VERDICT (vascular, extracellular and restricted diffusion for cytometry in tumours) estimates and maps microstructural features of cancerous tissue non‐invasively using diffusion MRI. The main purpose of this study is to address the high computational time of microstructural model fitting for prostate diagnosis, while retaining utility in terms of tumour conspicuity and repeatability. In this work, we adapt the accelerated microstructure imaging via convex optimization (AMICO) framework to linearize the estimation of VERDICT parameters for the prostate gland. We compare the original non‐linear fitting of VERDICT with the linear fitting, quantifying accuracy with synthetic data, and computational time and reliability (performance and precision) in eight patients. We also assess the repeatability (scan‐rescan) of the parameters. Comparison of the original VERDICT fitting versus VERDICT‐AMICO showed that the linearized fitting (1) is more accurate in simulation for a signal‐to‐noise ratio of 20 dB; (2) reduces the processing time by three orders of magnitude, from 6.55 seconds/voxel to 1.78 milliseconds/voxel; (3) estimates parameters more precisely; (4) produces similar parametric maps and (5) produces similar estimated parameters with a high Pearson correlation between implementations, r 2 > 0.7. The VERDICT‐AMICO estimates also show high levels of repeatability. Finally, we demonstrate that VERDICT‐AMICO can estimate an extra diffusivity parameter without losing tumour conspicuity and retains the fitting advantages. VERDICT‐AMICO provides microstructural maps for prostate cancer characterization in seconds.

Abstract. Prostate cancer diagnosis involves the highly subjective and time-consuming Gleason grading process. This paper proposes the use of Max-Margin Conditional Random Fields (CRFs) towards the aim of creating an automatic computer-aided diagnosis system. Unlike previous methods, this approach enables us to fuse information from multiple classifiers while leveraging CRFs to model spatial dependencies. We perform grading on superpixels which reduce redundancy and the size of data. Probabilistic outputs from independent classifiers are passed as input to a Max-Margin CRF, which then performs structured prediction on the biopsy core, segmenting the image into regions of benign tissue, Gleason grade 3 adenocarcinoma and Gleason grade 4 adenocarcinoma. The system achieves an accuracy of 83.0% with accuracies of 83.6%, 86.9% and 77.1% reported for benign, grade 3 and grade 4 classes respectively.

The VERDICT (Vascular, Extracellular and Restricted Diffusion for Cytometry in Tumours) technique estimates non-invasively cancer microstructure features. The clinical application of VERDICT for prostate cancer requires constraining some of the models parameter. This work uses the Accelerated Microstructure Imaging via Convex Optimization (AMICO) formulation for VERDICT (VERDICT-AMICO), to investigate parameter estimation for prostate tissue, in an attempt to minimize the parameter constraints. We examine various dictionaries for VERDICT-AMICO enabling different levels of flexibility on the choice of parameter values. Depending on the stability of the fitting this procedure leads to the selection of a dictionary (or dictionaries) with the fewest number of model parameter constraints. Results show that with the current VERDICT imaging acquisition, the model can have an extra free parameter to fit, the extracellular diffusivity. In conclusion, the AMICO adaptation for VERDICT allowed testing of different values for the previously fixed model parameters, and helped relax assumptions of fixed extracellular diffusivity that the model currently uses for prostate cancer characterisation.

Abstract. The benefits of deep neural networks can be hard to realise in medical imaging tasks because training sample sizes are often modest. Pre-training on large data sets and subsequent transfer learning to specific tasks with limited labelled training data has proved a successful strategy in other domains. Here, we implement and test this idea for detecting and classifying nuclei in histology, important tasks that enable quantifiable characterisation of prostate cancer. We pre-train a convolutional neural network for nucleus detection on a large colon histology dataset, and examine the effects of fine-tuning this network with different amounts of prostate histology data. Results show promise for clinical translation. However, we find that transfer learning is not always a viable option when training deep neural networks for nucleus classification. As such, we also demonstrate that semi-supervised ladder networks are a suitable alternative for learning a nucleus classifier with limited data.