoic echo" were 92.0%, 90.0%, and 92.4%. The analysis time of the CAD software was significantly shorter than for the sonographers (2.7 ± 1.6 vs. 29.7 ± 12.7 s, P < 0.001). Conclusion: The CAD system achieved high accuracy in describing thyroid nodule features. It might assist in clinical thyroid nodule analysis.

Assuntos

RESUMO

Neurodevelopment disorders can result in facial dysmorphisms. Therefore, the analysis of facial images using image processing and machine learning techniques can help construct systems for diagnosing genetic syndromes and neurodevelopmental disorders. The systems offer faster and cost-effective alternatives for genotyping tests, particularly when dealing with large-scale applications. However, there are still challenges to overcome to ensure the accuracy and reliability of computer-aided diagnosis systems. This article presents a systematic review of such initiatives, including 55 articles. The main aspects used to develop these diagnostic systems were discussed, namely datasets - availability, type of image, size, ethnicities and syndromes - types of facial features, techniques used for normalization, dimensionality reduction and classification, deep learning, as well as a discussion related to the main gaps, challenges and opportunities.

RESUMO

Training deep Convolutional Neural Networks (CNNs) presents challenges in terms of memory requirements and computational resources, often resulting in issues such as model overfitting and lack of generalization. These challenges can only be mitigated by using an excessive number of training images. However, medical image datasets commonly suffer from data scarcity due to the complexities involved in their acquisition, preparation, and curation. To address this issue, we propose a compact and hybrid machine learning architecture based on the Morphological and Convolutional Neural Network (MCNN), followed by a Random Forest classifier. Unlike deep CNN architectures, the MCNN was specifically designed to achieve effective performance with medical image datasets limited to a few hundred samples. It incorporates various morphological operations into a single layer and uses independent neural networks to extract information from each signal channel. The final classification is obtained by utilizing a Random Forest classifier on the outputs of the last neural network layer. We compare the classification performance of our proposed method with three popular deep CNN architectures (ResNet-18, ShuffleNet-V2, and MobileNet-V2) using two training approaches: full training and transfer learning. The evaluation was conducted on two distinct medical image datasets: the ISIC dataset for melanoma classification and the ORIGA dataset for glaucoma classification. Results demonstrate that the MCNN method exhibits reliable performance in melanoma classification, achieving an AUC of 0.94 (95% CI: 0.91 to 0.97), outperforming the popular CNN architectures. For the glaucoma dataset, the MCNN achieved an AUC of 0.65 (95% CI: 0.53 to 0.74), which was similar to the performance of the popular CNN architectures. This study contributes to the understanding of mathematical morphology in shallow neural networks for medical image classification and highlights the potential of hybrid architectures in effectively learning from medical image datasets that are limited by a small number of case samples.

RESUMO

Alzheimer's disease (AD) is a progressive type of dementia characterized by loss of memory and other cognitive abilities, including speech. Since AD is a progressive disease, detection in the early stages is essential for the appropriate care of the patient throughout its development, going from asymptomatic to a stage known as mild cognitive impairment (MCI), and then progressing to dementia and severe dementia; is worth mentioning that everyone suffers from cognitive impairment to some degree as we age, but the relevant task here is to identify which people are most likely to develop AD. Along with cognitive tests, evaluation of the brain morphology is the primary tool for AD diagnosis, where atrophy and loss of volume of the frontotemporal lobe are common features in patients who suffer from the disease. Regarding medical imaging techniques, magnetic resonance imaging (MRI) scans are one of the methods used by specialists to assess brain morphology. Recently, with the rise of deep learning (DL) and its successful implementation in medical imaging applications, it is of growing interest in the research community to develop computer-aided diagnosis systems that can help physicians to detect this disease, especially in the early stages where macroscopic changes are not so easily identified. This article presents a DL-based approach to classifying MRI scans in the different stages of AD, using a curated set of images from Alzheimer's Disease Neuroimaging Initiative and Open Access Series of Imaging Studies databases. Our methodology involves image pre-processing using FreeSurfer, spatial data-augmentation operations, such as rotation, flip, and random zoom during training, and state-of-the-art 3D convolutional neural networks such as EfficientNet, DenseNet, and a custom siamese network, as well as the relatively new approach of vision transformer architecture. With this approach, the best detection percentage among all four architectures was around 89% for AD vs. Control, 80% for Late MCI vs. Control, 66% for MCI vs. Control, and 67% for Early MCI vs. Control.

RESUMO

Purpose.This study aimed to develop a computer system for automatic detection of thermographic changes indicating breast malignancy risk.Materials and methods. The database contained 233 thermograms of women, including 43 with malignant lesions and 190 with no malignant lesions. Five classifiers were evaluated (k-Nearest Neighbor, Support Vector Machine, Decision Tree, Discriminant Analysis, and Naive Bayes) in combination with oversampling techniques. An attribute selection approach using genetic algorithms was considered. Performance was assessed using accuracy, sensitivity, specificity, AUC, and Kappa statistics.Results. Support vector machines combined with attribute selection by genetic algorithm and ASUWO oversampling obtained the best performance. Attributes were reduced by 41.38%, and accuracy was 95.23%, sensitivity was 93.65%, and specificity was 96.81%. The Kappa index was 0.90, and AUC was 0.99.Conclusion. The feature selection process lowered computational costs and improved diagnostic accuracy. A high-performance system using a new breast imaging modality could positively aid breast cancer screening.

Assuntos

RESUMO

BACKGROUND AND OBJECTIVE: Current studies based on digital biopsy images have achieved satisfactory results in detecting colon cancer despite their limited visual spectral range. Such methods may be less accurate when applied to samples taken from the tumor margin region or to samples containing multiple diagnoses. In contrast with the traditional computer vision approach, micro-FTIR hyperspectral images quantify the tissue-light interaction on a histochemical level and characterize different tissue pathologies, as they present a unique spectral signature. Therefore, this paper investigates the possibility of using hyperspectral images acquired over micro-FTIR absorbance spectroscopy to characterize healthy, inflammatory, and tumor colon tissues. METHODS: The proposed method consists of modeling hyperspectral data into a voxel format to detect the patterns of each voxel using fully connected deep neural network. A web-based computer-aided diagnosis tool for inference is also provided. RESULTS: Our experiments were performed using the K-fold cross-validation protocol in an intrapatient approach and achieved an overall accuracy of 99% using a deep neural network and 96% using a linear support vector machine. Through the experiments, we noticed the high performance of the method in characterizing such tissues using deep learning and hyperspectral images, indicating that the infrared spectrum contains relevant information and can be used to assist pathologists during the diagnostic process.

Assuntos

RESUMO

In December 2019, the spread of the SARS-CoV-2 virus to the world gave rise to probably the biggest public health problem in the world: the COVID-19 pandemic. Initially seen only as a disease of the respiratory system, COVID-19 is actually a blood disease with effects on the respiratory tract. Considering its influence on hematological parameters, how does COVID-19 affect cardiac function? Is it possible to support the clinical diagnosis of COVID-19 from the automatic analysis of electrocardiography? In this work, we sought to investigate how COVID-19 affects cardiac function using a machine learning approach to analyze electrocardiography (ECG) signals. We used a public database of ECG signals expressed as photographs of printed signals, obtained in the context of emergency care. This database has signals associated with abnormal heartbeat, myocardial infarction, history of myocardial infarction, COVID-19, and healthy heartbeat. We propose a system to support the diagnosis of COVID-19 based on hybrid deep architectures composed of pre-trained convolutional neural networks for feature extraction and Random Forests for classification. We investigated the LeNet, ResNet, and VGG16 networks. The best results were obtained with the VGG16 and Random Forest network with 100 trees, with attribute selection using particle swarm optimization. The instance size has been reduced from 4096 to 773 attributes. In the validation step, we obtained an accuracy of 94%, kappa index of 0.91, and sensitivity, specificity, and area under the ROC curve of 100%. This work showed that the influence of COVID-19 on cardiac function is quite considerable: COVID-19 did not present confusion with any heart disease, nor with signs of healthy individuals. It is also possible to build a solution to support the clinical diagnosis of COVID-19 in the context of emergency care from a non-invasive and technologically scalable solution, based on hybrid deep learning architectures.

Assuntos

RESUMO

Vertebral Compression Fracture (VCF) occurs when the vertebral body partially collapses under the action of compressive forces. Non-traumatic VCFs can be secondary to osteoporosis fragility (benign VCFs) or tumors (malignant VCFs). The investigation of the etiology of non-traumatic VCFs is usually necessary, since treatment and prognosis are dependent on the VCF type. Currently, there has been great interest in using Convolutional Neural Networks (CNNs) for the classification of medical images because these networks allow the automatic extraction of useful features for the classification in a given problem. However, CNNs usually require large datasets that are often not available in medical applications. Besides, these networks generally do not use additional information that may be important for classification. A different approach is to classify the image based on a large number of predefined features, an approach known as radiomics. In this work, we propose a hybrid method for classifying VCFs that uses features from three different sources: i) intermediate layers of CNNs; ii) radiomics; iii) additional clinical and image histogram information. In the hybrid method proposed here, external features are inserted as additional inputs to the first dense layer of a CNN. A Genetic Algorithm is used to: i) select a subset of radiomic, clinical, and histogram features relevant to the classification of VCFs; ii) select hyper-parameters of the CNN. Experiments using different models indicate that combining information is interesting to improve the performance of the classifier. Besides, pre-trained CNNs presents better performance than CNNs trained from scratch on the classification of VCFs.

Assuntos

RESUMO

The disease caused by the new type of coronavirus, Covid-19, has posed major public health challenges for many countries. With its rapid spread, since the beginning of the outbreak in December 2019, the disease transmitted by SARS-CoV-2 has already caused over 2 million deaths to date. In this work, we propose a web solution, called Heg.IA, to optimize the diagnosis of Covid-19 through the use of artificial intelligence. Our system aims to support decision-making regarding to diagnosis of Covid-19 and to the indication of hospitalization on regular ward, semi-ICU or ICU based on decision a Random Forest architecture with 90 trees. The main idea is that healthcare professionals can insert 41 hematological parameters from common blood tests and arterial gasometry into the system. Then, Heg.IA will provide a diagnostic report. The system reached good results for both Covid-19 diagnosis and to recommend hospitalization. For the first scenario we found average results of accuracy of 92.891%±0.851, kappa index of 0.858 ± 0.017, sensitivity of 0.936 ± 0.011, precision of 0.923 ± 0.011, specificity of 0.921 ± 0.012 and area under ROC of 0.984 ± 0.003. As for the indication of hospitalization, we achieved excellent performance of accuracies above 99% and more than 0.99 for the other metrics in all situations. By using a computationally simple method, based on the classical decision trees, we were able to achieve high diagnosis performance. Heg.IA system may be a way to overcome the testing unavailability in the context of Covid-19.Communicated by Ramaswamy H. Sarma.

Assuntos

RESUMO

INTRODUCTION: Hepatocellular carcinoma is the prevalent primary liver cancer, a silent disease that killed 782,000 worldwide in 2018. Multimodal deep learning is the application of deep learning techniques, fusing more than one data modality as the model's input. PURPOSE: A computer-aided diagnosis system for hepatocellular carcinoma developed with multimodal deep learning approaches could use multiple data modalities as recommended by clinical guidelines, and enhance the robustness and the value of the second-opinion given to physicians. This article describes the process of creation and evaluation of an algorithm for computer-aided diagnosis of hepatocellular carcinoma developed with multimodal deep learning techniques fusing preprocessed computed-tomography images with structured data from patient Electronic Health Records. RESULTS: The classification performance achieved by the proposed algorithm in the test dataset was: accuracy = 86.9%, precision = 89.6%, recall = 86.9% and F-Score = 86.7%. These classification performance metrics are closer to the state-of-the-art in this area and were achieved with data modalities which are cheaper than traditional Magnetic Resonance Imaging approaches, enabling the use of the proposed algorithm by low and mid-sized healthcare institutions. CONCLUSION: The classification performance achieved with the multimodal deep learning algorithm is higher than human specialists diagnostic performance using only CT for diagnosis. Even though the results are promising, the multimodal deep learning architecture used for hepatocellular carcinoma prediction needs more training and test processes using different datasets before the use of the proposed algorithm by physicians in real healthcare routines. The additional training aims to confirm the classification performance achieved and enhance the model's robustness.

RESUMO

Convolutional Neural Networks (CNNs) have been successfully applied in the medical diagnosis of different types of diseases. However, selecting the architecture and the best set of hyperparameters among the possible combinations can be a significant challenge. The purpose of this work is to investigate the use of the Hyperband optimization algorithm in the process of optimizing a CNN applied to the diagnosis of SARS-Cov2 disease (COVID-19). The test was performed with the Optuna framework, and the optimization process aimed to optimize four hyperparameters: (1) backbone architecture, (2) the number of inception modules, (3) the number of neurons in the fully connected layers, and (4) the learning rate. CNNs were trained on 2175 computed tomography (CT) images. The CNN that was proposed by the optimization process was a VGG16 with five inception modules, 128 neurons in the two fully connected layers, and a learning rate of 0.0027. The proposed method achieved a sensitivity, precision, and accuracy of 97%, 82%, and 88%, outperforming the sensitivity of the Real-Time Polymerase Chain Reaction (RT-PCR) tests (53-88%) and the accuracy of the diagnosis performed by human experts (72%).

Assuntos

RESUMO

Diabetes incidence has been a problem, because according with the World Health Organization and the International Diabetes Federation, the number of people with this disease is increasing very fast all over the world. Diabetic treatment is important to prevent the development of several complications, also lipid profile monitoring is important. For that reason the aim of this work is the implementation of machine learning algorithms that are able to classify cases, that corresponds to patients diagnosed with diabetes that have diabetes treatment, and controls that refers to subjects who do not have diabetes treatment but some of them have diabetes, bases on lipids profile levels. Logistic regression, K-nearest neighbor, decision trees and random forest were implemented, all of them were evaluated with accuracy, sensitivity, specificity and AUC-ROC curve metrics. Artificial neural network obtain an acurracy of 0.685 and an AUC value of 0.750, logistic regression achieve an accuracy of 0.729 and an AUC value of 0.795, K-nearest neighbor gets an accuracy of 0.669 and an AUC value of 0.709, on the other hand, decision tree reached an accuracy pg 0.691 and a AUC value of 0.683, finally random forest achieve an accuracy of 0.704 and an AUC curve of 0.776. The performance of all models was statistically significant, but the best performance model for this problem corresponds to logistic regression.

RESUMO

Histopathology is a well-established standard diagnosis employed for the majority of malignancies, including breast cancer. Nevertheless, despite training and standardization, it is considered operator-dependent and errors are still a concern. Fractal dimension analysis is a computational image processing technique that allows assessing the degree of complexity in patterns. We aimed here at providing a robust and easily attainable method for introducing computer-assisted techniques to histopathology laboratories. Slides from two databases were used: A) Breast Cancer Histopathological; and B) Grand Challenge on Breast Cancer Histology. Set A contained 2480 images from 24 patients with benign alterations, and 5429 images from 58 patients with breast cancer. Set B comprised 100 images of each type: normal tissue, benign alterations, in situ carcinoma, and invasive carcinoma. All images were analyzed with the FracLac algorithm in the ImageJ computational environment to yield the box count fractal dimension (Db) results. Images on set A on 40x magnification were statistically different (p = 0.0003), whereas images on 400x did not present differences in their means. On set B, the mean Db values presented promissing statistical differences when comparing. Normal and/or benign images to in situ and/or invasive carcinoma (all p < 0.0001). Interestingly, there was no difference when comparing normal tissue to benign alterations. These data corroborate with previous work in which fractal analysis allowed differentiating malignancies. Computer-aided diagnosis algorithms may beneficiate from using Db data; specific Db cut-off values may yield ~ 99% specificity in diagnosing breast cancer. Furthermore, the fact that it allows assessing tissue complexity, this tool may be used to understand the progression of the histological alterations in cancer.

RESUMO

Alzheimer's disease is the most prevalent dementia among the elderly population. Early detection is critical because it can help with future planning for those potentially affected. This paper uses a three-dimensional DenseNet architecture to detect Alzheimer's disease in magnetic resonance imaging. Our work is restricted to the use of freely available tools. We constructed a deep neural network classifier with metrics of 0.86¯ mean accuracy, 0.86¯ mean sensitivity (micro-average), 0.86¯ mean specificity (micro-average), and 0.91¯ area under the receiver operating characteristic curve (micro-average) for the task of discriminating between five different disease stages or classes. The use of tools available for free ensures the reproducibility of the study and the applicability of the classification system in developing countries.

Assuntos

RESUMO

Rheumatoid arthritis (RA) is an autoimmune disorder that typically affects people between 23 and 60 years old causing chronic synovial inflammation, symmetrical polyarthritis, destruction of large and small joints, and chronic disability. Clinical diagnosis of RA is stablished by current ACR-EULAR criteria, and it is crucial for starting conventional therapy in order to minimize damage progression. The 2010 ACR-EULAR criteria include the presence of swollen joints, elevated levels of rheumatoid factor or anti-citrullinated protein antibodies (ACPA), elevated acute phase reactant, and duration of symptoms. In this paper, a computer-aided system for helping in the RA diagnosis, based on quantitative and easy-to-acquire variables, is presented. The participants in this study were all female, grouped into two classes: class I, patients diagnosed with RA (n = 100), and class II corresponding to controls without RA (n = 100). The novel approach is constituted by the acquisition of thermal and RGB images, recording their hand grip strength or gripping force. The weight, height, and age were also obtained from all participants. The color layout descriptors (CLD) were obtained from each image for having a compact representation. After, a wrapper forward selection method in a range of classification algorithms included in WEKA was performed. In the feature selection process, variables such as hand images, grip force, and age were found relevant, whereas weight and height did not provide important information to the classification. Our system obtains an AUC ROC curve greater than 0.94 for both thermal and RGB images using the RandomForest classifier. Thirty-eight subjects were considered for an external test in order to evaluate and validate the model implementation. In this test, an accuracy of 94.7% was obtained using RGB images; the confusion matrix revealed our system provides a correct diagnosis for all participants and failed in only two of them (5.3%). Graphical abstract.

Assuntos

RESUMO

The projected burden of dementia by Alzheimer's disease (AD) represents a looming healthcare crisis as the population of most countries grows older. Although there is currently no cure, it is possible to treat symptoms of dementia. Early diagnosis is paramount to the development and success of interventions, and neuroimaging represents one of the most promising areas for early detection of AD. We aimed to deploy advanced deep learning methods to determine whether they can extract useful AD biomarkers from structural magnetic resonance imaging (sMRI) and classify brain images into AD, mild cognitive impairment (MCI), and cognitively normal (CN) groups. We tailored and trained Convolutional Neural Networks (CNNs) on sMRIs of the brain from datasets available in online databases. Our proposed method, ADNet, was evaluated on the CADDementia challenge and outperformed several approaches in the prior art. The method's configuration with machine-learning domain adaptation, ADNet-DA, reached 52.3% accuracy. Contributions of our study include devising a deep learning system that is entirely automatic and comparatively fast, presenting competitive results without using any patient's domain-specific knowledge about the disease. We were able to implement an end-to-end CNN system to classify subjects into AD, MCI, or CN groups, reflecting the identification of distinctive elements in brain images. In this context, our system represents a promising tool in finding biomarkers to help with the diagnosis of AD and, eventually, many other diseases.