Joana Rocha

Joana Rocha

Auto-generated profile, based on publicly available information

Joana Rocha

Universidade do Porto

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 March 2020 in Journal of Medical Systems
Verified
Journal Article
Closed access
PDF file
JRJoana Rocha
Joana Rocha
António CunhaAntónio Cunha
António Cunha
AMAna Maria Maria Rodrigues de Sousa Faria Mendonça
Ana Maria Maria Rodrigues de Sousa Faria Mendonça

30 November 2023 in Artificial Intelligence in Medicine
Verified
Journal Article
Closed access
JRJoana Rocha
Joana Rocha
SPSofia Cardoso Pereira
Sofia Cardoso Pereira
JPJoão Pedrosa
João Pedrosa
ACAurélio Campilho
Aurélio Campilho
AMAna Maria Mendonça
Ana Maria Mendonça

19 April 2023 in Computer Methods and Programs in Biomedicine
Verified
Journal Article
Open access
SPSofia C. Pereira
Sofia C. Pereira
JRJoana Rocha
Joana Rocha
ACAurelio Campilho
Aurelio Campilho
PSPedro Sousa
Pedro Sousa
AMAna Maria Maria Rodrigues de Sousa Faria Mendonça
Ana Maria Maria Rodrigues de Sousa Faria Mendonça
Convolutional neural networks are widely used to detect radiological findings in chest radiographs. Standard architectures are optimized for images of relatively small size (for example, 224 × 224 pixels), which suffices for most application domains. However, in medical imaging, larger inputs are often necessary to analyze disease patterns. A single scan can display multiple types of radiological findings varying greatly in size, and most models do not explicitly account for this. For a given network, whose layers have fixed-size receptive fields, smaller input images result in coarser features, which better characterize larger objects in an image. In contrast, larger inputs result in finer grained features, beneficial for the analysis of smaller objects. By compromising to a single resolution, existing frameworks fail to acknowledge that the ideal input size will not necessarily be the same for classifying every pathology of a scan. The goal of our work is to address this shortcoming by proposing a lightweight framework for multi-scale classification of chest radiographs, where finer and coarser features are combined in a parameter-efficient fashion. We experiment on CheXpert, a large chest X-ray database. A lightweight multi-resolution (224 × 224, 448 × 448 and 896 × 896 pixels) network is developed based on a Densenet-121 model where batch normalization layers are replaced with the proposed size-specific batch normalization. Each input size undergoes batch normalization with dedicated scale and shift parameters, while the remaining parameters are shared across sizes. Additional external validation of the proposed approach is performed on the VinDr-CXR data set. The proposed approach (AUC 83.27±0.17, 7.1M parameters) outperforms standard single-scale models (AUC 81.76±0.18, 82.62±0.11 and 82.39±0.13 for input sizes 224 × 224, 448 × 448 and 896 × 896, respectively, 6.9M parameters). It also achieves a performance similar to an ensemble of one individual model per scale (AUC 83.27±0.11, 20.9M parameters), while relying on significantly fewer parameters. The model leverages features of different granularities, resulting in a more accurate classification of all findings, regardless of their size, highlighting the advantages of this approach. Different chest X-ray findings are better classified at different scales. Our study shows that multi-scale features can be obtained with nearly no additional parameters, boosting performance.

1 July 2022
Verified
Conference Paper
Closed access
JRJoana Rocha
Joana Rocha
SPSofia Cardoso Pereira
Sofia Cardoso Pereira
JPJoão Pedrosa
João Pedrosa
ACAurelio Campilho
Aurelio Campilho
AMAna Maria Maria Rodrigues de Sousa Faria Mendonça
Ana Maria Maria Rodrigues de Sousa Faria Mendonça
Backed by more powerful computational resources and optimized training routines, deep learning models have attained unprecedented performance in extracting information from chest X-ray data. Preceding other tasks, an automated abnormality detection stage can be useful to prioritize certain exams and enable a more efficient clinical workflow. However, the presence of image artifacts such as lettering often generates a harmful bias in the classifier, leading to an increase of false positive results. Consequently, health care would benefit from a system that selects the thoracic region of interest prior to deciding whether an image is possibly pathologic. The current work tack-les this binary classification exercise using an attention-driven and spatially unsupervised Spatial Transformer Network (STN). The results indicate that the STN achieves similar results to using YOLO-cropped images, with fewer computational expenses and without the need for localization labels. More specifically, the system is able to distinguish between normal and abnormal CheXpert Images with a mean AUC of 84.22%.

30 August 2019
Verified
Book Chapter
Closed access
PDF file
JRJoana Rocha
Joana Rocha
António CunhaAntónio Cunha
António Cunha
AMAna Maria Mendonça
Ana Maria Mendonça

20 August 2019 in Medicine & Science in Sports & Exercise
Verified
Journal Article
Open access
JRJoana N. Rocha
Joana N. Rocha
CBClaire M. Barnes
Claire M. Barnes
PRPaul Rees
Paul Rees
Cain ClarkCain Clark
Cain Clark
Gareth StrattonGareth Stratton
Gareth Stratton
HSHuw Summers
Huw Summers
Purpose: (i) To develop an automated measurement technique for the assessment of both the form and intensity of physical activity undertaken by children during play. (ii) To profile the varying activity across a cohort of children using a multivariate analysis of their movement patterns. Methods: Ankle-worn accelerometers were used to record 40 min of activity during a school recess, for 24 children over five consecutive days. Activity events of 1.1 s duration were identified within the acceleration time trace and compared with a reference motif, consisting of a single walking stride acceleration trace, obtained on a treadmill operating at a speed of 4 km h−1. Dynamic time warping of motif and activity events provided metrics of comparative movement duration and intensity, which formed the data set for multivariate mapping of the cohort activity using a principal component analysis (PCA). Results: The two-dimensional PCA plot provided clear differentiation of children displaying diverse activity profiles and clustering of those with similar movement patterns. The first component of the PCA correlated to the integrated intensity of movement over the 40-min period, whereas the second component informed on the temporal phasing of activity. Conclusions: By defining movement events and then quantifying them by reference to a motion-standard, meaningful assessment of highly varied activity within free play can be obtained. This allows detailed profiling of individual children’s activity and provides an insight on social aspects of play through identification of matched activity time profiles for children participating in conjoined play.