Atypical teratoid rhabdoid tumors (ATRTs) are rare central nervous system tumors that comprise approximately 1–2% of all pediatric brain tumors; however, in patients less than 3 years of age this tumor accounts for up to 20% of cases. ATRT is characterized by loss of the long arm of chromosome 22 which results in loss of the hSNF5/INI-1 gene. INI1, a member of the SWI/SNF chromatin remodeling complex, is important in maintenance of the mitotic spindle and cell cycle control. Overall survival in ATRT is poor with median survival around 17 months. Radiation is an effective component of therapy but is avoided in patients younger than 3 years of age due to long term neurocognitive sequelae. Most long term survivors undergo radiation therapy as a part of their upfront or salvage therapy, and there is a suggestion that sequencing the radiation earlier in therapy may improve outcome. There is no standard curative chemotherapeutic regimen, but anecdotal reports advocate the use of intensive therapy with alkylating agents, high-dose methotrexate, or therapy that includes high-dose chemotherapy with stem cell rescue. Due to the rarity of this tumor and the lack of randomized controlled trials it has been challenging to define optimal therapy and advance treatment. Recent laboratory investigations have identified aberrant function and/or regulation of cyclin D1, aurora kinase, and insulin-like growth factor pathways in ATRT. There has been significant interest in identifying and testing therapeutic agents that target these pathways.

Deep learning (DL) is a widely applied mathematical modeling technique. Classically, DL models utilize large volumes of training data, which are not available in many healthcare contexts. For patients with brain tumors, non-invasive diagnosis would represent a substantial clinical advance, potentially sparing patients from the risks associated with surgical intervention on the brain. Such an approach will depend upon highly accurate models built using the limited datasets that are available. Herein, we present a novel genetic algorithm (GA) that identifies optimal architecture parameters using feature embeddings from state-of-the-art image classification networks to identify the pediatric brain tumor, adamantinomatous craniopharyngioma (ACP). We optimized classification models for preoperative Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and combined CT and MRI datasets with demonstrated test accuracies of 85.3%, 83.3%, and 87.8%, respectively. Notably, our GA improved baseline model performance by up to 38%. This work advances DL and its applications within healthcare by identifying optimized networks in small-scale data contexts. The proposed system is easily implementable and scalable for non-invasive computer-aided diagnosis, even for uncommon diseases.

Head Start strategies may produce long-term remission in young children with newly diagnosed CPC with avoidance of cranial irradiation.

Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Medulloblastoma (MDB) represents a major form of malignant brain tumors in the pediatric population. A vast spectrum of research on MDB has advanced our understanding of the underlying mechanism, however, a significant need still exists to develop novel therapeutics on the basis of gaining new knowledge about the characteristics of cell signaling networks involved. The Ras signaling pathway, one of the most important proto-oncogenic pathways involved in human cancers, has been shown to be involved in the development of neurological malignancies. We have studied an important effector down-stream of Ras, namely RalA (Ras-Like), for the first time and revealed overactivation of RalA in MDB. Affinity precipitation analysis of active RalA (RalA-GTP) in eight MDB cell lines (DAOY, RES256, RES262, UW228-1, UW426, UW473, D283 and D425) revealed that the majority contained elevated levels of active RalA (RalA-GTP) as compared with fetal cerebellar tissue as a normal control. Additionally, total RalA levels were shown to be elevated in 20 MDB patient samples as compared to normal brain tissue. The overall expression of RalA, however, was comparable in cancerous and normal samples. Other important effectors of RalA pathway including RalA binding protein-1 (RalBP1) and protein phosphatase A (PP2A) down-stream of Ral and Aurora kinase A (AKA) as an upstream RalA activator were also investigated in MDB. Considering the lack of specific inhibitors for RalA, we used gene specific silencing in order to inhibit RalA expression. Using a lentivirus expressing anti-RalA shRNA we successfully inhibited RalA expression in MDB and observed a significant reduction in proliferation and invasiveness. Similar results were observed using inhibitors of AKA and geranyl-geranyl transferase (non-specific inhibitors of RalA signaling) in terms of loss of in vivo tumorigenicity in heterotopic nude mouse model. Finally, once tested in cells expressing CD133 (a marker for MDB cancer stem cells), higher levels of RalA activation was observed. These data not only bring RalA to light as an important contributor to the malignant phenotype of MDB but introduces this pathway as a novel target in the treatment of this malignancy.