Medical Student Research Forum Posters

Neurofibromatosis type 1 (NF1) is an inherited tumor syndrome caused by heterozygous germline mutations in the NF1 gene, occurring in approximately 1/2600 individuals. A subset of patients with neurofibromatosis type 1 (NF1) develop juvenile xanthogranulomas (JXGs), a non-Langerhans cell histiocytosis, and some of these patients also develop juvenile myelomonocytic leukemia (JMML).Yet, these associations are poorly delineated.JXG is a benign proliferation of non-Langerhans cells histiocytes characterized by small yellow/brown papulonodules ranging from 1-20 mm in size. JMML is a mixed myeloproliferative-myelodysplastic disorder that affects children, most often before age 6.4. The first and only systematic review on this described therisk of developing JMML 20 to 30 times higher in patients with NF1 with JXG lesions compared to those without JXG. Since then, mostly isolated case reports have either refuted or confirmed this triple association.

Cutaneous melanoma results from the malignant transformation of melanocytes in skin and accounts for 75% of deaths related to skin cancer. Between 5 to 12% of melanoma cases can be attributed to hereditary melanoma, melanoma caused by inherited germline mutations in melanoma predisposition genes. Although several inherited melanoma predisposition syndromes have been identified, a subset of melanoma families lack pathogenic mutations in known highly penetrant predisposition genes, including CDKN2A, CDK4, and BAP1. Here we report the case of a woman with a history of melanoma and multiple primary tumors with one pathogenic germline mutation in BUB1B

Early diagnosis of melanoma is critical for improved survival as melanoma is the deadliest of the common forms of skin cancer. The gold standard for the diagnosis of melanoma is a biopsy followed by histopathological analysis. Melanocytic nevi, which are very common benign neoplasms of the melanocytes, are often biopsied because they are mimics and precursor for melanoma. Annually, 4.5 million pigmented lesions are biopsied in the United States. A subset of melanoma is difficult to distinguish from melanocytic nevi, resulting in diagnostic errors and worsened patient outcomes. Therefore, improved diagnostic tests, including the utilization of novel biomarkers, are being developed to improve clinical and histological diagnostic accuracy of melanoma. Moreover, non-invasive tests prior to surgical biopsy have been introduced, including an adhesive patch-based assay that tests the expression of melanoma biomarkers PRAME and noncoding long RNA LINC00518. Our prior work identified that S100A8, a member of the calcium-binding S100 family, is differentially expressed in melanomas versus nevi4 . Specifically, S100A8 is expressed by the keratinocyte microenvironment of melanomas but not nevi. S100A8 is a melanoma biomarker of interest for an adhesive patch-based assay, because it is expressed in the epidermis, the most superficial layer of the skin.

Machine learning (ML) is an application of computational and statistical techniques to allow computers to learn and predict without explicit programming. In recent years, with the increasing availability of large scale and low-cost computing power, ML capacity has expanded vastly and has begun to change how many industries operate. The ability of machines to analyze large, complex datasets and to detect patterns beyond the scope of the human mind provides a powerful opportunity for application in a healthcare setting. ML has introduced new approaches to many dimensions of medicine including, but not limited to, Pathology, Radiology, drug development, enhancing existing clinical predictive tools, and the management of many diseases including cancer and autoimmune diseases. Currently, ML remains in its infancy but has already started to make an impact in various healthcare disciplines. This research project aimed to provide the foundational training and understanding of the modern approaches to ML and develop the skill set necessary to use available healthcare data to develop and deploy new ML models to assist in the delivery of future healthcare.

Introduction: In the era of molecular diagnostics and personalized medicine, it is becoming increasingly important to save tissue for downstream testing for optimal pathologic diagnosis. Unfortunately, conventional histology processing and its expenditure of tissue for H&E imaging often results in inadequate material for essential molecular tests downstream. Microscopy Using Ultraviolet Excitation (MUSE) has emerged as a promising potential answer in providing a novel tissuesparing method of generating morphologic imaging without the need to fix or cut fresh tissue. We aim to standardize protocols for imaging an array of CNS tumor samples and demonstrate equivalency to traditional FFPE H&E in terms of generating images for tumor diagnostics.

Materials and Methods: 24 CNS tumor biopsy specimens were imaged using the MUSE interface, then subsequently fixed and paraffinembedded for traditional H&E staining. Each pair of slides (MUSE and H&E) were then read by a panel of 4 neuropathologists, and the diagnosis by each reader was recorded as correct or wrong. Combined accuracy was calculated within each diagnosis category and for each pathologist.

Results: In surgical resections of 24 adult patients (mean age 54 years) with newly diagnosed brain and spinal cord tumors, 7/24 were diagnosed by conventional methodology with diffuse astrocytic/oligodendroglial tumors, 8/24 with meningiomas, 3/24 with ependymal/choroid plexus tumors, 3/24 with tumors of cranial/paraspinal nerves, and 3/24 with metastatic tumors. 97% concordance was observed among MUSE versus light microscopy diagnostics, with 94% within the pathologist panel.

Conclusions: MUSE imaging appears to have been successful in reliably generating diagnostic-quality histological images of CNS tumors. This is supported by inter-pathologist concordance on diagnoses made through both MUSE and traditional H&E images. Ongoing studies are expected to expand to assessments of grading MUSE images of more diagnostically difficult brain and spinal cord tumors.