Data scientists at the Communications Engineering Branch (CEB) of the Lister Hill National Center for Biomedical Communications (LHC) at the National Library of Medicine (NLM) have developed new intelligent methods to screen for infectious diseases, such as malaria and tuberculosis. These methods take advantage of the latest results in machine learning, which is a subfield of artificial intelligence. Particularly, scientists used deep learning methods to detect and count infected red blood cells and parasites. Deep learning is a family of machine learning methods based on artificial neural networks, which are inspired by biological neural networks of the central nervous system of animals. Scientists at CEB advanced and adapted these methods to take account of the specific features of blood smear images, in particular the relatively small size of parasites. Other challenges that scientists had to overcome include variations of color, lighting, and similar variations introduced by smear preparation and subsequent image capture. The data scientists at CEB developed deep learning networks for both thin and thick smears, which are the two types of bloods smears used by experts in the field to diagnose malaria. The networks can detect blood cells in thin smears and classify them into infected and uninfected cells. They can also detect parasites in thick smears, which is a new feature developed this fiscal year, thus covering the full screening spectrum in practice. Scientists laid special focus on executing trained deep learning networks on Android smartphones. This required research into designing new and smaller network structures that can cope with the lower hardware specifications of smartphones in terms of processing units and memory. Scientists were able to develop and execute smaller networks than can provide the same performance and that can run in reasonable time on smartphones. They were able to do so for thin and thick smears. The developed system is the first smartphone application for malaria screening that can process both thin and thick smears using deep learning. A better implementation of the software, exploiting parallelization, and modifications to the user interface improved the runtime and general usability of the smartphone application. This also includes a cloud-based image upload function for uploading images to a central server for further processing or archiving, and language support for multiple languages. To train the mobile smartphone application for thick smears, scientists used an annotated image set from 200 patients, which they acquired the year before at a hospital in Chittagong, Bangladesh. This set includes about 3000 thick smear images with 85,000 parasites and 60,000 white blood cells, all manually annotated. This data was acquired for Plasmodium falciparum, which is the deadliest parasite species causing malaria in humans. In this fiscal year, similar data has been acquired for thin smears featuring Plasmodium vivax, which is another common parasite species causing malaria. In addition, data acquisition for Plasmodium vivax in thick smears has started. Once this data acquisition is completed, it will allow training of deep learning networks for detecting and discriminating parasite species. CEB has made the software publicly available in the Google Play Store (NLM MalariaScreener), where researchers can download it for testing or contributing training data. Scientists at CEB have been actively collaborating with several sites in Thailand, Bangladesh, Kenya, Uganda, Pakistan, Thailand, and Mali for field-testing the software with blood smear images from different labs.
| Poostchi, Mahdieh; Silamut, Kamolrat; Maude, Richard J et al. (2018) Image analysis and machine learning for detecting malaria. Transl Res 194:36-55 |
| Rajaraman, Sivaramakrishnan; Antani, Sameer K; Poostchi, Mahdieh et al. (2018) Pre-trained convolutional neural networks as feature extractors toward improved malaria parasite detection in thin blood smear images. PeerJ 6:e4568 |
