NanoSTIMA – RL5
The emergent clinical scenarios that are envisioned in this project will give rise to large amounts of data coming from different sources, including ubiquitous sensing, clinical data from electronic health records, and diverse devices being increasingly used for medical imaging and biomedical signal monitoring.
In this context, the use computer-aided detection and diagnosis (CAD) tools will gain increased importance in order to be able to deal with the data processing and analysis in a tractable way, so that clinical research and practice can take advantage of the gathered information, giving rise to more effective workflows.
Nowadays CAD tools aim to assist clinicians in the stages of clinical practice, either for screening, objective detection, measurement and characterization of relevant information from images and signals, or providing support for diagnosis, prognosis and decisions (e.g. treatment, surgery, ...) by easing the integration of the available information.
Medical researchers are now building analytic tools that can turn raw data into actionable intelligence. Some are mining existing databases of medical treatments and outcomes to gather statistical evidence. Others are looking to the patient specific data to find cues about the state of the patient. Medical imaging, especially x-ray, ultrasonography and magnetic resonance imaging, are crucial in treatment and diagnosis, since effective decisions depend on correct diagnosis. Though medical/clinical judgment may be sufficient in the treatment of many conditions, the use of diagnostic imaging services is paramount in confirming, correctly assessing and documenting the course of the disease as well as in assessing response to treatment.
Although research on Computer-Aided Diagnosis (CAD) and quantitative medical image analysis of lesions has been ongoing for decades, major gains must be made before such detection and analysis are universally accepted for use in the clinical arena. Currently, use of computers in the medical image interpretation process is mainly reserved as an aid to the radiologist, serving only to register or enhance images and give secondary interpretations. Although computer-aided detection (CADe) is widely used to assist radiologists with the reading, there is a need to improve performance of existing CADe algorithms, particularly with respect to sensitivity and the false-positive rates. In addition, the current acquisition modes of some medical imaging devices, like ultrasound, are very operator dependent, and their automatic analysis is still in its infancy.
Apart from the improvement of the currently used tools, new features and challenges are envisioned for the future CADe applications. For example, novel ways to present CADe results to clinicians should be investigated to improve its potential use in the clinical decision, as well as with better explanatory capabilities of the diagnostics and prognostics. Furthermore, new screening technologies are emerging, and there is a growing need for CADe applications for these new modalities, as well as for incorporation of biomechanical modelling into the analysis chain to account for motion and correlation across multiple views and modalities (i.e., image fusion) - multimodal data integration for personalized health.
One of the primary aims of this research line is not merely to follow trends or evaluate the latest imaging technology, but to develop new capabilities for future research and development as well as potential commercial clinical outcomes. We will face the problem of a faster evolution of modalities, with increased complexity, requiring a faster evolution and adaption of CAD tools. This requires research on innovative methodologies for CAD development, that allow to switch from ad hoc engineering approaches, driven by the automation of direct expert knowledge, to more automated approaches, driven by the intrinsic structure of data, knowledge discovery and expert supervision. Problems tackled will be generic in that appropriate outcomes can be applied universally to medical imaging (e.g. radiology) practice as a whole. Such research may be termed “Generic Enabling Research (GER)”. In order to tackle the specific objectives of this research line, we approach the problem from a number of different inter-related work packages, each with its own requirements and objectives.