Alzheimer’s disease is the most common form of dementia, affecting more than 55 million people globally and accounting for around 70 percent of dementia cases. In Portugal, it is estimated that 200,000 people live with this condition, posing significant challenges to health systems due to the need for long-term care and the high associated costs. It is characterized by a progressive and irreversible degeneration of brain cells, resulting in cognitive decline, memory loss and functional impairment. In addition to the impact on patients’ quality of life, Alzheimer’s burdens carers and health services, requiring continuous supervision and specialized support. Although there is no cure, early diagnosis is essential to optimize therapeutic interventions and slow down the progression of symptoms.

Home / Publications / Publication

Home / Publications / Publication

Mão com Puzzle IA

DEEP NEURAL NETWORKS AND THE FUTURE OF EARLY DETECTION OF ALZHEIMER’S DISEASE

March 02, 25

Publication type: Article Summary
Original title: Diagnóstico da doença de Alzheimer com redes neuronais profundas
Article publication date: July 2022
Source: Repositório da Universidade do Minho
Author: Mateus Ferreira da Silva
Supervisor: António Esteves

What is the goal, target audience, and areas of digital health it addresses?
     The study aims to evaluate the application of deep neural networks in the early diagnosis of Alzheimer’s disease, exploring advanced computational methods for analysing and interpreting Magnetic Resonance Imaging (MRI). The target audience for this work includes researchers, health professionals, biomedical engineers and health technology companies. The study falls within areas of digital health, such as computer-aided diagnosis, computational neuroimaging and artificial intelligence applied to medicine, with an emphasis on the use of convolutional neural networks (CNNs).

What is the context?
     Alzheimer’s disease is the most common form of dementia, affecting more than 55 million people globally and accounting for around 70 percent of dementia cases. In Portugal, it is estimated that 200,000 people live with this condition, posing significant challenges to health systems due to the need for long-term care and the high associated costs.

     It is characterized by a progressive and irreversible degeneration of brain cells, resulting in cognitive decline, memory loss and functional impairment. In addition to the impact on patients’ quality of life, Alzheimer’s burdens carers and health services, requiring continuous supervision and specialized support. Although there is no cure, early diagnosis is essential to optimize therapeutic interventions and slow down the progression of symptoms.

What are the current approaches?
     Traditional methodologies for diagnosing Alzheimer’s disease include neuropsychological tests used to assess cognitive functions, but they have limitations, such as subjectivity in interpretation and low sensitivity for identifying the disease in its early stages.

     Analysing biomarkers in the cerebrospinal fluid, combined with neuroimaging tests, such PET (Positron Emission Tomography) and MRI, has been fundamental in identifying brain alterations characteristic of the disease. Structural MRI images make it possible to detect morphological changes associated with neurodegeneration, including volumetric reduction of the hippocampus, cortical atrophy and expansion of the cerebral ventricles. These structural biomarkers are widely used in clinical practice to differentiate between healthy control, mild cognitive impairment and Alzheimer’s disease. However, inter-observer variability and dependence on the specialist’s experience make it difficult to standardize the diagnosis, limiting its clinical applicability.

     To overcome these limitations, machine learning techniques have been widely explored. Supervised models, such as Support Vector Machines and Random Forest, are used to classify medical images and clinical data, allowing patterns associated with the disease to be identified. Unsupervised approaches, such as clustering, make it easier to analyze disease progression by grouping together patients with similar neurodegenerative characteristics.

     Deep learning has shown great potential in the automated analysis of neuroimaging, with CNNs standing out in the detection of structural patterns associated with neurodegeneration. In parallel, Recurrent Neural Networks or Transformers have been applied to monitor cognitive progression over time, providing insights into the evolution of the disease. Hybrid models, which integrate neuroimaging, cognitive tests and structural biomarkers detected with machine learning, demonstrate greater reliability by providing a comprehensive analysis of disease progression and supporting clinical decision-making.

     Despite advances, significant challenges remain, including the interpretability of deep learning models, the variability of clinical data and the need for large-scale validation. The clinical acceptance of these technologies depends on the development of models that are explainable and adaptable to the needs of healthcare professionals.

What does innovation consist of? How is the impact of this study assessed?
     The innovation of this study consisted of applying deep learning to the automated detection of structural alterations in the brain associated with Alzheimer’s disease, overcoming the limitations of conventional methods.

     The analysis process began with the acquisition of high-resolution structural MRI images from the Alzheimer’s Disease Neuroimaging Initiative, one of the largest data repositories for Alzheimer’s research, containing mainly data from North American patients. The use of this database allowed the model to be trained and validated with a diverse set of images of healthy individuals and patients at different stages of the disease, ensuring greater robustness and generalization of the results. To ensure standardization and optimization of the images for automated analysis, rigorous pre-processing was carried out, including spatial normalization, adjusting the images to a standard format to facilitate comparison between individuals and large-scale analysis. The removal of the skull, carried out using the Brain Extraction Tool, eliminated non-brain structures such as the skull and adjacent tissues, allowing the analysis to focus exclusively on brain tissue. In addition, brain segmentation was carried out, separating different brain regions to facilitate the identification of areas affected by neurodegeneration.

     To enable the images to be used in the deep learning models, the three-dimensional MRI images were transformed into axial slices and organized into 4×4 RGB matrices, making them compatible with both 2D and 3D CNNs models. To optimize image storage and loading during model training, the TFRecords structure was used, an efficient binary format that reduces data reading and processing time, improving the scalability of computational analysis.

     After pre-processing, the CNN models analysed structural biomarkers of the disease, including the volumetric reduction of the hippocampus, cortical atrophy and the expansion of the cerebral ventricles. Based on these biomarkers, these models were trained to classify the images into three categories: healthy control, characterized by a preserved brain structure with no signs of neurodegeneration; mild cognitive impairment, an intermediate stage in which there is some brain atrophy without significant progression of the disease; and Alzheimer’s disease, in which there is severe atrophy in the hippocampus and entorhinal cortex, associated with marked cognitive decline. To mitigate the limitation imposed by the small number of samples available, a data augmentation strategy was applied that includes techniques such as rotating the images at different angles, varying the intensity to simulate differences in image acquisition and random cuts, creating artificial variations that improve the robustness of the model and reduce the risk of overfitting.

     The impact of this study was evaluated using rigorous performance metrics, including the matriz de confusão, which quantified classification accuracy and identified error patterns (such as false negatives). In addition, the Receiver Operating Characteristic curve and its related Area Under the Curve (scale from 0 to 1, where 1 represents a perfect classification) were used to measure the models’ ability to differentiate the three categories.

What are the main results? What is the future of this approach?
     The results of this study showed that 3D CNN models generally outperformed 2D models in detecting Alzheimer’s disease, benefiting from the complete volumetric analysis of the brain. Among the models tested, ResNext50 3D was the best performer, achieving an Area Under the Curve of 0.70 for the Alzheimer’s disease category, 0.65 for the mild cognitive impairment category and 0.73 for the healthy control category. However, classifying the mild cognitive impairment category proved more challenging, with a high false negative rate, reflecting the difficulty in correctly distinguishing this condition from the others. The Receiver Operating Characteristic curve confirmed this trend, with a lower Area Under the Curve for mild cognitive impairment, indicating the model’s difficulties in distinguishing this condition from healthy control and Alzheimer’s disease. The increase in data did not significantly improve the results, suggesting that the variations generated were highly correlated with the originals and did not contribute to diversifying the training set.

     For the future, it will be essential to expand the database to include more representative samples, optimize pre-processing methods and explore binary classifications (Alzheimer’s disease vs. healthy control) to improve diagnostic accuracy. The development of explainable artificial intelligence techniques will be key to increasing the interpretability of the models, ensuring greater reliability and clinical acceptance. In addition, the integration of multimodal analyses, combining neuroimaging with genetic and liquid biomarkers, could improve diagnosis and allow more detailed monitoring of disease progression.

Would you like to know all the details?

Know more here

Do you have an innovative idea in healthcare field?

Share it with us and see it come to life.
We will help bring your projects to life!

Share my idea

Newsletter

Receive the latest updates from the InovarSaúde portal.

Subscribe

Support

FAQs

Contacts

República Portuguesa logo
logotipo SNS
SPMS logotipo

Follow Us

Co-funded by

PRR Logotipo
república Portuguesa logo
União Europeia Logo

Newsletter

Receive the latest updates from the InovarSaúde portal.

Subscribe

Support

FAQs

Contacts

República Portuguesa logo
SNS Logo
SPMS Logo

Follow Us

Co-funded by

PRR Logotipo
República Portuguesa logo
União Europeia Logo

Home / Publications / Publication

Mão com Puzzle IA

DEEP NEURAL NETWORKS AND THE FUTURE OF EARLY DETECTION OF ALZHEIMER’S DISEASE

March 02, 25

Publication type: Article Summary
Original title: Diagnóstico da doença de Alzheimer com redes neuronais profundas
Article publication date: July 2022
Source: Repositório da Universidade do Minho
Author: Mateus Ferreira da Silva
Supervisor: António Esteves

What is the goal, target audience, and areas of digital health it addresses?
     The study aims to evaluate the application of deep neural networks in the early diagnosis of Alzheimer’s disease, exploring advanced computational methods for analysing and interpreting Magnetic Resonance Imaging (MRI). The target audience for this work includes researchers, health professionals, biomedical engineers and health technology companies. The study falls within areas of digital health, such as computer-aided diagnosis, computational neuroimaging and artificial intelligence applied to medicine, with an emphasis on the use of convolutional neural networks (CNNs).

What is the context?
     Alzheimer’s disease is the most common form of dementia, affecting more than 55 million people globally and accounting for around 70 percent of dementia cases. In Portugal, it is estimated that 200,000 people live with this condition, posing significant challenges to health systems due to the need for long-term care and the high associated costs.

     It is characterized by a progressive and irreversible degeneration of brain cells, resulting in cognitive decline, memory loss and functional impairment. In addition to the impact on patients’ quality of life, Alzheimer’s burdens carers and health services, requiring continuous supervision and specialized support. Although there is no cure, early diagnosis is essential to optimize therapeutic interventions and slow down the progression of symptoms.

What are the current approaches?
     Traditional methodologies for diagnosing Alzheimer’s disease include neuropsychological tests used to assess cognitive functions, but they have limitations, such as subjectivity in interpretation and low sensitivity for identifying the disease in its early stages.

     Analysing biomarkers in the cerebrospinal fluid, combined with neuroimaging tests, such PET (Positron Emission Tomography) and MRI, has been fundamental in identifying brain alterations characteristic of the disease. Structural MRI images make it possible to detect morphological changes associated with neurodegeneration, including volumetric reduction of the hippocampus, cortical atrophy and expansion of the cerebral ventricles. These structural biomarkers are widely used in clinical practice to differentiate between healthy control, mild cognitive impairment and Alzheimer’s disease. However, inter-observer variability and dependence on the specialist’s experience make it difficult to standardize the diagnosis, limiting its clinical applicability.

     To overcome these limitations, machine learning techniques have been widely explored. Supervised models, such as Support Vector Machines and Random Forest, are used to classify medical images and clinical data, allowing patterns associated with the disease to be identified. Unsupervised approaches, such as clustering, make it easier to analyze disease progression by grouping together patients with similar neurodegenerative characteristics.

     Deep learning has shown great potential in the automated analysis of neuroimaging, with CNNs standing out in the detection of structural patterns associated with neurodegeneration. In parallel, Recurrent Neural Networks or Transformers have been applied to monitor cognitive progression over time, providing insights into the evolution of the disease. Hybrid models, which integrate neuroimaging, cognitive tests and structural biomarkers detected with machine learning, demonstrate greater reliability by providing a comprehensive analysis of disease progression and supporting clinical decision-making.

     Despite advances, significant challenges remain, including the interpretability of deep learning models, the variability of clinical data and the need for large-scale validation. The clinical acceptance of these technologies depends on the development of models that are explainable and adaptable to the needs of healthcare professionals.

What does innovation consist of? How is the impact of this study assessed?
     The innovation of this study consisted of applying deep learning to the automated detection of structural alterations in the brain associated with Alzheimer’s disease, overcoming the limitations of conventional methods.

     The analysis process began with the acquisition of high-resolution structural MRI images from the Alzheimer’s Disease Neuroimaging Initiative, one of the largest data repositories for Alzheimer’s research, containing mainly data from North American patients. The use of this database allowed the model to be trained and validated with a diverse set of images of healthy individuals and patients at different stages of the disease, ensuring greater robustness and generalization of the results. To ensure standardization and optimization of the images for automated analysis, rigorous pre-processing was carried out, including spatial normalization, adjusting the images to a standard format to facilitate comparison between individuals and large-scale analysis. The removal of the skull, carried out using the Brain Extraction Tool, eliminated non-brain structures such as the skull and adjacent tissues, allowing the analysis to focus exclusively on brain tissue. In addition, brain segmentation was carried out, separating different brain regions to facilitate the identification of areas affected by neurodegeneration.

     To enable the images to be used in the deep learning models, the three-dimensional MRI images were transformed into axial slices and organized into 4×4 RGB matrices, making them compatible with both 2D and 3D CNNs models. To optimize image storage and loading during model training, the TFRecords structure was used, an efficient binary format that reduces data reading and processing time, improving the scalability of computational analysis.

     After pre-processing, the CNN models analysed structural biomarkers of the disease, including the volumetric reduction of the hippocampus, cortical atrophy and the expansion of the cerebral ventricles. Based on these biomarkers, these models were trained to classify the images into three categories: healthy control, characterized by a preserved brain structure with no signs of neurodegeneration; mild cognitive impairment, an intermediate stage in which there is some brain atrophy without significant progression of the disease; and Alzheimer’s disease, in which there is severe atrophy in the hippocampus and entorhinal cortex, associated with marked cognitive decline. To mitigate the limitation imposed by the small number of samples available, a data augmentation strategy was applied that includes techniques such as rotating the images at different angles, varying the intensity to simulate differences in image acquisition and random cuts, creating artificial variations that improve the robustness of the model and reduce the risk of overfitting.

     The impact of this study was evaluated using rigorous performance metrics, including the matriz de confusão, which quantified classification accuracy and identified error patterns (such as false negatives). In addition, the Receiver Operating Characteristic curve and its related Area Under the Curve (scale from 0 to 1, where 1 represents a perfect classification) were used to measure the models’ ability to differentiate the three categories.

What are the main results? What is the future of this approach?
     The results of this study showed that 3D CNN models generally outperformed 2D models in detecting Alzheimer’s disease, benefiting from the complete volumetric analysis of the brain. Among the models tested, ResNext50 3D was the best performer, achieving an Area Under the Curve of 0.70 for the Alzheimer’s disease category, 0.65 for the mild cognitive impairment category and 0.73 for the healthy control category. However, classifying the mild cognitive impairment category proved more challenging, with a high false negative rate, reflecting the difficulty in correctly distinguishing this condition from the others. The Receiver Operating Characteristic curve confirmed this trend, with a lower Area Under the Curve for mild cognitive impairment, indicating the model’s difficulties in distinguishing this condition from healthy control and Alzheimer’s disease. The increase in data did not significantly improve the results, suggesting that the variations generated were highly correlated with the originals and did not contribute to diversifying the training set.

     For the future, it will be essential to expand the database to include more representative samples, optimize pre-processing methods and explore binary classifications (Alzheimer’s disease vs. healthy control) to improve diagnostic accuracy. The development of explainable artificial intelligence techniques will be key to increasing the interpretability of the models, ensuring greater reliability and clinical acceptance. In addition, the integration of multimodal analyses, combining neuroimaging with genetic and liquid biomarkers, could improve diagnosis and allow more detailed monitoring of disease progression.

Would you like to know all the details?

Know more here
Sistema robótico autónomo INSIDE

Autonomous Robotics System for Autism Therapy

Autism spectrum disorder is a neurodevelopmental condition with significant clinical, social and economic repercussions throughout life. According to the World Health Organization, it is estimated to affect approximately 1 in 160 children worldwide. Its origin…

Read more
Enfermeira com um telefone

Mobile Application to Improve Workflows in Nursing Homes

Portugal has one of the highest aging populations in the world, placing increasing pressure on elderly care services, especially in nursing homes. Healthcare professionals in these facilities are often overwhelmed due to the increasing number…

Read more
troca de informações de saúde e interoperabilidade

New Era of Interoperability in Healthcare Systems

The growing use of electronic health records, digital diagnostic systems and remote monitoring technologies has led to a significant increase in the volume and complexity of health data. This increase intensifies the need for continuous,…

Read more
robótica colaborativa

Collaborative Robotics Improves Working Conditions

Workers face growing challenges in the industrial environment. Among the most critical are fatigue and inappropriate postures, often associated with repetitive tasks and working conditions that lack ergonomic suitability. These factors represent significant risks for…

Read more
Benefícios da Eletrônica Médica

Detection of Anxiety and Panic Attacks in Real Time

The growing number of people with anxiety disorders, along with increased awareness of mental health, drives the need for new technological tools that provide remote and continuous monitoring of anxiety and panic disorders. Thus, the…

Read more
tele-ecografia

A Novel Approach for Robotic-assisted Tele-echography

Currently, robotic systems for ultrasound diagnostic procedures fall into two main categories: portable robots that require manual positioning and fully autonomous robotic systems that independently control the ultrasound probe’s orientation and positioning. Portable robots rely…

Read more
Personalização e tecnologia na gestão da Diabetes

Personalization and Technology in Diabetes Management

IPDM has significant potential to improve diabetes management and drive health system reforms to become high-performing, effective, equitable, accessible, and sustainable. Evidence and good practices inspire health system transformation. Adopting person-centred approaches like co-creation and…

Read more
TEF-HEALTH Logo

SPMS Integrates the TEF-Health Initiative

SPMS participates in the TEF-Health initiative as a partner in a consortium composed of 51 entities from 9 European Union countries. This action is co-financed by the European Commission and has a duration of five…

Read more
Global Digital Health Partnership Logo

SPMS Represents Portugal as Vice-president of GDHP

The GDHP is an intergovernmental organization in the digital health sector that facilitates cooperation and collaboration between government representatives and the World Health Organization (WHO). Its purpose is to foster policymaking that promote the digitalization…

Read more
Portugal INCoDe.2030

Digital Transformation of Health at INCoDe.2030 in Tomar

The “National Digital Skills Initiative e.2030, Portugal” (INCoDe.2030) is an initiative that aims to improve the Portuguese population’s level of digital skills, placing Portugal at the level of the most advanced European countries in this…

Read more
HealthData@PT Logo

HealthData@PT: New SPMS Initiative for Health Data

Action HealthData@PT is launched in the context of the implementation of the European Health Data Space, and is an initiative approved by the European Commission under the EU4Health 2021-2027 programme. This initiative contributes to the…

Read more

Do you have an innovative idea in healthcare field?

Share it with us and see it come to life.
We will help bring your projects to life!

Share my idea

Newsletter

Receive the latest updates from the Inovarsaúde portal.

Subscribe

Support

FAQs

Contacts

República Portuguesa logo
SNS Logo
SPMS Logo

Follow Us

Co-funded by

PRR Logotipo
República Portuguesa logo
União Europeia Logo
Scroll to Top