ReCasNet: Improving consistency within the two-stage mitosis detection framework.

Mitotic count (MC) is an important histological parameter for cancer diagnosis and grading, but the manual process for obtaining MC from whole-slide histopathological images is very time-consuming and prone to error. Therefore, deep learning models have been proposed to facilitate this process. Existing approaches utilize a two-stage pipeline: the detection stage for identifying the locations of potential mitotic cells and the classification stage for refining prediction confidences. However, this pipeline formulation can lead to inconsistencies in the classification stage due to the poor prediction quality of the detection stage and the mismatches in training data distributions between the two stages. In this study, we propose a Refine Cascade Network (ReCasNet), an enhanced deep learning pipeline that mitigates the aforementioned problems with three improvements. First, window relocation was used to reduce the number of poor quality false positives generated during the detection stage. Second, object re-cropping was performed with another deep learning model to adjust poorly centered objects. Third, improved data selection strategies were introduced during the classification stage to reduce the mismatches in training data distributions. ReCasNet was evaluated on two large-scale mitotic figure recognition datasets, canine cutaneous mast cell tumor (CCMCT) and canine mammary carcinoma (CMC), which resulted in up to 4.8% percentage point improvements in the F1 scores for mitotic cell detection and 44.1% reductions in mean absolute percentage error (MAPE) for MC prediction. Techniques that underlie ReCasNet can be generalized to other two-stage object detection pipeline and should contribute to improving the performances of deep learning models in broad digital pathology applications.

Postoperative radiotherapy timing, molecular subgroups and treatment outcomes of Thai pediatric patients with medulloblastoma.

INTRODUCTION: Medulloblastoma (MB) is the most common childhood malignant brain tumor worldwide. Recently, molecular classification was established and started to play a role in the management of MB; however, studies involving molecular defined MB in Southeast Asia have been limited. We aimed to describe, and correlate clinical characteristics and molecular subgroups with therapeutic outcomes of Thai pediatric patients with MB. MATERIALS AND METHODS: Pediatric MB patients treated at King Chulalongkorn Memorial Hospital in Thailand from 2006 to 2018 were recruited. Patients were classified by clinical characteristics into standard- and high-risk groups, which determined treatment regimen. Retrospectively, available tumor tissues were classified into 3 molecular subgroups using immunohistochemistry: 1) WNT, 2) SHH, and 3) non-WNT/non-SHH. The primary outcome was 5-year overall survival (OS). Risk factors associated with OS were analyzed using cox regression analysis. RESULTS: Fifty-three Thai pediatric patients with MB were enrolled. The median follow-up time was 60 months. The 5-year OS for all patients, and patients with standard-risk and high-risk were 74.2%, 76.3% and 71.4%, respectively. Tumor tissues of 24 patients were available, of which 23 could be molecularly classified. Two, one and 20 were in the WNT, SHH and non-WNT/non-SHH subtypes with 5-year OS of 100%, 100% and 78.9%, respectively. Using multivariate analysis, the interval of more than 8 weeks between surgery and radiotherapy was significantly correlated with a decrease in the 5-year OS. CONCLUSION: Interval between surgery and radiotherapy within 8 weeks was associated with good therapeutic outcomes among Thai pediatric patients with MB. Simplified molecular subtyping combined with clinical characteristics is practical in risk classification of patients with MB in institutes with limited resources.

KRAS Mutation in Pediatric Intracranial Germ Cell Tumors.

BACKGROUND: Intracranial germ cell tumors (IGCTs) are rare, highly curable neoplasms. KRAS is a gene in the KIT/RAS signaling pathway, and KRAS mutations have been reported in patients diagnosed with IGCTs. OBJECTIVES: To describe the clinicopathologic and molecular features of KRAS mutation and the treatment outcome of children diagnosed with IGCTs. METHODS: Patients diagnosed with IGCTs at the Department of Pediatrics, King Chulalongkorn Memorial Hospital from 2007 to 2016 were retrospectively reviewed. DNA was extracted from formalin-fixed, paraffin-embedded tissue and used for molecular study. Mutations in codons 12, 13, and 61 of the KRAS gene were detected using the cobas® KRAS mutation test and pyrosequencing. RESULTS: Eighteen patients were diagnosed with IGCTs (11 males and 7 females): nine with germinomas and nine with non-germinomatous GCTs (NGGCTs). The age range of the patients was 5-14 years (median 10.5 years). Elevated markers were revealed in approximately 25% of the patients. Four patients (two with germinomas and two with NGGCTs) had leptomeningeal involvement. All patients underwent tumor biopsy and received neoadjuvant chemotherapy. Radiotherapy was administered in 16 patients, and craniospinal radiation was administered only in patients with leptomeningeal metastasis. With a median follow-up of 26 months, overall survival was 88.9% in the patients with germinomas and 37% in the patients with NGGCTs. Mutation of the KRAS gene was detected using pyrosequencing in one patient. The mutation located at codon 61, with frequency 38.3% units, nucleotide substitution CAA > CTA, and amino acid substitution, was Q61L. The patient carrying the mutant gene was diagnosed with germinoma with cerebrospinal fluid metastasis and eventually died from treatment-related toxicity. CONCLUSION: Our study revealed the treatment outcomes of IGCTs in Thai children. The metastatic germinoma patient with KRAS codon 61 mutation had a poor outcome, supporting that Q61L has a clinical correlation with IGCTs.

Pediatric fibromyxoid tumor with PLAG1 fusion: An emerging entity with a novel intracranial location.

Translocations involving PLAG1 occur in several tumors, most commonly pleomorphic adenoma and lipoblastoma. Recently, a distinctive soft tissue tumor with a PLAG1 fusion has been reported in the pediatric age group. These are low grade tumors with a fibroblastic or mixed fibroblastic and myxoid morphology but no other lines of differentiation. They are typically immunopositive for desmin and CD34. The partner genes for these tumors have included YWHAZ, EEF1A1, ZFHX4l, CHCHD7, and PCMTD1. We report another case of this fibromyxoid tumor with a PLAG1 fusion, this time with COL3A1 as the partner gene. The fusion placed expression of a full-length PLAG1 protein under the control of the constitutively active COL3A1 promoter. Overexpression of PLAG1 was confirmed by diffusely positive immunostaining for PLAG1. The most novel aspect of this tumor is the intracranial location. Opinion has been divided over whether these tumors are a specific entity, or related to lipoblastoma, since that tumor also typically occurs in soft tissue in the pediatric age group and shows many of the same gene fusions. However, lipoblastoma has never been reported in an intracranial location and, thus, our case provides compelling evidence that this fibromyxoid tumor is indeed a distinct entity.

Idiopathic pericardial effusion in patients with hypertrophic cardiomyopathy.

The aims of this study were to examine the prevalence of moderate to large (moderate-large) idiopathic pericardial effusion (i-PEF) in patients with hypertrophic cardiomyopathy (HCM) and to identify clinical and echocardiographic hemodynamic profiles associated with pericardial effusion. A total of 292 adult patients with HCM were studied. Fifteen patients with a history of factors associated with pericardial effusion including myocardial infarction, heart surgery or cardiac procedure within the last 12 months, autoimmune disease, hydralazine use, chronic kidney disease stage 3-4, tuberculosis, and malignancy were excluded. Of 277 eligible patients with HCM, 11 patients (4%) with moderate-large i-PEF were identified. Clinical tamponade was present in 1 patient. Compared to patients with HCM who had no or small pericardial effusion, patients with moderate-large i-PEF were younger and more likely to have right ventricular (RV) hypertrophy and reverse septal curvature. These patients also exhibited a greater maximal septal thickness, mean and systolic pulmonary pressure, and right atrial pressure (p < 0.05 for all). Pericardial fluid analysis and histopathological exams were performed in 7 and 3 patients, respectively. All examinations revealed transudative and nonspecific etiology of pericardial effusion. Idiopathic pericardial effusion and cardiac tamponade in patients with HCM was uncommon. The pathophysiology involved in pericardial effusion remains undetermined. Patients with moderate-large i-PEF frequently exhibited a phenotype of pulmonary hypertension and RV pressure overload.

Simplified approach for pathological diagnosis of diffuse gliomas in adult patients.

The most recent WHO classification (2016) for gliomas introduced integrated diagnoses requiring both phenotypic and genotypic data. This approach presents difficulties for countries with limited resources for laboratory testing. The present study describes a series of 118 adult Thai patients with diffuse gliomas, classified by the WHO 2016 classification. The purpose was to demonstrate how a diagnosis can still be achieved using a simplified approach that combines clinical, morphological, immunohistochemical, and fewer molecular assays than typically performed. This algorithm starts with tumor location (midline vs. non-midline) with diffuse midline glioma identified by H3 K27M immunostaining. All other tumors are placed into one of 6 categories, based on morphologic features rather than specific diagnoses. Molecular testing is limited to IDH1/IDH2 mutations, plus co-deletion of 1p/19q for cases with oligodendroglial features and TERT promoter mutation for cases without such features. Additional testing for co-deletion of 1p/19q, TERT promoter mutation and BRAF mutations are only used in selected cases to refine diagnosis and prognosis. With this approach, we were able to reach the integrated diagnosis in 117/118 cases, saving 50 % of the costs of a more inclusive testing panel. The demographic data and tumor subtypes were found to be similar to series from other regions of the world. To the best of our knowledge, this is to the first reported series of diffuse gliomas in South-East Asia categorized by the WHO 2016 classification system.

Giant choroid plexus cysts with calvarial erosion: a case report and literature review.

Choroid plexus cysts rarely grow to be symptomatic. Few large choroid plexus cysts have been reported in the pediatric population. The authors report a 15-month-old boy with increased head circumference and a bony deformity in the left parietal region due to mass effect from a giant choroid plexus cyst. The child had a craniotomy for open resection of the cyst, and made an excellent recovery. The differential diagnosis for intraventricular cysts and the literature surrounding choroid plexus cysts are discussed.

MIB-1 Index as a Surrogate for Mitosis-Karyorrhexis Index in Neuroblastoma.

Neuroblastoma, the most common extracranial solid tumor in infancy, shows marked biological heterogeneity. Multiple prognostic markers are combined to risk-stratify neuroblastoma patients for treatment. One marker assesses histology, dividing patients into favorable and unfavorable categories based, in part, on the mitosis-karyorrhexis index (MKI). The recommended scoring of 5000 cells is, however, time-consuming and observer-dependent, and accurate counts may not always be performed. In the present study, we investigated using MIB-1 as a surrogate marker for the MKI. Twenty-five cases of neuroblastoma, ranging from low to high MKI, were immunostained for MIB-1. A total of 375 microscopic fields were digitally captured with > 100,000 cells scored. The MIB-1 index was determined by image analysis and MKI, by manual counting of the same immunostained fields. There was a significant correlation between the MIB-1 index and MKI comparing all fields (r = 0.7869, P < 0.01) and an even better correlation comparing individual cases (r = 0.9147, P < 0.01). Using a linear regression model, a formula was generated to calculate MKI from the MIB-1 index as follows: MKI = (MIB-1 index × 0.124) + 1.412. With this formula, a low MKI corresponds to an MIB-1 index < 4.74, intermediate MKI to an MIB-1 index of 4.74 to 20.87, and high MKI to an MIB-1 index > 20.87. For comparison, the calculations were repeated using a manual MIB-1 count on the same images. Similar significant correlations were obtained, with nearly identical cutoff values for MKI categories. This approach can facilitate determination of the MKI by assessing the MIB-1 index, either by image analysis or manual counting.