ISSLS PRIZE in Clinical Science 2023: comparison of degenerative MRI features of the intervertebral disc between those with and without chronic low back pain. An exploratory study of two large female populations using automated annotation.

OBJECTIVES: The relationship of degeneration to symptoms has been questioned. MRI detects apparently similar disc degeneration and degenerative changes in subjects both with and without back pain. We aimed to overcome these problems by re-annotating MRIs from asymptomatic and symptomatics groups onto the same grading system. METHODS: We analysed disc degeneration in pre-existing large MRI datasets. Their MRIs were all originally annotated on different scales. We re-annotated all MRIs independent of their initial grading system, using a verified, rapid automated MRI annotation system (SpineNet) which reported degeneration on the Pfirrmann (1-5) scale, and other degenerative features (herniation, endplate defects, marrow signs, spinal stenosis) as binary present/absent. We compared prevalence of degenerative features between symptomatics and asymptomatics. RESULTS: Pfirrmann degeneration grades in relation to age and spinal level were very similar for the two independent groups of symptomatics over all ages and spinal levels. Severe degenerative changes were significantly more prevalent in discs of symptomatics than asymptomatics in the caudal but not the rostral lumbar discs in subjects < 60 years. We found high co-existence of degenerative features in both populations. Degeneration was minimal in around 30% of symptomatics < 50 years. CONCLUSIONS: We confirmed age and disc level are significant in determining imaging differences between asymptomatic and symptomatic populations and should not be ignored. Automated analysis, by rapidly combining and comparing data from existing groups with MRIs and information on LBP, provides a way in which epidemiological and 'big data' analysis could be advanced without the expense of collecting new groups. LEVEL OF EVIDENCE I: Diagnostic: individual cross-sectional studies with consistently applied reference standard and blinding.

Temporal models for mitotic phase labelling.

With the widespread use of time-lapse data to understand cellular function, there is a need for tools which facilitate high-throughput analysis of data. Fluorescence microscopy of genetically engineered cell lines in culture can be used to visualise the progression of these cells through the cell cycle, including distinctly identifiable sequential stages of cell division (mitotic phases). We present a system for automated segmentation and mitotic phase labelling using temporal models. This work takes the novel approach of using temporal features evaluated over the whole of the mitotic phases rather than over single frames, thereby capturing the distinctive behaviour over the phases. We compare and contrast three different temporal models: Dynamic Time Warping, Hidden Markov Models, and Semi Markov Models. A new loss function is proposed for the Semi Markov model to make it more robust to inconsistencies in data annotation near transition boundaries. The models are tested under two different experimental conditions to explore robustness to changes in biological conditions.

OBJCUT: efficient segmentation using top-down and bottom-up cues.

We present a probabilistic method for segmenting instances of a particular object category within an image. Our approach overcomes the deficiencies of previous segmentation techniques based on traditional grid conditional random fields (CRF), namely that 1) they require the user to provide seed pixels for the foreground and the background and 2) they provide a poor prior for specific shapes due to the small neighborhood size of grid CRF. Specifically, we automatically obtain the pose of the object in a given image instead of relying on manual interaction. Furthermore, we employ a probabilistic model which includes shape potentials for the object to incorporate top-down information that is global across the image, in addition to the grid clique potentials which provide the bottom-up information used in previous approaches. The shape potentials are provided by the pose of the object obtained using an object category model. We represent articulated object categories using a novel layered pictorial structures model. Nonarticulated object categories are modeled using a set of exemplars. These object category models have the advantage that they can handle large intraclass shape, appearance, and spatial variation. We develop an efficient method, OBJCUT, to obtain segmentations using our probabilistic framework. Novel aspects of this method include: 1) efficient algorithms for sampling the object category models of our choice and 2) the observation that a sampling-based approximation of the expected log-likelihood of the model can be increased by a single graph cut. Results are presented on several articulated (e.g., animals) and nonarticulated (e.g., fruits) object categories. We provide a favorable comparison of our method with the state of the art in object category specific image segmentation, specifically the methods of Leibe and Schiele and Schoenemann and Cremers.

Segmentation and measurement of brain structures in MRI including confidence bounds.

The advent of new and improved imaging devices has allowed an impressive increase in the accuracy and precision of MRI acquisitions. However, the volumetric nature of the image formation process implies an inherent uncertainty, known as the partial volume effect, which can be further affected by artifacts such as magnetic inhomogeneities and noise. These degradations seriously challenge the application to MRI of any segmentation method, especially on data sets where the size of the object or effect to be studied is small relative to the voxel size, as is the case in multiple sclerosis and schizophrenia. We develop an approach to this problem by estimating a set of bounds on the spatial location of each organ to be segmented. First, we describe a method for 3D segmentation from voxel data which combines statistical classification and geometry-driven segmentation; then we discuss how the partial volume effect is estimated and object measurements are obtained. A comprehensive validation study and a set of results on clinical applications are also described.