Page 10 - FIMM_Brochure_2022
P. 10
Research Director DfGoirrgaPintradelcCDishiiaaoglnlneonsgteic:s
Johan Lundin Medicine
Johan Lundin, M.D., Ph.D., A paradigm shift from human expert-based interpretations to
was appointed as Research computerised analysis and readout has vast implications for
Director at FIMM in 2011. He both clinical medicine and biomedical research and poses a
is also a Professor of Medical grand challenge for the research community and healthcare in
Technology at the Depart- general.
ment of Global Public Health,
Karolinska Institutet. His Characterisation of biological samples for diagnostic purposes is undergoing a transi-
research is especially focused tion from analogue to digital technology. Along with advancements in assay technol-
on digital diagnostics and ogies, an increasing number of steps in the process can now be effectively enhanced
medical AI to support patient and supported by machine learning and AI. For example, within pathology, cancer
care. Specific research areas research, and microbiology, an expert’s decisions can be supported with an array of
include image-based diag- readouts performed by computer vision algorithms.
nostics of cancer and infec-
tious diseases, with special The goal of FIMM’s Digital Diagnostics for Precision Medicine Grand Challenge is to
emphasis on point-of-care achieve superhuman performance in tissue-based disease outcome prediction and
solutions for resource-limited fully automated digital diagnostics for major diseases, with the support of AI. The
settings. Together with his effort focuses on the many visual tasks currently performed manually as part of the
research group, he has diagnostic process that can be automated by training deep learning classifiers.
invented several cloud-based One of the major advantages of novel AI-based algorithms is the ability to train clas-
and mobile solutions that sifiers for diagnoses that exhibit a high level of complexity. This means that during
allow the diagnostic process the next few years, it will not only become possible to replicate what highly trained
to be performed remotely, by experts do in the assessment of a biological sample, but also supersede human per-
a human observer or using formance with regard to diagnostic precision, accuracy, and consistency. The target
automated computerised diseases we study at FIMM, such as breast and cervical cancer, malaria and neglected
analysis based on AI. tropical diseases, pose major public health challenges worldwide and more efficient
and increasingly automated diagnostics will have a clear global impact.
10 |
FIMM researchers are also developing tailored high-content image analysis methods
and finding computational solutions for phenotyping diseases and drug responses
at a single cell level. The techniques in use combine high-resolution microscopy, laser
microcapture microscopy, image analysis, and machine learning, to enable scalable
molecular analysis of single cells, targetable by proteomics, morphology or location
within the sample.
Digital imaging and readout technologies enable novel solutions for the pharmaceu-
tical industry, particularly with respect to early phase drug development within oncol-
ogy and for novel point-of-care diagnostics. Algorithms that have been developed at
FIMM during the past few years include the analysis of multiple antigens in the same
tissue section (multiplexed immunohistochemistry), analysis of the tumour microen-
vironment, detection of cell subtypes, cancer tissue viability classifiers and a series of
protein expression readout methods. The expertise gained within this grand chal-
lenge is also offered as a service at the FIMM Technology Centre (p. 22).
