Research at the Department of Computing is both theoretical and applied; the researchers collaborate widely across many fields. Thus, when viewing Tapio Pahikkala’s list of publications you can find publications from several different disciplines varying from medicine to geography.
Tapio Pahikkala’s research interests consists of theory and algorithmics of machine learning and data analysis methods. In the field of medicine, artificial intelligence (AI) is extensively used in various applications, such as image processing and interpretation. In the best scenario, AI might be a future tool to automatically detect cancer or tumors in medical imaging or if not replacing, helping radiologists in their daily work. However, extensive amount of research is still needed in order to get there. When creating algorithmics, it is crucial to measure how the “black box of artificial intelligence” actually works. Thus, the spesific topic which Tapio wants to highlight, is the importance of experimental design and performance evaluation to guarantee generalizability of the model. As Tapio explains, it is possible that algorithmics might work in scanning spesific patients’ MRI images, but fail when it comes to analyzing different patients’ data. Tapio uses so called resampling methods, which try to overcome the generalizability challenge.
Tapio has been involved in several Academy of Finland consortium projects. Currently the group consisting of researhers from University of Turku, University of Helsinki and Aalto University are investigating machine learning framework for predicting the responses of drug combinations to certain diseases, such as cancer. The screening of different drug combinations and doses with traditional methods is almost impossible and the computational methods are needed to guide the selection of effective combinations to be prioritized for further laboratory testing. Recent finding published in Nature communications (https://www.nature.com/articles/s41467-020-19950-z) suggests that certain machine learning framework created for predicting drug combination responses of cancer cell lines (comboFM) provides an effective means for systematic pre-screening of drug combinations.
The other big project Tapio is currently working with, is Business Finland funded PRIVASA consortium project, which is led by Tapio Pahikkala together with Antti Airola. The aim of the project is to develop artificial intelligence methods for the safe utilization of sensitive data in the health sector. Privacy-preserving algorithms convert sensitive data into a safe but useful anonymous format that is suitable for medical research, testing and validation.
For more information on PRIVASA project, please visit:
Important research topics and lots of possibilities for collaboration with neuroscience field. As Tapio said to me, he is always willing to discuss the collaboration possibilities. In addition, he sees the popularization of this difficult field of study as important.