A short remaining life expectancy and a high mortality characterise brain tumours as a particularly dangerous disease. In this scope, a software tool OncoFEM is presented, that fully automatic processes from non-invasive magnetic resonance image scans used in diagnostics and prepares a predictive tumour evolution. Relevant information, including the microstructural composition of the healthy tissue and the position of the tumour are identified with different machine learning tools, such as convolutional neural networks. Preparing this into referential states of formed initial boundary value problems are simulated with the relevant processes of tumour growth and regression by embedding a continuum-mechanical model in the framework of the Theory of Porous Media (TPM), cf. Wagner~[1]. The resulting set of governing partial differential equations is treated with the finite element software package FEniCS [2], that allows a flexible implementation of a bi-phasic TPM model. By inclusion of multiple solids and fluid-resolved additive components, relevant clinical questions can be studied from multi-scientific perspectives. The complex phenomenon of tumour growth and regression in brain tissue, cf. Wolf et al.\ [3], becomes even more complex when adding the layer of therapeutic accessibility. In addition to the basic effects of diffusive spreading and the mass effect, actual situations from the tumours emergence growing up to the pre-operative state, its resection and the application of a drug are simulated with a well parametrised model, which leads to the visionary clinical support in diagnosis and planning of a treatment strategy.