Automatic Auditing System for Endoscopic Exploration of the Stomach with Artificial Intelligence-Gastro UNAL: Gastroendoscopy UNit for Automatic Labeling / Sistema de auditoría automática para la exploración endoscópica del estómago con Inteligencia Artificial - Gastro UNAL: Gastroendoscopy UNit for Automatic Labeling

Abstract Introduction: Upper endoscopy is the standard method for diagnosing early-stage gastric cancer. However, according to estimates, up to 20% of tumors are not detected, and their accuracy may be affected by the variability in their performance. In Colombia, most diagnoses take place in advanced stages, which aggravates the problem. Protocols have been proposed to ensure the complete observation of areas prone to premalignant lesions to address variability. Objective: To build and validate an automatic audit system for endoscopies using artificial intelligence techniques. Methodology: In this study, 96 patients from a teaching hospital underwent video-documented endoscopies, spanning 22 stations rearranged to minimize overlaps and improve the identification of 13 key gastric regions. An advanced convolutional network was used to process the images, extracting visual characteristics, which facilitated the training of artificial intelligence in the classification of these areas. Results: the model, called Gastro UNAL, was trained and validated with images of 67 patients (70% of cases) and tested with 29 different patients (30% of cases), which reached an average sensitivity of 85,5% and a specificity of 98,8% in detecting the 13 gastric regions. Conclusions: The effectiveness of the model suggests its potential to ensure the quality and accuracy of endoscopies. This approach could confirm the regions evaluated, alerting less experienced or trained endoscopists about blind spots in the examinations, thus, increasing the quality of these procedures.

Resumen Introducción: La endoscopia digestiva alta es el método estándar para diagnosticar el cáncer gástrico en etapas tempranas. Sin embargo, su precisión puede verse afectada por la variabilidad en su realización, y se estiman hasta 20% de tumores no detectados. En Colombia, la mayoría de los diagnósticos se realizan en etapas avanzadas, lo que agrava el problema. Para abordar la variabilidad, se han propuesto protocolos con el fin de asegurar la observación completa de áreas propensas a lesiones premalignas. Objetivo: Construir y validar un sistema de auditoría automática para endoscopias usando técnicas de inteligencia artificial. Metodología: En este estudio, 96 pacientes de un hospital universitario se sometieron a endoscopias documentadas en video, abarcando 22 estaciones reorganizadas para minimizar solapamientos y mejorar la identificación de 13 regiones gástricas clave. Se utilizó una red convolucional avanzada para procesar las imágenes, extrayendo características visuales, lo que facilitó el entrenamiento de la inteligencia artificial en la clasificación de estas áreas. Resultados: El modelo, llamado Gastro UNAL, fue entrenado y validado con imágenes de 67 pacientes (70% de los casos) y probado con 29 pacientes distintos (30% de los casos), con lo que alcanzó una sensibilidad promedio del 85,5% y una especificidad del 98,8% en la detección de las 13 regiones gástricas. Conclusiones: La eficacia del modelo sugiere su potencial para asegurar la calidad y precisión de las endoscopias. Este enfoque podría confirmar las regiones evaluadas, alertando puntos ciegos en la exploración a los endoscopistas con menos experiencia o en entrenamiento, de tal forma que se aumente la calidad de estos procedimientos.

Deep Density Estimation for Cone Counting and Diagnosis of Genetic Eye Diseases From Adaptive Optics Scanning Light Ophthalmoscope Images.

Purpose: Adaptive optics scanning light ophthalmoscope (AOSLO) imaging offers a microscopic view of the living retina, holding promise for diagnosing and researching eye diseases like retinitis pigmentosa and Stargardt's disease. The technology's clinical impact of AOSLO hinges on early detection through automated analysis tools. Methods: We introduce Cone Density Estimation (CoDE) and CoDE for Diagnosis (CoDED). CoDE is a deep density estimation model for cone counting that estimates a density function whose integral is equal to the number of cones. CoDED is an integration of CoDE with deep image classifiers for diagnosis. We use two AOSLO image datasets to train and evaluate the performance of cone density estimation and classification models for retinitis pigmentosa and Stargardt's disease. Results: Bland-Altman plots show that CoDE outperforms state-of-the-art models for cone density estimation. CoDED reported an F1 score of 0.770 ± 0.04 for disease classification, outperforming traditional convolutional networks. Conclusions: CoDE shows promise in classifying the retinitis pigmentosa and Stargardt's disease cases from a single AOSLO image. Our preliminary results suggest the potential role of analyzing patterns in the retinal cellular mosaic to aid in the diagnosis of genetic eye diseases. Translational Relevance: Our study explores the potential of deep density estimation models to aid in the analysis of AOSLO images. Although the initial results are encouraging, more research is needed to fully realize the potential of such methods in the treatment and study of genetic retinal pathologies.

Etiology of Macular Edema Defined by Deep Learning in Optical Coherence Tomography Scans.

Purpose: To develop an automated method based on deep learning (DL) to classify macular edema (ME) from the evaluation of optical coherence tomography (OCT) scans. Methods: A total of 4230 images were obtained from data repositories of patients attended in an ophthalmology clinic in Colombia and two free open-access databases. They were annotated with four biomarkers (BMs) as intraretinal fluid, subretinal fluid, hyperreflective foci/tissue, and drusen. Then the scans were labeled as control or ocular disease among diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), and retinal vein occlusion (RVO) by two expert ophthalmologists. Our method was developed by following four consecutive phases: segmentation of BMs, the combination of BMs, feature extraction with convolutional neural networks to achieve binary classification for each disease, and, finally, multiclass classification of diseases and control images. Results: The accuracy of our model for nAMD was 97%, and for DME, RVO, and control were 94%, 93%, and 93%, respectively. Area under curve values were 0.99, 0.98, 0.96, and 0.97, respectively. The mean Cohen's kappa coefficient for the multiclass classification task was 0.84. Conclusions: The proposed DL model may identify OCT scans as normal and ME. In addition, it may classify its cause among three major exudative retinal diseases with high accuracy and reliability. Translational Relevance: Our DL approach can optimize the efficiency and timeliness of appropriate etiological diagnosis of ME, thus improving patient access and clinical decision making. It could be useful in places with a shortage of specialists and for readers that evaluate OCT scans remotely.

Grading diabetic retinopathy and prostate cancer diagnostic images with deep quantum ordinal regression.

Although for many diseases there is a progressive diagnosis scale, automatic analysis of grade-based medical images is quite often addressed as a binary classification problem, missing the finer distinction and intrinsic relation between the different possible stages or grades. Ordinal regression (or classification) considers the order of the values of the categorical labels and thus takes into account the order of grading scales used to assess the severity of different medical conditions. This paper presents a quantum-inspired deep probabilistic learning ordinal regression model for medical image diagnosis that takes advantage of the representational power of deep learning and the intrinsic ordinal information of disease stages. The method is evaluated on two different medical image analysis tasks: prostate cancer diagnosis and diabetic retinopathy grade estimation on eye fundus images. The experimental results show that the proposed method not only improves the diagnosis performance on the two tasks but also the interpretability of the results by quantifying the uncertainty of the predictions in comparison to conventional deep classification and regression architectures. The code and datasets are available at https://github.com/stoledoc/DQOR.

Classification of diabetes-related retinal diseases using a deep learning approach in optical coherence tomography.

BACKGROUND AND OBJECTIVES: Spectral Domain Optical Coherence Tomography (SD-OCT) is a volumetric imaging technique that allows measuring patterns between layers such as small amounts of fluid. Since 2012, automatic medical image analysis performance has steadily increased through the use of deep learning models that automatically learn relevant features for specific tasks, instead of designing visual features manually. Nevertheless, providing insights and interpretation of the predictions made by the model is still a challenge. This paper describes a deep learning model able to detect medically interpretable information in relevant images from a volume to classify diabetes-related retinal diseases. METHODS: This article presents a new deep learning model, OCT-NET, which is a customized convolutional neural network for processing scans extracted from optical coherence tomography volumes. OCT-NET is applied to the classification of three conditions seen in SD-OCT volumes. Additionally, the proposed model includes a feedback stage that highlights the areas of the scans to support the interpretation of the results. This information is potentially useful for a medical specialist while assessing the prediction produced by the model. RESULTS: The proposed model was tested on the public SERI-CUHK and A2A SD-OCT data sets containing healthy, diabetic retinopathy, diabetic macular edema and age-related macular degeneration. The experimental evaluation shows that the proposed method outperforms conventional convolutional deep learning models from the state of the art reported on the SERI+CUHK and A2A SD-OCT data sets with a precision of 93% and an area under the ROC curve (AUC) of 0.99 respectively. CONCLUSIONS: The proposed method is able to classify the three studied retinal diseases with high accuracy. One advantage of the method is its ability to produce interpretable clinical information in the form of highlighting the regions of the image that most contribute to the classifier decision.

Accurate and reproducible invasive breast cancer detection in whole-slide images: A Deep Learning approach for quantifying tumor extent.

With the increasing ability to routinely and rapidly digitize whole slide images with slide scanners, there has been interest in developing computerized image analysis algorithms for automated detection of disease extent from digital pathology images. The manual identification of presence and extent of breast cancer by a pathologist is critical for patient management for tumor staging and assessing treatment response. However, this process is tedious and subject to inter- and intra-reader variability. For computerized methods to be useful as decision support tools, they need to be resilient to data acquired from different sources, different staining and cutting protocols and different scanners. The objective of this study was to evaluate the accuracy and robustness of a deep learning-based method to automatically identify the extent of invasive tumor on digitized images. Here, we present a new method that employs a convolutional neural network for detecting presence of invasive tumor on whole slide images. Our approach involves training the classifier on nearly 400 exemplars from multiple different sites, and scanners, and then independently validating on almost 200 cases from The Cancer Genome Atlas. Our approach yielded a Dice coefficient of 75.86%, a positive predictive value of 71.62% and a negative predictive value of 96.77% in terms of pixel-by-pixel evaluation compared to manually annotated regions of invasive ductal carcinoma.

An automatic method for skeletal patterns classification using craniomaxillary variables on a Colombian population.

BACKGROUND: The mandibular bone is an important part of the forensic facial reconstruction and it has the possibility of getting lost in skeletonized remains; for this reason, it is necessary to facilitate the identification process simulating the mandibular position only through craniomaxillary measures, for this task, different modeling techniques have been performed, but they only contemplate a straight facial profile that belong to skeletal pattern Class I, but the 24.5% corresponding to the Colombian skeletal patterns Class II and III are not taking into account, besides, craniofacial measures do not follow a parametric trend or a normal distribution. OBJECTIVE: The aim of this study was to employ an automatic non-parametric method as the Support Vector Machines to classify skeletal patterns through craniomaxillary variables, in order to simulate the natural mandibular position on a contemporary Colombian sample. MATERIALS AND METHODS: Lateral cephalograms (229) of Colombian young adults of both sexes were collected. Landmark coordinates protocols were used to create craniomaxillary variables. A Support Vector Machine with a linear kernel classifier model was trained on a subset of the available data and evaluated over the remaining samples. The weights of the model were used to select the 10 best variables for classification accuracy. RESULTS: An accuracy of 74.51% was obtained, defined by Pr-A-N, N-Pr-A, A-N-Pr, A-Te-Pr, A-Pr-Rhi, Rhi-A-Pr, Pr-A-Te, Te-Pr-A, Zm-A-Pr and PNS-A-Pr angles. The Class Precision and the Class Recall showed a correct distinction of the Class II from the Class III and vice versa. CONCLUSIONS: Support Vector Machines created an important model of classification of skeletal patterns using craniomaxillary variables that are not commonly used in the literature and could be applicable to the 24.5% of the contemporary Colombian sample.

Visual pattern mining in histology image collections using bag of features.

OBJECTIVE: The paper addresses the problem of finding visual patterns in histology image collections. In particular, it proposes a method for correlating basic visual patterns with high-level concepts combining an appropriate image collection representation with state-of-the-art machine learning techniques. METHODOLOGY: The proposed method starts by representing the visual content of the collection using a bag-of-features strategy. Then, two main visual mining tasks are performed: finding associations between visual-patterns and high-level concepts, and performing automatic image annotation. Associations are found using minimum-redundancy-maximum-relevance feature selection and co-clustering analysis. Annotation is done by applying a support-vector-machine classifier. Additionally, the proposed method includes an interpretation mechanism that associates concept annotations with corresponding image regions. The method was evaluated in two data sets: one comprising histology images from the different four fundamental tissues, and the other composed of histopathology images used for cancer diagnosis. Different visual-word representations and codebook sizes were tested. The performance in both concept association and image annotation tasks was qualitatively and quantitatively evaluated. RESULTS: The results show that the method is able to find highly discriminative visual features and to associate them to high-level concepts. In the annotation task the method showed a competitive performance: an increase of 21% in f-measure with respect to the baseline in the histopathology data set, and an increase of 47% in the histology data set. CONCLUSIONS: The experimental evidence suggests that the bag-of-features representation is a good alternative to represent visual content in histology images. The proposed method exploits this representation to perform visual pattern mining from a wider perspective where the focus is the image collection as a whole, rather than individual images.

Content-based histopathology image retrieval using a kernel-based semantic annotation framework.

Large amounts of histology images are captured and archived in pathology departments due to the ever expanding use of digital microscopy. The ability to manage and access these collections of digital images is regarded as a key component of next generation medical imaging systems. This paper addresses the problem of retrieving histopathology images from a large collection using an example image as query. The proposed approach automatically annotates the images in the collection, as well as the query images, with high-level semantic concepts. This semantic representation delivers an improved retrieval performance providing more meaningful results. We model the problem of automatic image annotation using kernel methods, resulting in a unified framework that includes: (1) multiple features for image representation, (2) a feature integration and selection mechanism (3) and an automatic semantic image annotation strategy. An extensive experimental evaluation demonstrated the effectiveness of the proposed framework to build meaningful image representations for learning and useful semantic annotations for image retrieval.

Automatic annotation of histopathological images using a latent topic model based on non-negative matrix factorization.

Histopathological images are an important resource for clinical diagnosis and biomedical research. From an image understanding point of view, the automatic annotation of these images is a challenging problem. This paper presents a new method for automatic histopathological image annotation based on three complementary strategies, first, a part-based image representation, called the bag of features, which takes advantage of the natural redundancy of histopathological images for capturing the fundamental patterns of biological structures, second, a latent topic model, based on non-negative matrix factorization, which captures the high-level visual patterns hidden in the image, and, third, a probabilistic annotation model that links visual appearance of morphological and architectural features associated to 10 histopathological image annotations. The method was evaluated using 1,604 annotated images of skin tissues, which included normal and pathological architectural and morphological features, obtaining a recall of 74% and a precision of 50%, which improved a baseline annotation method based on support vector machines in a 64% and 24%, respectively.

A semi-automatic method for quantification and classification of erythrocytes infected with malaria parasites in microscopic images.

Visual quantification of parasitemia in thin blood films is a very tedious, subjective and time-consuming task. This study presents an original method for quantification and classification of erythrocytes in stained thin blood films infected with Plasmodium falciparum. The proposed approach is composed of three main phases: a preprocessing step, which corrects luminance differences. A segmentation step that uses the normalized RGB color space for classifying pixels either as erythrocyte or background followed by an Inclusion-Tree representation that structures the pixel information into objects, from which erythrocytes are found. Finally, a two step classification process identifies infected erythrocytes and differentiates the infection stage, using a trained bank of classifiers. Additionally, user intervention is allowed when the approach cannot make a proper decision. Four hundred fifty malaria images were used for training and evaluating the method. Automatic identification of infected erythrocytes showed a specificity of 99.7% and a sensitivity of 94%. The infection stage was determined with an average sensitivity of 78.8% and average specificity of 91.2%.