The study, recently published in Biophysics Reviews, shows how artificial intelligence can be used to develop faster, cheaper and more reliable information about cells, while also eliminating the disadvantages from using chemicals in the process.
Extracting quantitative biological information from bright-field cell images using deep learning
Saga Helgadottir, Benjamin Midtvedt, Jesús Pineda, Alan Sabirsh, Caroline B. Adiels, Stefano Romeo, Daniel Midtvedt, Giovanni Volpe
Biophysics Rev. 2, 031401 (2021)
Quantitative analysis of cell structures is essential for biomedical and pharmaceutical research. The standard imaging approach relies on fluorescence microscopy, where cell structures of interest are labeled by chemical staining techniques. However, these techniques are often invasive and sometimes even toxic to the cells, in addition to being time-consuming, labor-intensive, and expensive. Here, we introduce an alternative deep-learning-powered approach based on the analysis of bright-field images by a conditional generative adversarial neural network (cGAN). We show that this approach can extract information from the bright-field images to generate virtually-stained images, which can be used in subsequent downstream quantitative analyses of cell structures. Specifically, we train a cGAN to virtually stain lipid droplets, cytoplasm, and nuclei using bright-field images of human stem-cell-derived fat cells (adipocytes), which are of particular interest for nanomedicine and vaccine development. Subsequently, we use these virtually-stained images to extract quantitative measures about these cell structures. Generating virtually-stained fluorescence images is less invasive, less expensive, and more reproducible than standard chemical staining; furthermore, it frees up the fluorescence microscopy channels for other analytical probes, thus increasing the amount of information that can be extracted from each cell.
Neural Network Training with Highly Incomplete Datasets
Yu-Wei Chang, Laura Natali, Oveis Jamialahmadi, Stefano Romeo, Joana B. Pereira, Giovanni Volpe
Neural network training and validation rely on the availability of large high-quality datasets. However, in many cases only incomplete datasets are available, particularly in health care applications, where each patient typically undergoes different clinical procedures or can drop out of a study. Since the data to train the neural networks need to be complete, most studies discard the incomplete datapoints, which reduces the size of the training data, or impute the missing features, which can lead to artefacts. Alas, both approaches are inadequate when a large portion of the data is missing. Here, we introduce GapNet, an alternative deep-learning training approach that can use highly incomplete datasets. First, the dataset is split into subsets of samples containing all values for a certain cluster of features. Then, these subsets are used to train individual neural networks. Finally, this ensemble of neural networks is combined into a single neural network whose training is fine-tuned using all complete datapoints. Using two highly incomplete real-world medical datasets, we show that GapNet improves the identification of patients with underlying Alzheimer’s disease pathology and of patients at risk of hospitalization due to Covid-19. By distilling the information available in incomplete datasets without having to reduce their size or to impute missing values, GapNet will permit to extract valuable information from a wide range of datasets, benefiting diverse fields from medicine to engineering.
Virtual genetic diagnosis for familial hypercholesterolemia powered by machine learning
Anna Pina, Saga Helgadottir, Rosellina Margherita Mancina, Chiara Pavanello, Carlo Pirazzi, Tiziana Montalcini, Roberto Henriques, Laura Calabresi, Olov Wiklund, M Paula Macedo, Luca Valenti, Giovanni Volpe, Stefano Romeo
European Journal of Preventive Cardiology (2020)
Familial hypercholesterolemia (FH) is the most common genetic disorder of lipid metabolism. The gold standard for FH diagnosis is genetic testing, available, however, only in selected university hospitals. Clinical scores – for example, the Dutch Lipid Score – are often employed as alternative, more accessible, albeit less accurate FH diagnostic tools. The aim of this study is to obtain a more reliable approach to FH diagnosis by a “virtual” genetic test using machine-learning approaches.
Methods and results
We used three machine-learning algorithms (a classification tree (CT), a gradient boosting machine (GBM), a neural network (NN)) to predict the presence of FH-causative genetic mutations in two independent FH cohorts: the FH Gothenburg cohort (split into training data (N = 174) and internal test (N = 74)) and the FH-CEGP Milan cohort (external test, N = 364). By evaluating their area under the receiver operating characteristic (AUROC) curves, we found that the three machine-learning algorithms performed better (AUROC 0.79 (CT), 0.83 (GBM), and 0.83 (NN) on the Gothenburg cohort, and 0.70 (CT), 0.78 (GBM), and 0.76 (NN) on the Milan cohort) than the clinical Dutch Lipid Score (AUROC 0.68 and 0.64 on the Gothenburg and Milan cohorts, respectively) in predicting carriers of FH-causative mutations.
In the diagnosis of FH-causative genetic mutations, all three machine-learning approaches we have tested outperform the Dutch Lipid Score, which is the clinical standard. We expect these machine-learning algorithms to provide the tools to implement a virtual genetic test of FH. These tools might prove particularly important for lipid clinics without access to genetic testing.