IPMN-LEARN: A linear support vector machine learning model for predicting low-grade intraductal papillary mucinous neoplasms.

Backgrounds/Aims: We aimed to build a machine learning tool to help predict low-grade intraductal papillary mucinous neoplasms (IPMNs) in order to avoid unnecessary surgical resection. IPMNs are precursors to pancreatic cancer. Surgical resection remains the only recognized treatment for IPMNs yet carries some risks of morbidity and potential mortality. Existing clinical guidelines are imperfect in distinguishing low-risk cysts from high-risk cysts that warrant resection. Methods: We built a linear support vector machine (SVM) learning model using a prospectively maintained surgical database of patients with resected IPMNs. Input variables included 18 demographic, clinical, and imaging characteristics. The outcome variable was the presence of low-grade or high-grade IPMN based on post-operative pathology results. Data were divided into a training/validation set and a testing set at a ratio of 4:1. Receiver operating characteristics analysis was used to assess classification performance. Results: A total of 575 patients with resected IPMNs were identified. Of them, 53.4% had low-grade disease on final pathology. After classifier training and testing, a linear SVM-based model (IPMN-LEARN) was applied on the validation set. It achieved an accuracy of 77.4%, with a positive predictive value of 83%, a specificity of 72%, and a sensitivity of 83% in predicting low-grade disease in patients with IPMN. The model predicted low-grade lesions with an area under the curve of 0.82. Conclusions: A linear SVM learning model can identify low-grade IPMNs with good sensitivity and specificity. It may be used as a complement to existing guidelines to identify patients who could avoid unnecessary surgical resection.

Thymic Epithelial Neoplasms: Radiologic-Pathologic Correlation.

Thymic epithelial neoplasms, as classified by the World Health Organization, include thymoma, thymic carcinoma, and thymic carcinoid. They are a rare group of tumors and are often diagnosed incidentally in the work-up of parathymic syndrome, such as myasthenia gravis, or when mass effect or local invasion causes other symptoms. In each of these scenarios, understanding the radiologic-pathologic relationship of these tumors allows clinical imagers to contribute meaningfully to management decisions and overall patient care. Integrating important imaging features, such as local invasion, and pathologic features, such as necrosis and immunohistochemistry, ensures a meaningful contribution by clinical imagers to the care team.

The metrial gland in the rat and its similarities to granular cell tumors.

Metrial glands are normal structures located in the mesometrial triangle of the pregnant rat uterus from gestational day (GD) 8 through termination of pregnancy. Metrial glands are composed of a dynamic mixed cell population of granulated metrial gland (GMG) cells, endometrial stromal cells, trophoblasts, blood vessels, and fibroblasts. Collections of similar cells may be seen in association with pseudopregnancy and other hormonal disturbances. Granulated metrial gland cells are the hallmark cell of the metrial gland. They are bone-marrow-derived, perforin-positive, natural killer cells that proliferate in the pregnant uterus. Understanding the normal histogenesis of the metrial gland and recognizing the possible existence of GMG cells and a reactive metrial gland in the nonpregnant state are important when examining any uterine lesion that contains granulated cells. This report demonstrates that the cellular composition, morphology, and immunohistochemical staining profile of normal metrial glands are similar to reported granular cell neoplasms in rats and mice. The possibility of a non-neoplastic lesion involving the metrial gland should be considered when proliferative lesions involving granulated cells are observed in the uterus of mice and rats from nonclinical toxicity studies. Positive immunohistochemical staining for perforin and S100 would assist in the classification of such lesions as a reactive metrial gland or decidual reaction.

Ovarian follicle counts using proliferating cell nuclear antigen (PCNA) and semi-automated image analysis in rats.

Ovarian follicle counting is a method to assess ovarian toxicity in reproductive toxicity studies in rats. Although ovarian follicle counting has been traditionally performed manually on hematoxylin and eosin (H&E)-stained sections, the use of immunohistochemical methods, including human cytochrome P450 1B1 (CYP1B1) and proliferating cell nuclear antigen (PCNA), have been used to enhance the visibility of the primordial and primary follicles to facilitate manual counting. In this study, serial sections from both ovaries from ten 3-month-old female Sprague Dawley rats were stained using routine H&E and immunohistochemistry for PCNA. Counting of primordial and primary follicles was performed manually using these two stains and by semi-automated image analysis of PCNA-stained slides. Although manual counting of PCNA-stained slides is preferable to manual counting of H&E-stained slides, manual counting involves variability between individual counters. Semi-automated image analysis of PCNA-stained slides yields an accurate and consistent count of these primordial/primary follicles and eliminates variability between individual counters.