The Event Will Cover Topics Addressing
Surge In Medical Data Through Collaboration And Innovation:
Registration to the event is free of charge and only the best talks will be chosen by the committee of this workshop to be presented.
Deadline for Registration
(free of charge)
29 May 2024
The exponential growth of medical and biomedical data from laboratories and hospitals worldwide has imposed a significant burden on healthcare professionals. With a shortage of experts capable of analyzing this vast dataset, there is an urgent demand for the automation of data analysis processes. Prominent examples of this necessity include the analysis of DNA damage and repair in individual cells, which plays a pivotal role in genetic damage assessment and human biomonitoring within medical and biological research, cellular injury, comet assay images, and accurate segmentation of cancer cells via plasma membrane recognition, among others.
This workshop primarily focuses on fostering collaboration among data-collecting specialists in laboratories, image-processing experts, and machine-learning researchers, aiming to share ongoing research obstacles that require automated solutions.
The goal of this workshop is to exploit collective expertise and gather different ideas on cutting-edge automated data analysis through image processing and machine learning techniques. Through networking meetings, researchers will work on how to empower accurate detection, annotation, and segmentation of images obtained in laboratory settings. By bridging the gap between domain-specific knowledge and advanced computational methodologies, we aspire to enhance medical and biomedical data analysis, thereby advancing healthcare and scientific research.
Lavdie Rada
Supported by: Networking Grants of the International Science Partnerships Fund (ISPF) https://acmedsci.ac.uk/networking-grants, NGR1\1718
City University of London
Acıbadem University
Bahçeşehir University
London Metropolitan University
Universidad Anahuac Mayab
Universidad del Valle de Mexico
University of Yucatan
Kadirhas University
Bahçeşehir University
Politeknik Caltex Riau
3rd June
Assessment in Human Biomonitoring and Medical Studies: Unleashing the Power of Image Processing and Machine Learning Fusion
Lavdie Rada - BAU
10:00-10:40
Elda Pacheco-Pantoja -Universidad Anahuac Mayab
Genomic Instability: Comet Assay and Image Analysis for Phenotypic Assessment
10:40-11:20
Tea Break
11:20-11:40
Lavdie Ahmeti -Max Planck Institute for Multidisciplinary Science
Studies of Protein Aggregation in Mouse Oocytes
11:40-12:10
Lunch Break
12:10-13:10
Cefa Karabağ and D. Brito-Pacheco, C. Karabağ, C. Brito-Loeza, P. Giannopoulos, C.C. Reyes-Aldasoro
-London Metropolitan University
Relationship Between Nuclear Invaginations And Mitochondria In Hela Cells Observed With Electron Microscopy
13:10-13:50
Mauricio Alberto Ortega-Ruiz
-Universidad del Valle de Mexico
Deep learning applications for breast cancer imaging segmentation, advances in digital pathology in Latin America
13:50-14:30
14:30-15:00
Tea Break
Süreyya Akyüz and Mohammed Thamer Kamil Al-Khazraji
-BAU
SFG MKL: A Novel Machine Learning Model for Enhanced Robustness and Accuracy
15:00-15:30
Advancing Interdisciplinary Methodologies in Machine Learning and Big Data Analytics: A Journey of Integration and Innovation
Ridvan Bunjaku-AAB College, Prishtina
15:30-16:00
4th June
Carlos Reyes-City
University of London
Biomedical Image Analysis: Applications to Cancer, Microcirculation and Immunology
10:00-10:40
Raşit Eskicioglu
BAU
Current Trends for Health Care Utilizing AI
10:40-11:20
11:20-11:40
Tea Break
A Weight-Induced Sparse Regression Algorithm to Deconvolute Cell-type Medleys in Spatial Transcriptomics Using Single-cell RNA Sequencing Data
Nuray Erdoğan
Kadir Has University
11:40-12:10
Lunch Break
12:10-13:10
Ata Akin
Acibadem University
AI in Neuropsychiatry
13:10-13:50
From Fields to Health: Leveraging Collaborative Approaches in Image Analysis, Machine Learning, and Generative AI for Well-being and Food Security.
Ananda Anada
Politeknik Caltex Riau
13:50-14:30
14:30-15:00
Tea Break
Network Dynamics Reconstruction from Data: Emergent Hypergraphs and Critical Transitions to Synchronization
Deniz Eroğlu
Kadir Has University
15:00-15:40
Osman Bayraktar, Başak Ekinci, Hale Sert, Şafak Yasun, Güray Gürkan, Bora Işıldak, Suat Özkorucuklu
Istanbul University
Enhancing X-Ray Image Classification: A Comparative Study of CNN Models and YOLOv8
15:40-16:10
Carlos Reyes-City University of London
Introduction to biomedical image analysis with Matlab
16:30-19:30
5th June
Carlos Britos University of Yucatan
On Variational and Deep Learning Models for Biomedical Image Processing
10:00-10:40
Imaging, Machine Learning and Computational Fluid Dynamics in Cardiovascular Medicine
İbrahim Başar Aka
Bilgi Universitesi
10:40-11:10
Early Detection of Autistic Children with Eye
Tracker and Artificial Intelligence:
A Case Study Application with Matlab
Erol Duymaz
Ostim Technical University
11:10-11:30
11:30-11:50
Tea Break
Soheil Salahshour
Okan University
A New Approach to Solve SIR Epidemic Dynamic System
11:50-12:20
12:20-13:20
Lunch Break
Spatiotemporal domain-dependent dynamics of a cross-diffusive model in biological pattern generation
Gülsemay Yiğit
BAU
13:20-13:50
Bayesian Approximation in Microwave Medical Imaging
Evrim Tetik
BAU
13:50-14:20
14:20-15:10
Tea Break
Advancements in Automated Image Recognition for Enhanced Data Identification and Fraud Detection
Ossama Krawi
BAU
15:10-15:40
Free Discussions
15:40-16:10
Çırağan Caddesi, No: 4 - 6, 34353
Beşiktaş / İstanbul
444 28 64 • www.bau.edu.tr
The axial skeleton is frequently affected by a chronic, inflammatory disease, impacting a substantial number of individuals. This ailment induces pain and physical limitations, potentially leading to complete erosion, disfusion, or fractures if left untreated. Diagnosis relies on a combination of radiographic imaging and manual assessment, presenting challenges attributed to image clarity, the diagnostic expertise of medical professionals, and the obscured visibility of inflammation, especially in the disease's early stages. This presentation offers an insightful overview of diagnostic procedures and disease progression, shedding light on the constraints inherent in conventional assessments. It advocates for the integration of advanced techniques, specifically proposing the combined utilization of image processing and machine learning. These synergistic approaches aim to enhance image visualization and classification. The study's findings underscore the potential of deep learning in refining both diagnosis and the monitoring of disease progression. Simultaneously, it recognizes the vital supportive role played by image processing in augmenting overall diagnostic accuracy.
"Genomic instability, a hallmark of various human diseases including cancer and genetic disorders, underscores the importance of robust methods for assessing DNA damage at the cellular level. The comet assay emerged as a versatile and widely used technique for evaluating DNA damage and repair kinetics in individual cells. Despite being developed several years ago, this procedure remains a topic of significant interest in biomedical research. This review explores the application of the comet assay in conjunction with advanced image analysis techniques for comprehensive phenotypic assessment of genomic instability. The comet assay, also known as single-cell gel electrophoresis, provides a sensitive measure of DNA strand breaks and fragmentation. By subjecting cells to electrophoresis, damaged DNA fragments migrate away from the nucleus, forming a ""comet-like"" tail whose length and intensity correlate with the extent of DNA damage. This assay enables the detection of genotoxic effects induced by various agents such as chemicals, radiation, and environmental pollutants, as well as the assessment of cellular responses to DNA-damaging insults. In recent years, the integration of advanced image analysis software has significantly enhanced the utility of the comet assay for phenotypic assessment of genomic instability. Image analysis algorithms allow for automated quantification of comet parameters, including tail length, tail moment, and DNA content, facilitating high-throughput data processing and objective interpretation. This enables researchers to elucidate the mechanisms underlying DNA damage and repair, as well as to identify potential genotoxic stressors and evaluate their impact on cellular DNA integrity. Furthermore, the application of the comet assay and image analysis extends beyond basic research to clinical settings, offering valuable insights into disease pathogenesis and therapeutic response. Biomonitoring studies utilize comet assay-based phenotypic assessment to evaluate individual susceptibility to genotoxic agents and to monitor DNA damage in exposed populations. In personalized medicine, the comet assay aids in stratifying patients based on their DNA repair capacity, with the possibility to guide treatment decisions and optimizing therapeutic outcomes.
In addition to these advancements, concerted efforts are underway to establish a Latin American network or work group dedicated to advancing research and standardization efforts in genomic instability assessment using techniques such as the comet assay and image analysis. By fostering collaboration and knowledge exchange among researchers in
the region, this initiative aims to strengthen research capabilities and enhance the impact of genomic instability studies on human health. In summary, the integration of the comet assay with advanced image analysis techniques represents a comprehensive approach for evaluating genomic instability. Through continued technological advancements, interdisciplinary collaboration, and regional networking efforts, this methodology holds promise for advancing our understanding of DNA damage mechanisms and their implications for human health."
Oocytes are one of the most long-lived cells in the organism. Maintaining oocytes in a healthy state is crucial for female fertility. One of the main problems associated with long- lived cells is the accumulation of protein aggregates with age. This has been extensively studied in neurons. However, whether oocytes accumulate protein aggregates with age, and how they cope with such a problem is still unclear. Here, we aim to characterize the types and features of protein aggregates in oocytes, and to monitor if protein aggregates accumulate with age using confocal microscopy. We identified a novel compartment containing protein aggregates, which does not contain ubiquitinated residues. In addition, we revealed that this compartment co-localizes with the Lysosome, and is distinct from the EndoLysosomal Vesicular Assemblies (ELVAs). We also showed that protein aggregates do not increase with age. This suggests that oocytes may have some special mechanisms to prevent further accumulation of protein aggregates as they age. To test this hypothesis, we have established two mammalian cell line systems expressing aggregation-prone proteins, the first is stable inducible HEK293 cells expressing wild-type (Q23), and Mutant (Q74) Huntingtin exon 1. The second is HEK293 cells transiently expressing Aggregating Destabilizing Domains (AgDD). In the future, we plan to express some candidate proteins selected from the oocyte mass spectrometry dataset in these cell lines and test how they affect the formation or clearance of protein aggregates. These findings will not only increase our knowledge on protein aggregation in the oocyte, but also suggest the potential for transplanting oocyte-specific mechanisms to counteract the aggregation of toxic proteins in neurons.
"This paper describes a methodology to analyse the complexity of HeLa cells as observed with electron microscopy, in particular the relationship between mitochondria and the roughness of the nuclear envelope as reflected by the invaginations of the surface. For this purpose, several algorithms to segment the mitochondria were quantitatively compared, namely: Topology, Image Processing, Topology and Image Processing, and
Deep Learning which provides the highest accuracy. The invaginations were segmented with an image processing algorithm. Correlations between the segmented structures were explored for 25 segmented cells. It was found that there was a positive correlation between the volume of invaginations and the volume of mitochondria, and negative correlations between the number and the mean volume of mitochondria, and between the volume of the cytoplasm and the aspect ratio of mitochondria. These results suggest that there is a relationship between the shape of a cell, its nucleus and its mitochondria, as well as a relationship between the number of mitochondria and their shapes. Whilst these results were obtained from a small set of only one type of cell, they encourage further studies as the methodology proposed can be easily applied to other datasets."
This talk introduces SFG MKL, a novel machine learning model that integrates the strengths of multiple kernel learning (MKL) and stochastic functional gradient (SFG) for robust optimization within generative adversarial networks (GANs). SFG MKL leverages SFG for efficient learning from large datasets and incorporates MKL to augment model resilience against adversarial perturbations. We conduct a rigorous comparison between SFG MKL, empirical risk minimization (ERM), and SFG RKHS on challenging image datasets. The results demonstrate that SFG MKL consistently achieves lower error rates and superior performance under adversarial attacks, highlighting the effectiveness of the combined SFG-MKL approach. Interestingly, SFG MKL exhibits competitive efficacy with SFG RKHS in standard scenarios, suggesting that MKL strengthens resilience without compromising accuracy. These findings position SFG MKL as a promising contender for machine learning tasks demanding both accuracy and robustness. Furthermore, the flexibility of exploring diverse kernels within the MKL framework opens exciting avenues for future exploration, particularly in critical domains like biomedical imaging and autonomous systems. Combining gradients with MKL holds significant potential for advancing the development of robust and reliable machine learning algorithms, potentially paving the way for a new paradigm in the field.
This talk encapsulates a journey of exploration and integration at the intersection of Engineering Mathematics, Computer Science, and Data Analysis. With a solid foundation in these disciplines, the research aims to uncover the synergies that exist between them, particularly focusing on advancing methodologies in Machine Learning (ML) and Big Data. The presenter's journey spans over two decades of academic rigor and practical application, manifested in a seamless integration of teaching and industry engagement. Recent studies have delved deeper into Artificial Intelligence (AI), with a specific emphasis on ML and Big Data, enriching the understanding of these transformative technologies and their potential to revolutionize statistical analysis. Two key areas of investigation drive the research: 1) Interdisciplinary Methodologies: This segment focuses on the development of frameworks that integrate mathematical modeling, AI, ML, Big Data analytics, and statistical analysis techniques. The objective is to enhance decision-making and optimization processes by leveraging the collective power of these disciplines. 2) Optimizing Data-Driven Decision-Making: The research explores methodologies to streamline data collection, analysis, and application processes. By identifying efficient approaches, the aim is to empower organizations to make informed decisions based on robust data insights. Through this talk, the presenter expresses a keen interest in fostering collaboration and innovation in the field, with a commitment to advancing methodologies that drive meaningful impact in Machine Learning and Big Data analytics.
Constantino Carlos Reyes-Aldasoro studied Electrical Engineering (BSEE UNAM-Mexico, MSc Imperial College) and Computer Science (PhD Warwick) before joining a group of Cancer Biologists at The University of Sheffield. His role as a postdoctoral associate and then fellow was based at the Department of Surgical Oncology of the Royal Hallamshire Hospital was the analysis of the data acquired by the biologists and clinicians. As he was the only person who was not trained in Medicine or Biology the work offered interesting challenges and opportunities. He has continued to work on analysis of biomedical images of cancer, immunology and microcirculation for several years and has now directed several PhD students in this area. In this presentation, he will share some of the experiences of an Engineer doing Computer Vision immersed in the world of Biology.
In this talk, I will delve into the transformative impact of Artificial Intelligence (AI) on contemporary healthcare practices, elucidating key trends and advancements. With the rapid evolution of technology, AI has emerged as a pivotal tool in revolutionizing various facets of healthcare delivery, from diagnostics to treatment optimization and patient care management. I will also navigate through current trends, shedding light on innovative applications such as predictive analytics, personalized medicine, virtual health assistants, and precision diagnostics.One prominent trend is the integration of AI-powered predictive analytics to forecast disease outbreaks, optimize resource allocation, and enhance decision-making processes for healthcare professionals. Virtual health assistants, enabled by AI algorithms, are reshaping patient engagement and care delivery, providing timely information, monitoring vital signs, and offering personalized recommendations.Furthermore, AI is driving advancements in precision diagnostics, enhancing the accuracy and efficiency of medical imaging interpretation, pathology analysis, and genetic testing. Moreover, AI-powered telemedicine platforms are bridging geographical barriers, expanding access to healthcare services, especially in remote areas. Nevertheless, alongside these advancements, challenges persist, including data privacy concerns, algorithm biases, and ethical considerations. Addressing these challenges is imperative to ensure the responsible and equitable implementation of AI in healthcare. Looking ahead, continued research, collaboration, and regulatory frameworks will be crucial in harnessing the full potential of AI to enhance healthcare delivery, improve patient outcomes, and foster a healthier society.
The state-of-the-art tools of modern science have turned to collecting large and complex data sets with the increase in technological developments and innovations. By using single cell RNA sequencing (scRNA-Seq) methods as today's cutting-edge technology in molecular biology, the heterogeneity of cell populations can be obtained with high resolution through gene expression data at the cellular level. However, this technology cannot detect the spatial positions of cells. On the other hand, barcode-based spatially resolved gene expression profiles obtained from spatial transcriptomics (ST) are playing key role to understand tissue organization and function. But this ST technology lacks single cell resolution. Therefore, deconvolution techniques are key to better understand the cellular profiles and their high-resolution spatial organizations and interaction patterns. In this talk, we will be discussing about a newly developed data-driven machine learning algorithm weight-induced sparse regression (WISpR), that leverages high resolution cell-type data from scRNA-Seq to deconvolute cellular profiles from ST. We will explain how WISpR correctly mapped the cell profiles of developing embryonic human heart and mouse brain and defined overall tissue architecture, successfully. Finally, we will be talking about how WISpR revealed the zone-specific cellular cancer heterogeneity in human breast cancer. Overall, we will be discussing the principle of WISpR, which was driven by the sparsity of nature, and how it will provide promising contributions to personalized medicine as it enables the high-resolution molecular profiling of biological tissues.
Predicting critical transitions in complex systems, such as those observed in the climate or the brain, constitutes a central challenge due to the inherent difficulty in anticipating these abrupt shifts within intricate systems. These transitions can potentially cause significant disruptions to the system's normal functioning. The comprehension of these complex systems fundamentally relies on our capacity to reconstruct appropriate models based on empirical observations. This presentation will delve into the intricacies of reconstructing network dynamics from empirical data. We also discuss the unexpected outcomes, such as revealing phenomena like hypernetwork dynamics, even when the constituent agents of the original system engage in pairwise interactions within the network structure. Ultimately, this recovery process will equip us with the means to forecast critical transitions based on the insights derived from the reconstructed model.
In this study, we focused on X-ray image classification and relatedly disease diagnoses, which has been funded by TUBITAK 2232 Program, an International Fellowship for Outstanding Researchers, with a project number of 121C090. In this project, we will receive data from Medicine Faculty of the Istanbul University. However, in this talk we will represent the studies performed by using public Montgomery and Shenzhen datasets. There are a lot of state of art models for computer vision and image processing. Convolutional Neural Network (CNN) models are frequently used models in image classification. We compared the performance of various CNN and we chose couple of model such as ResNet, DenseNet and AlexNet, as starting point for their high performance in classification problems. In addition, we combined three individual CNN models by applying feature extraction method to create an ensemble model. With the ensemble model, we obtained higher performance compared to individual models. We modified it by adding a classifier at the end of model to enhance generalization ability. This adjustment have yielded satisfactory results. We further extended our approach by incorporating the YOLO algorithm, specifically YOLOv8. The individual classification performance of YOLOv8 exceeded our expectations. In this presentation we will summarize the studies performed and present the preliminary results.
This presentation will examine the application of certain differential operators as regularizers with geometric properties in machine learning models. We will also discuss the stability of neural networks and explore how the principles of ordinary differential equations might be utilized to design more robust networks against adversarial attacks. Finally, we present some deep learning models that we built at the computational learning and imaging research laboratory. These models are designed to address important issues in the southeastern region of Mexico
Cardiovascular disease remains a leading cause of death globally. This talk explores the growing synergy of data, medical imaging, machine learning (ML), and Computational Fluid Dynamics (CFD) in transforming cardiovascular medicine. We discuss the vast amount of data generated from various imaging modalities like echocardiography and CT scans. We delve into how ML algorithms can analyze this data to extract hidden patterns and insights. Additionally, we explore the role of CFD simulations in modeling blood flow dynamics within the cardiovascular system. This integration allows for a deeper understanding of hemodynamics, contributing to improved diagnoses, predictions of disease progression, and personalized treatment plans for cardiovascular patients. Ultimately, this convergence of technologies holds immense potential to enhance patient care and outcomes.
In this talk, we provide a new approach to solve the system of a dynamic system that represents the SIR epidemic model. Since the given model has no exact solution, working on the advanced numerical approach seems a way to go. For this purpose, we first convert the original system to the equivalent nonlinear second-order differential equations by a novel operator approach. Then, we apply some numerical methods to find the solution to the differential equation that was obtained. Keywords: Epidemic Model, Operator Method, Nonlinear differential Equation, Numerical method
"Traditional methods for diagnosing autism often rely on subjective observations and general screening information, leading to inadequacies in early detection. The limitations in self-expression among young children further underscore the need for objective data collection methods. Utilizing technological advancements like eye tracking sensors offers a promising avenue for early diagnosis, particularly in detecting eye contact anomalies, a key symptom of autism.
This study explores the integration of eye tracker data with artificial intelligence algorithms for autism diagnosis. Initial findings demonstrate the high accuracy of our algorithm in diagnosing autism, highlighting the potential of eye tracking sensors as a crucial tool in early detection. The dataset comprises 195 participants, including diagnosed autistic individuals, non-diagnosed individuals, and those with unavailable diagnostic information. Each entry in the dataset is labeled either ASD (autism spectrum disorder) or TD (typically developing), ensuring balanced representation. Through meticulous data cleaning and organization, the raw sensor data is transformed into a tidy format conducive to analysis. Training and test data are partitioned at a 70-30 split, with 66 attributes in total, including personal information columns.
By leveraging eye tracker data and advanced AI techniques, this study demonstrates a promising approach to early autism detection. The presented methodology offers a reliable and objective means of identifying individuals at risk, potentially facilitating timely intervention and support for autistic children. This research underscores the importance of technological innovations in enhancing diagnostic practices and improving outcomes for individuals with autism spectrum disorders."
In this work, a domain-dependent stability analysis for a cross-diffusive system is investigated to understand the role of geometry and cross-diffusion in the evolution of the biological pattern formation. Bifurcation analysis of the system is conducted to understand the spatiotemporal dynamics for the cross-diffusive reaction-diffusion model. We show the conditions on the domain size and generate parameter spaces associated with Turing difusion-driven instability, Hopf and trascritical instabilities. To support theoretical findings, finite element simulations on general two dimensional geometries are presented. We show that finite element simulations reveal the spatial and spatiotemporal behaviour in the dynamics of a cross-diffusive model. These observed patterns resemble which found in the growth process.
This talk explores an ongoing interest of study about microwave medical imaging for breast cancer. Microwave imaging was proposed as a complementary or alternative method to X-ray mammography since it does not share the same disadvantages. Since different tissues and tumors have different electrical properties such as dielectric permittivity and conductivity at microwave frequencies, microwave imaging consists of illuminating the area, measuring the scattered field, and then solving an inverse problem to reconstruct a map of permittivity and conductivity. Most methods that have been developed do not use any a priori information rather than a simple initial guess. However, since the tissue composition of breast, although different for everyone or at different times in a person’s life, is not random. Therefore, implementing statistical or probabilistic methods such as Bayesian Approximation could improve the imaging leading to an early diagnosis.
The convergence of advanced technologies and interdisciplinary collaboration offers profound opportunities to enhance health outcomes and food security. This presentation examines technology areas such as wrist X-ray images, shrimp farming for food security and the analysis of child malnutrition through digital images.
In healthcare, the presentation focuses on using deep learning to analyse the abnormalities within X-ray images. The other is using images as modalities for early detection of child malnutrition. By using deep learning models to indicate the area of abnormalities within the wrist X-ray. The use of digital images can also open opportunities to diagnose growth issues, facilitating timely interventions and improving health outcomes for children.
Vaname shrimp cultivation (Litopenaeus vannamei) is a fish farming activity unit that is widely carried out in several tropical countries such as Indonesia. One of the routine activities in cultivating Vaname shrimp is shrimp sampling in ponds. Rapid sampling should be carried out to avoid reducing the quality of the shrimp and not bringing in diseases from outside the pond. Image processing and analysis are used to detect size (weight and length) through images of Litopenaeus Vannamei and predict its growth, thereby contributing to food security.
Through these case studies, I try to show how a collaborative approach to technology and domain expertise can drive innovation. Integrating these three things can develop effective solutions to global challenges in food security and public health, ultimately improving community welfare.
In an age defined by the ubiquitous presence of digital imagery and the relentless progress of machine vision technologies, the demand for automated techniques to accurately identify specific images of paramount importance within a multitude of datasets is increasingly pressing. This need extends to various domains, including combatting fraud, where the ability to pinpoint desired or missing data is critical.
We present our ongoing research aimed at precisely classifying a given data image within a blind scenario, offering an automated solution capable of constructing a comprehensive database for image recognition tasks. This approach streamlines the cumbersome manual search process typically associated with sifting through vast datasets.
Neuropsychiatry is the clinical field that treats neurological and psychiatric patients. Due to the nature of these diseases the symptoms can sometimes be confusing. New imaging and other psychological tests have been developed to provide more objective measures of these diseases. This talk will emphasize the role of AI in neuropsychiatry and how it can dissolve some of the uncertainty faced by clinicians with the addition of these objective measures. The talk will also introduce the use of fNIRS, a novel neuroimaging tool in understanding the brain dynamics of such patients. The talk aims to convince the audience that employment of AI in medicine is not be feared but should be accepted as a digital consultant.
This presentation is focused on the study of techniques to analyze, quantify, and segment breast cancer tumors from H&E-type images and has a special interest in the segmentation of the different regions of the breast tissue. The main objective is to determine and implement an optimal model for multiclass segmentation, i.e., for the recognition of different regions of breast tissue based on Deep Learning techniques. We also present a brief description of the Automatization and AI implementation in Pathology laboratories in the Latin American region (LATAM) and introduce the newly born research group DigPatho.