Automated Detection of Brain Metastases on T1-Weighted MRI Using a Convolutional Neural Network: Impact of Volume Aware Loss and Sampling Strategy.

BACKGROUND: Detection of brain metastases (BM) and segmentation for treatment planning could be optimized with machine learning methods. Convolutional neural networks (CNNs) are promising, but their trade-offs between sensitivity and precision frequently lead to missing small lesions. HYPOTHESIS: Combining volume aware (VA) loss function and sampling strategy could improve BM detection sensitivity. STUDY TYPE: Retrospective. POPULATION: A total of 530 radiation oncology patients (55% women) were split into a training/validation set (433 patients/1460 BM) and an independent test set (97 patients/296 BM). FIELD STRENGTH/SEQUENCE: 1.5 T and 3 T, contrast-enhanced three-dimensional (3D) T1-weighted fast gradient echo sequences. ASSESSMENT: Ground truth masks were based on radiotherapy treatment planning contours reviewed by experts. A U-Net inspired model was trained. Three loss functions (Dice, Dice + boundary, and VA) and two sampling methods (label and VA) were compared. Results were reported with Dice scores, volumetric error, lesion detection sensitivity, and precision. A detected voxel within the ground truth constituted a true positive. STATISTICAL TESTS: McNemar's exact test to compare detected lesions between models. Pearson's correlation coefficient and Bland-Altman analysis to compare volume agreement between predicted and ground truth volumes. Statistical significance was set at P ≤ 0.05. RESULTS: Combining VA loss and VA sampling performed best with an overall sensitivity of 91% and precision of 81%. For BM in the 2.5-6 mm estimated sphere diameter range, VA loss reduced false negatives by 58% and VA sampling reduced it further by 30%. In the same range, the boundary loss achieved the highest precision at 81%, but a low sensitivity (24%) and a 31% Dice loss. DATA CONCLUSION: Considering BM size in the loss and sampling function of CNN may increase the detection sensitivity regarding small BM. Our pipeline relying on a single contrast-enhanced T1-weighted MRI sequence could reach a detection sensitivity of 91%, with an average of only 0.66 false positives per scan. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Stereotyped spatial patterns of functional synaptic connectivity in the cerebellar cortex.

Motor coordination is supported by an array of highly organized heterogeneous modules in the cerebellum. How incoming sensorimotor information is channeled and communicated between these anatomical modules is still poorly understood. In this study, we used transgenic mice expressing GFP in specific subsets of Purkinje cells that allowed us to target a given set of cerebellar modules. Combining in vitro recordings and photostimulation, we identified stereotyped patterns of functional synaptic organization between the granule cell layer and its main targets, the Purkinje cells, Golgi cells and molecular layer interneurons. Each type of connection displayed position-specific patterns of granule cell synaptic inputs that do not strictly match with anatomical boundaries but connect distant cortical modules. Although these patterns can be adjusted by activity-dependent processes, they were found to be consistent and predictable between animals. Our results highlight the operational rules underlying communication between modules in the cerebellar cortex.

Limited changes in spinal lamina I dorsal horn neurons following the cytotoxic ablation of non-peptidergic C-fibers.

BACKGROUND: Non-peptidergic nociceptive neurons are a sub-population of small diameter primary sensory neurons that comprise approximately 50 % of the C fiber population. Together with the peptidergic sub-population, they transmit nociceptive information from the periphery to the superficial dorsal horn of the spinal cord. Despite the numerous studies investigating the role of the non-peptidergic primary afferents, their role in normal nociception and in pain remains poorly understood. Our lab has previously demonstrated that, in rat models of neuropathic and inflammatory pain, there is a de novo expression of substance P receptors (NK-1r) by lamina I pyramidal projection neurons, a neuronal population that normally does not express these receptors. RESULTS: In this study, we used a ribosomal toxin, saporin, conjugated to the lectin IB4 to selectively ablate the non-peptidergic nociceptive C fibers, to investigate if the loss of these fibers was enough to induce a change in NK-1r expression by lamina I projection neurons. IB4-saporin treatment led to the permanent ablation of the IB4-positive afferents but also to a small non-significant reduction in CGRP-positive afferents. An overall increase in immunoreactivity for the NK-1r was observed in lamina I projection neurons, however, the lack of non-peptidergic afferents did not increase the number of lamina I pyramidal projection neurons immunoreactive for the receptor. CONCLUSIONS: Our results demonstrate that the deletion of the non-peptidergic afferents, at the L4-L5 spinal levels, is not sufficient to trigger the de novo expression of NK-1r by projection pyramidal neurons but increases the expression of NK-1r in fusiform and multipolar projection neurons. Furthermore, our data suggest that a neuropathic component is essential to trigger the expression of NK-1r by pyramidal neurons.

gamma-Aminobutyric acid cells with cocaine-induced DeltaFosB in the ventral tegmental area innervate mesolimbic neurons.

BACKGROUND: The transcription factor DeltaFosB is implicated in the plasticity induced by drugs of abuse. We showed that psychostimulants induce DeltaFosB in gamma-aminobutyric acid (GABA) cells of a caudal subregion of the ventral tegmental area (VTA) that was named tail of the VTA (tVTA). Although tVTA mostly shares VTA inputs, its outputs remain to be characterized. METHODS: The tVTA efferents were studied by iontophoretic injections of the anterograde tracer biotinylated dextran amine (BDA). To further study VTA inputs arising from tVTA, injections of the retrograde tracer Fluoro-Gold were combined with multiple labeling by immunohistochemistry in rats treated with cocaine. Indirect projections from the tVTA to the nucleus accumbens were assessed with a double-tracing approach, cholera toxin B subunit (CTB) being delivered in the nucleus accumbens and BDA in the tVTA. RESULTS: Tract-tracing studies showed that tVTA heavily projects to the midbrain dopaminergic system and revealed terminal appositions with dopamine cells in the VTA. Double-labeling studies demonstrated that this tVTA output is mostly GABAergic, includes cells in which cocaine exposure induces DeltaFosB, and displays appositions to dopamine cells projecting to the nucleus accumbens. CONCLUSIONS: The GABA neurons expressing DeltaFosB in the tVTA after cocaine exposure project to the dopamine mesolimbic neurons.

Afferents to the GABAergic tail of the ventral tegmental area in the rat.

We previously showed that chronic psychostimulant exposure induces the transcription factor DeltaFosB in gamma-aminobutyric acid (GABA)ergic neurons of the caudal tier of the ventral tegmental area (VTA). This subregion was defined as the tail of the VTA (tVTA). In the present study, we showed that tVTA can also be visualized by analyzing FosB/DeltaFosB response following acute cocaine injection. This induction occurs in GABAergic neurons, as identified by glutamic acid decarboxylase (GAD) expression. To characterize tVTA further, we mapped its inputs by using the retrograde tracers Fluoro-Gold or cholera toxin B subunit. Retrogradely labeled neurons were observed in the medial prefrontal cortex, the lateral septum, the ventral pallidum, the bed nucleus of the stria terminalis, the substantia innominata, the medial and lateral preoptic areas, the lateral and dorsal hypothalamic areas, the lateral habenula, the intermediate layers of the superior colliculus, the dorsal raphe, the periaqueductal gray, and the mesencephalic and pontine reticular formation. Projections from the prefrontal cortex, the hypothalamus, and the lateral habenula to the tVTA were also shown by using the anterograde tracer biotinylated dextran amine (BDA). We showed that the central nucleus of the amygdala innervates the anterior extent of the VTA but not the tVTA. Moreover, the tVTA mainly receives non-aminergic inputs from the dorsal raphe and the locus coeruleus. Although the tVTA has a low density of dopaminergic neurons, its afferents are mostly similar to those targeting the rest of the VTA. This suggests that the tVTA can be considered as a VTA subregion despite its caudal location.