Marcelo U. Ferreira | Searcharxiv

arXiv2022

Unfolding selection to infer individual risk heterogeneity for optimising disease forecasts and policy development

Mathematical models are increasing adopted for setting targets for disease prevention and control. As model-informed policies are implemented, however, the inaccuracies of some forecasts become apparent, for example overprediction of infection burdens and overestimation of intervention impacts. Here, we attribute these discrepancies to methodological limitations in capturing the heterogeneities of real-world systems. The mechanisms underpinning single factors for infection and their interactions determine individual propensities to acquire disease. These are potentially so numerous that to attain a full mechanistic description may be unfeasible. To contribute constructively to the development of health policies, model developers either leave factors out (reductionism) or adopt a broader but coarse description (holism). In our view, predictive capacity requires holistic descriptions of heterogeneity which are currently underutilised in infectious disease epidemiology but common in other disciplines.

q-bio.PE↗

arXiv2019

Applying Faster R-CNN for Object Detection on Malaria Images

Deep learning based models have had great success in object detection, but the state of the art models have not yet been widely applied to biological image data. We apply for the first time an object detection model previously used on natural images to identify cells and recognize their stages in brightfield microscopy images of malaria-infected blood. Many micro-organisms like malaria parasites are still studied by expert manual inspection and hand counting. This type of object detection task is challenging due to factors like variations in cell shape, density, and color, and uncertainty of some cell classes. In addition, annotated data useful for training is scarce, and the class distribution is inherently highly imbalanced due to the dominance of uninfected red blood cells. We use Faster Region-based Convolutional Neural Network (Faster R-CNN), one of the top performing object detection models in recent years, pre-trained on ImageNet but fine tuned with our data, and compare it to a baseline, which is based on a traditional approach consisting of cell segmentation, extraction of several single-cell features, and classification using random forests. To conduct our initial study, we collect and label a dataset of 1300 fields of view consisting of around 100,000 individual cells. We demonstrate that Faster R-CNN outperforms our baseline and put the results in context of human performance.

cs.CV↗