Automatic Lung Nodule Detection Combined With Gaze Information Improves Radiologists’ Screening Performance

Aresta, Guilherme; Ferreira, Carlos; Pedrosa, João; Araújo, Teresa; Rebelo, Joana; Negrão, Eduardo; Morgado, Margarida; Alves, Filipe; Cunha, António; Ramos, Isabel; Campilho, Aurélio

doi:10.1109/jbhi.2020.2976150

Cited by 22 publications

(12 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the past five decades eye tracking has been extensively used in radiology for education, perception understanding, and fatigue measurement (example reviews: [4][5][6][7] ). More recently, efforts [8][9][10][11] have used eye tracking data to improve segmentation and disease classification in Computed Tomography (CT) radiography by integrating them in deep learning techniques. With such evidence and with the lack of public datasets that capture eye gaze data in the chest X-Ray (CXR) space, we present a new dataset that can help improve the way machine learning models are developed for radiology applications and we demonstrate its use in some popular deep learning architectures.…”

Section: Background and Summarymentioning

confidence: 99%

Creation and validation of a chest X-ray dataset with eye-tracking and report dictation for AI development

et al. 2021

View full text Add to dashboard Cite

We developed a rich dataset of Chest X-Ray (CXR) images to assist investigators in artificial intelligence. The data were collected using an eye-tracking system while a radiologist reviewed and reported on 1,083 CXR images. The dataset contains the following aligned data: CXR image, transcribed radiology report text, radiologist’s dictation audio and eye gaze coordinates data. We hope this dataset can contribute to various areas of research particularly towards explainable and multimodal deep learning/machine learning methods. Furthermore, investigators in disease classification and localization, automated radiology report generation, and human-machine interaction can benefit from these data. We report deep learning experiments that utilize the attention maps produced by the eye gaze dataset to show the potential utility of this dataset.

show abstract

Section: Background and Summarymentioning

confidence: 99%

Creation and validation of a chest X-ray dataset with eye-tracking and report dictation for AI development

et al. 2021

View full text Add to dashboard Cite

show abstract

“…The research conducted by [17], where the researchers decided to route an intelligent system to detect lung infections in asthmatic patients applied to improve diagnosis in radiologists. Obtaining as a result an efficiency of 91.02%, but this system due to its low sensitivity, can suffer from environmental interference and this can affect the results according to the evaluation of the patient.…”

Section: Discussionmentioning

confidence: 99%

“…In [17], researchers rely on the rapid identification of any pathology that may affect lung conditions by using various technologies, such as computed tomography, which is a tool that will facilitate the doctor in visualizing pathologies, as well as could reduce patient deaths because their analysis response is rapid, So it could be applied in asthmatics, to avoid that over time it can be related to lung cancer, which would undoubtedly affect patients to the point of causing death, therefore, they decided to develop an intelligent system to detect lung infections in asthmatic patients applied to improve diagnosis in radiologists. The procedure of the researchers focuses on the classification of lung infections, based on this, to be able to detect the pathologies that may occur in the pulmonary system of patients using lung ultrasound as a tool, then perform an image processing through Python to generate a more accurate vision for medical specialists when using this system that has a low percentage of error.…”

Section: International Journal On Recent and Innovation Trends In Com...mentioning

confidence: 99%

Automatic System for Detecting Pathologies in the Respiratory System for the Care of Patients with Bronchial Asthma Visualized by Computerized Radiography

Meneses-Claudio

Perez-Siguas²,

Matta-Solís³

et al. 2023

IJRITCC

View full text Add to dashboard Cite

COVID-19 is a disease that directly affects the respiratory tract, being harmful in asthmatic patients because this condition causes a lack of oxygen, even to the extent that it requires external equipment to combat drowning. Likewise, it highlights the importance of maintaining a treatment or therapy correctly to prevent your disease from worsening or being exposed to other diseases, putting your health at risk. The diseases that asthmatic patients can acquire can result from pathologies, which have been growing over the years due to lack of equipment or efficient examinations that generate complete information about respiratory conditions about asthmatic patients, therefore, by developing an advanced system, the chances of detecting pathologies prematurely increase considerably, which is an essential tool today. According to the problem exposed, in this research an automatic system of detection of pathologies in the respiratory system was carried out for the care of the patient with bronchial asthma visualized by computerized radiography, so that any pathology can be detected by means of a premature diagnosis in the respiratory system and the doctor can perform a correct treatment on the asthmatic patient. Through the tests carried out by the system, its performance was accurate and efficient, being suitable to be implemented in various hospitals so that the doctor can treat the disease in time, since the system presented a 98.79% efficiency in the detection of pathologies.

show abstract

“…The studies mentioned previously used ETD as GT and images as input to the models, while other approach is to use ETD as input. In [6], a gaze analysis algorithm was used to cluster In [60], another study was performed using ETD regarding chest CT images. In this case, a CNN architecture based on the YOLOv3 model [61] was developed and applied in the task of lung nodule segmentation.…”

Section: Use Of Eye-tracking Datamentioning

confidence: 99%

Automatic Eye-Tracking-Assisted Chest Radiography Pathology Screening

Santos

Pedrosa

Mendonça

et al. 2023

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

The lungs and heart are targets of a myriad of life-threatening diseases. Chest radiography, one of the most common medical procedures worldwide, is a vital tool to study such conditions, since it is easy to perform and affordable. However, the high demand for Chest X-Ray (CXR) examinations creates a huge burden on radiologists, who could benefit from the support of new Artificial Intelligence (AI) methods.With the ongoing digital transformation in clinical practice chest radiographs are being stored in massive amounts, which can be collected and annotated to create large datasets with the goal of developing Deep Learning (DL) models. Such models can be applied in the task of pathology classification, for instance, to provide a second opinion or for patient stratification. Nevertheless, DL models are undergoing a limited implementation in real world scenarios, due to the lack of trustworthy explanations that provide insights about their reasoning. Consequently, DL generates reluctance in the medical community.The need for explanations generated a change of paradigm in DL, and an increasing focus in the topic of explainability. Several methods have been developed to accommodate this need, including model-agnostic methods that can be used with virtually any DL model. This also prompted the design of models capable of rendering better explanations, since providing further evidence alone is insufficient.In this work, Eye-Tracking Data (ETD) was used to train models with the goal of performing CXR pathology classification in a more explainable way. ETD consists, essentially, in a set of data points relative to the gaze fixations of radiologists, including the respective coordinates and time duration. ETD was converted to images, namely heatmaps, in order to be used with Convolutional Neural Networks (CNN). The first stage consisted in reconstructing the heatmaps using CXR images as inputs. This has the main advantage of making the proposed framework independent of the ETD during inference. Then, two approaches were developed. The first one consists in using the reconstructed heatmaps to segment the thorax in CXR images. By providing only the thorax area to a subsequent classifier, it is guaranteed that the model does not use irrelevant information present in other parts of the images. The second approach leverages the fact that reconstructed heatmaps contain higher pixel intensities in pathological areas, and multiplies them by the corresponding CXR images. Ultimately, the objective is to guide the classifier into focusing on important features for the prediction.Three ETD datasets available in the literature, EGD, REFLACX, and CXR-P, were used to train and/or test the models. Furthermore, several experiments were performed to select the best architectures for both the heatmap reconstruction and pathology classification tasks. A UNet with a pretrained DenseNet121 encoder was chosen for the first task, and a pretrained EfficientNet-b0 as the classifier for both approaches. The developed models were compared with the st...

show abstract

Automatic Lung Nodule Detection Combined With Gaze Information Improves Radiologists’ Screening Performance

Cited by 22 publications

References 23 publications

Creation and validation of a chest X-ray dataset with eye-tracking and report dictation for AI development

Creation and validation of a chest X-ray dataset with eye-tracking and report dictation for AI development

Automatic System for Detecting Pathologies in the Respiratory System for the Care of Patients with Bronchial Asthma Visualized by Computerized Radiography

Automatic Eye-Tracking-Assisted Chest Radiography Pathology Screening

Contact Info

Product

Resources

About