Daniel H. Turnbull

The segmentation of the brain ventricle (BV) and body in embryonic mice high-frequency ultrasound (HFU) volumes can provide useful information for biological researchers. However, manual segmentation of the BV and body requires substantial time and expertise. This work proposes a novel deep learning based end-to-end auto-context refinement framework, consisting of two stages. The first stage produces a low resolution segmentation of the BV and body simultaneously. The resulting probability map for each object (BV or body) is then used to crop a region of interest (ROI) around the target object in both the original image and the probability map to provide context to the refinement segmentation network. Joint training of the two stages provides significant improvement in Dice Similarity Coefficient (DSC) over using only the first stage (0.818 to 0.906 for the BV, and 0.919 to 0.934 for the body). The proposed method significantly reduces the inference time (102.36 to 0.09 s/volume ≈1000x faster) while slightly improves the segmentation accuracy over the previous methods using slide-window approaches.

Medulloblastoma (MB), the most common malignant pediatric brain tumor, is a classic example of dysregulation of developmental pathways leading to tumorogenesis. Despite advancements in multi-modal therapies, most patients suffer from long-term neurocognitive and neuroendocrine disabilities. The Sonic Hedgehog subgroup of MB (SHH-MB) accounts for ~30% of all cases and originates from ATOH1+ cerebellar granule cell precursors (GCPs). Experimental data in mice has shown that activating mutations in the SHH pathway induce tumors only in rare GCPs, suggesting that additional mutations and epigenetic changes are required to influence tumor progression. The KMT2D gene, encoding the histone-lysine N-methyltransferase 2D, is amongst the ten most frequently mutated genes in MB, with somatic mutations seen in ~15% of all SHH-MB patients. We developed sporadic mouse models of SHH-MB with a low penetrance to enable studies of secondary mutations (Tan, PNAS, 2018). Immunofluorescence staining for KMT2D on early-stage SHH-MB lesions, mid-stage and late-stage tumors revealed that a subset of lesions/tumors (16/98) do not express KMT2D and are negative for H3K4me3. Interestingly, P53 and KMT2D expression showed a positive correlation in ~94% of tumors/lesions and NeuN and KMT2D showed a positive correlation in ~92% of tumors/lesions. In order to determine the roles for KMT2D in GCP proliferation and differentiation, and uncover whether and how KMT2D promotes SHH-MB tumorigenesis, we are using genetic mouse-models whereby Kmt2d is heterozygously or homozygously deleted alone, or in conjunction with activation of the SHH pathway. Mice with SHH-MB tumors expressing SmoM2 and a loss of Kmt2d develop aggressive tumors at high penetrance, with metastatic leptomeningeal spread in the brain stem and spinal cord. Thus, loss of Kmt2d increases SHH-MB tumor progression and leads to malignancy. Ongoing studies are determining how the chromatin landscape and gene expression are changed when Kmt2d is deleted in GCPs.

High-frequency ultrasound (HFU) is well suited for imaging embryonic mice in vivo because it is non-invasive and real-time. Manual segmentation of the brain ventricles (BVs) and whole body from 3D HFU images is time-consuming and requires specialized training. This paper presents a deep-learning-based segmentation pipeline which automates several time-consuming, repetitive tasks currently performed to study genetic mutations in developing mouse embryos. Namely, the pipeline accurately segments the BV and body regions in 3D HFU images of mouse embryos, despite significant challenges due to position and shape variation of the embryos, as well as imaging artifacts. Based on the BV segmentation, a 3D convolutional neural network (CNN) is further trained to detect embryos with the Engrailed-1 (En1) mutation. The algorithms achieve 0.896 and 0.925 Dice Similarity Coefficient (DSC) for BV and body segmentation, respectively, and 95.8% accuracy on mutant classification. Through gradient based interrogation and visualization of the trained classifier, it is demonstrated that the model focuses on the morphological structures known to be affected by the En1 mutation.

Purpose: Genetically engineered mouse models of sporadic cancers are critical for studying tumor biology and for preclinical testing of therapeutics. We present an MRI‐based pipeline designed to produce high resolution, quantitative information about tumor progression and response to novel therapies in mouse models of medulloblastoma (MB).Methods: Sporadic MB was modeled in mice by inducing expression of an activated form of the Smoothened gene (aSmo) in a small number of cerebellar granule cell precursors. aSmo mice were imaged and analyzed at defined time‐points using a 3D manganese‐enhanced MRI‐based pipeline optimized for high‐throughput.Results: A semi‐automated segmentation protocol was established that estimates tumor volume in a time‐frame compatible with a high‐throughput pipeline. Both an empirical, volume‐based classifier and a linear discriminant analysis‐based classifier were tested to distinguish progressing from nonprogressing lesions at early stages of tumorigenesis. Tumor centroids measured at early stages revealed that there is a very specific location of the probable origin of the aSmo MB tumors. The efficacy of the manganese‐enhanced MRI pipeline was demonstrated with a small‐scale experimental drug trial designed to reduce the number of tumor associated macrophages and microglia.Conclusion: Our results revealed a high level of heterogeneity between tumors within and between aSmo MB models, indicating that meaningful studies of sporadic tumor progression and response to therapy could not be conducted without an imaging‐based pipeline approach.

Models based in differential expansion of elastic material, axonal constraints, directed growth, or multi-phasic combinations have been proposed to explain brain folding. However, the cellular and physical processes present during folding have not been defined. We used the murine cerebellum to challenge folding models with in vivo data. We show that at folding initiation differential expansion is created by the outer layer of proliferating progenitors expanding faster than the core. However, the stiffness differential, compressive forces, and emergent thickness variations required by elastic material models are not present. We find that folding occurs without an obvious cellular pre-pattern, that the outer layer expansion is uniform and fluid-like, and that the cerebellum is under radial and circumferential constraints. Lastly, we find that a multi-phase model incorporating differential expansion of a fluid outer layer and radial and circumferential constraints approximates the in vivo shape evolution observed during initiation of cerebellar folding.

Volumetric analysis of brain ventricle (BV) structure is a key tool in the study of central nervous system development in embryonic mice. High-frequency ultrasound (HFU) is the only non-invasive, real-time modality available for rapid volumetric imaging of embryos in utero. However, manual segmentation of the BV from HFU volumes is tedious, time-consuming, and requires specialized expertise. In this paper, we propose a novel deep learning based BV segmentation system for whole-body HFU images of mouse embryos. Our fully automated system consists of two modules: localization and segmentation. It first applies a volumetric convolutional neural network on a 3D sliding window over the entire volume to identify a 3D bounding box containing the entire BV. It then employs a fully convolutional network to segment the detected bounding box into BV and background. The system achieves a Dice Similarity Coefficient (DSC) of 0.8956 for BV segmentation on an unseen 111 HFU volume test set surpassing the previous state-of-the-art method (DSC of 0.7119) by a margin of 25%.