The ERK inhibitor LY3214996 augments anti-PD-1 immunotherapy in preclinical mouse models of BRAFV600E melanoma brain metastasis

BACKGROUND

Immune checkpoint inhibitors (ICI) have revolutionized cancer treatment; however, only a subset of patients with brain metastasis (BM) respond to ICI. Activating mutations in the mitogen-activated protein kinase signaling pathway are frequent in BM. The objective of this study was to evaluate whether therapeutic inhibition of extracellular signal-regulated kinase (ERK) can improve the efficacy of ICI for BM.

METHODS

We used immunotypical mouse models of BM bearing dual extracranial/intracranial tumors to evaluate the efficacy of single-agent and dual-agent treatment with selective ERK inhibitor LY3214996 (LY321) and anti-programmed death receptor 1 (PD-1) antibody. We verified target inhibition and drug delivery, then investigated treatment effects on T-cell response and tumor-immune microenvironment using high-parameter flow cytometry, multiplex immunoassays, and T-cell receptor profiling.

RESULTS

We found that dual treatment with LY321 and anti-PD-1 significantly improved overall survival in 2 BRAFV600E-mutant murine melanoma models but not in KRAS-mutant murine lung adenocarcinoma. We demonstrate that although LY321 has limited blood-brain barrier (BBB) permeability, combined LY321 and anti-PD-1 therapy increases tumor-infiltrating CD8+ effector T cells, broadens the T-cell receptor repertoire in the extracranial tumor, enriches T-cell clones shared by the periphery and brain, and reduces immunosuppressive cytokines and cell populations in tumors.

CONCLUSIONS

Despite the limited BBB permeability of LY321, combined LY321 and anti-PD-1 treatment can improve intracranial disease control by amplifying extracranial immune responses, highlighting the role of extracranial tumors in driving intracranial response to treatment. Combined ERK and PD-1 inhibition is a promising therapeutic approach, worthy of further investigation for patients with melanoma BM.

Region-specific DNA hydroxymethylation along the malignant progression of IDH-mutant gliomas

The majority of low-grade isocitrate dehydrogenase-mutant (IDHmt) gliomas undergo malignant progression (MP), but their underlying mechanism remains unclear. IDHmt gliomas exhibit global DNA methylation, and our previous report suggested that MP could be partly attributed to passive demethylation caused by accelerated cell cycles. However, during MP, there is also active demethylation mediated by ten-eleven translocation, such as DNA hydroxymethylation. Hydroxymethylation is reported to potentially contribute to gene expression regulation, but its role in MP remains under investigation. Therefore, we conducted a comprehensive analysis of hydroxymethylation during MP of IDHmt astrocytoma. Five primary/malignantly progressed IDHmt astrocytoma pairs were analyzed with oxidative bisulfite and the Infinium EPIC methylation array, detecting 5-hydroxymethyl cytosine at over 850,000 locations for region-specific hydroxymethylation assessment. Notably, we observed significant sharing of hydroxymethylated genomic regions during MP across the samples. Hydroxymethylated CpGs were enriched in open sea and intergenic regions (p < 0.001), and genes undergoing hydroxymethylation were significantly associated with cancer-related signaling pathways. RNA sequencing data integration identified 91 genes with significant positive/negative hydroxymethylation-expression correlations. Functional analysis suggested that positively correlated genes are involved in cell-cycle promotion, while negatively correlated ones are associated with antineoplastic functions. Analyses of The Cancer Genome Atlas clinical data on glioma were in line with these findings. Motif-enrichment analysis suggested the potential involvement of the transcription factor KLF4 in hydroxymethylation-based gene regulation. Our findings shed light on the significance of region-specific DNA hydroxymethylation in glioma MP and suggest its potential role in cancer-related gene expression and IDHmt glioma malignancy.

Transcriptomic and epigenetic dissection of spinal ependymoma (SP-EPN) identifies clinically relevant subtypes enriched for tumors with and without NF2 mutation

Ependymomas encompass multiple clinically relevant tumor types based on localization and molecular profiles. Tumors of the methylation class "spinal ependymoma" (SP-EPN) represent the most common intramedullary neoplasms in children and adults. However, their developmental origin is ill-defined, molecular data are scarce, and the potential heterogeneity within SP-EPN remains unexplored. The only known recurrent genetic events in SP-EPN are loss of chromosome 22q and NF2 mutations, but neither types and frequency of these alterations nor their clinical relevance have been described in a large, epigenetically defined series. Transcriptomic (n = 72), epigenetic (n = 225), genetic (n = 134), and clinical data (n = 112) were integrated for a detailed molecular overview on SP-EPN. Additionally, we mapped SP-EPN transcriptomes to developmental atlases of the developing and adult spinal cord to uncover potential developmental origins of these tumors. The integration of transcriptomic ependymoma data with single-cell atlases of the spinal cord revealed that SP-EPN display the highest similarities to mature adult ependymal cells. Unsupervised hierarchical clustering of transcriptomic data together with integrated analysis of methylation profiles identified two molecular SP-EPN subtypes. Subtype A tumors primarily carried previously known germline or sporadic NF2 mutations together with 22q loss (bi-allelic NF2 loss), resulting in decreased NF2 expression. Furthermore, they more often presented as multilocular disease and demonstrated a significantly reduced progression-free survival as compared to SP-EP subtype B. In contrast, subtype B predominantly contained samples without NF2 mutation detected in sequencing together with 22q loss (monoallelic NF2 loss). These tumors showed regular NF2 expression but more extensive global copy number alterations. Based on integrated molecular profiling of a large multi-center cohort, we identified two distinct SP-EPN subtypes with important implications for genetic counseling, patient surveillance, and drug development priorities.

Glioblastoma with high O6-methyl-guanine DNA methyltransferase expression are more immunologically active than tumors with low MGMT expression

Background

Glioblastoma (GBM) is a highly lethal brain tumor. The effectiveness of temozolomide (TMZ) treatment in GBM is linked to the methylation status of O6-methyl-guanine DNA methyltransferase (MGMT) promoter. Patients with unmethylated MGMT promoter have limited treatment options available. Consequently, there is a pressing need for alternative therapeutic strategies for such patients.

Methods

Data, including transcriptomic and clinical information, as well as information on MGMT promoter methylation status in primary GBM, were obtained from The Cancer Genome Atlas (TCGA) (n=121) and Chinese Glioma Genome Atlas (CGGA) (n=83) datasets. Samples were categorized into high and low MGMT expression groups, MGMT-high (MGMT-H) and MGMT-low (MGMT-L) tumors. A comprehensive transcriptome analysis was conducted to explore the tumor-immune microenvironment. Furthermore, we integrated transcriptome data from 13 GBM patients operated at our institution with findings from tumor-infiltrating lymphocyte (TIL) cultures, specifically investigating their response to autologous tumors.

Results

Gene signatures associated with various immune cells, including CD8 T cells, helper T cells, B cells, and macrophages, were noted in MGMT-H tumors. Pathway analysis confirmed the enrichment of immune cell-related pathways. Additionally, biological processes involved in the activation of monocytes and lymphocytes were observed in MGMT-H tumors. Furthermore, TIL culture experiments showed a greater presence of tumor-reactive T cells in MGMT-H tumors compared to MGMT-L tumors. These findings suggest that MGMT-H tumors has a potential for enhanced immune response against tumors mediated by CD8 T cells.

Conclusion

Our study provides novel insights into the immune cell composition of MGMT-H tumors, which is characterized by the infiltration of type 1 helper T cells and activated B cells, and also the presence of tumor-reactive T cells evidenced by TIL culture. These findings contribute to a better understanding of the immune response in MGMT-H tumors, emphasizing their potential for immunotherapy. Further studies are warranted to investigate on the mechanisms of MGMT expression and antitumor immunity.

DDDR-02. CDK4/6 INHIBITION WITH ABEMACICLIB SENSITIZES INTRACRANIAL TUMORS TO CHECKPOINT BLOCKADE IN PRE-CLINICAL MODELS OF BRAIN METASTASIS

Although immune checkpoint inhibitors (ICI) have become a useful tool in the management of central nervous system (CNS) metastases, only a subset of patients experience durable clinical responses and prognosis continues to remain poor. In particular, the role of ICI for CNS metastases from breast cancer has also not been adequately explored. CDK4/6 inhibition has been shown to sensitize extracranial tumors to ICI, and we previously reported intracranial efficacy of combination CDK4/6 inhibition with abemaciclib and anti-PD-1 in preclinical models of melanoma brain metastasis. We have now investigated this combination strategy further in a mouse model of triple negative breast cancer brain metastasis. 4T1 mammary carcinoma cells were first injected into the mammary fat pad and then intracranially 3 days later. While single agent anti-PD-1 antibody or abemaciclib were both ineffective at reducing tumor growth, we found a significant reduction in intracranial and extracranial tumor burden when abemaciclib was combined with anti-PD-1. Immunofluorescence for CD8 and CD3 revealed a significant increase in tumor-infiltrating CD8+ T cells with abemaciclib monotherapy as well as combination therapy in intracranial 4T1 tumors. We observed a similar increase in intracranial tumor-infiltrating CD8+ T cells in two melanoma brain metastasis models, YUMM1.7 and B16-F10, following combination therapy. We conclude that combination with abemaciclib can improve intracranial efficacy of ICI by driving CD8+ T cell infiltration into intracranial tumors. Future work will focus on investigating additional treatment-induced changes in the tumor immune microenvironment and dissecting the effect of CDK4/6 inhibition on T cell activation and function.

MRI-Based Radiomics Differentiates Skull Base Chordoma and Chondrosarcoma: A Preliminary Study

Simple Summary

In this study, we created a novel MRI-based machine learning model to differentiate skull base chordoma and chondrosarcoma with multiparametric signatures. While these tumors share common radiographic characteristics, clinical behavior is distinct. Therefore, distinguishing these tumors before initial surgical intervention would be useful, potentially impacting the surgical strategy. Although there are some limitations, such as the risk of overfitting and the lack of an extramural cohort for truly independent final validation, our machine learning model distinguishing chordoma from chondrosarcoma yielded superior diagnostic accuracy to that achieved by 20 board-certified neurosurgeons.

Abstract

Chordoma and chondrosarcoma share common radiographic characteristics yet are distinct clinically. A radiomic machine learning model differentiating these tumors preoperatively would help plan surgery. MR images were acquired from 57 consecutive patients with chordoma (N = 32) or chondrosarcoma (N = 25) treated at the University of Tokyo Hospital between September 2012 and February 2020. Preoperative T1-weighted images with gadolinium enhancement (GdT1) and T2-weighted images were analyzed. Datasets from the first 47 cases were used for model creation, and those from the subsequent 10 cases were used for validation. Feature extraction was performed semi-automatically, and 2438 features were obtained per image sequence. Machine learning models with logistic regression and a support vector machine were created. The model with the highest accuracy incorporated seven features extracted from GdT1 in the logistic regression. The average area under the curve was 0.93 ± 0.06, and accuracy was 0.90 (9/10) in the validation dataset. The same validation dataset was assessed by 20 board-certified neurosurgeons. Diagnostic accuracy ranged from 0.50 to 0.80 (median 0.60, 95% confidence interval 0.60 ± 0.06%), which was inferior to that of the machine learning model (p = 0.03), although there are some limitations, such as the risk of overfitting and the lack of an extramural cohort for truly independent final validation. In summary, we created a novel MRI-based machine learning model to differentiate skull base chordoma and chondrosarcoma from multiparametric signatures.

EPEN-27. Epigenetic dissection of spinal ependymomas (SP-EPN) separates tumors with and withoutNF2 mutation

Ependymomas encompass multiple, clinically relevant tumor types based on localization, genetic alterations, and epigenetic and transcriptomic profiles. Tumors belonging to the methylation class of spinal ependymoma (SP-EPN) represent the most common intramedullary neoplasms in children and adults. However, molecular data of SP-EPN are scarce, and clear treatment recommendations are lacking. The only known recurrent genetic events in SP-EPN are loss of chromosome 22q and NF2 mutations. Yet, it remains unclear whether SP-EPN with germline or sporadic NF2 mutations or with NF2 wild type status differ clinically or molecularly. To provide a comprehensive molecular profile of SP-EPN, we integrated epigenetic, genomic, transcriptomic, and histological analyses of up to 237 cases. Clustering of methylation data revealed two distinct molecular SP-EPN subtypes. The distribution of NF2 mutated cases differed significantly across these subtypes (p <0.0001): The vast majority of tumors harboring either a previously known NF2 germline mutation or a sporadic mutation were assigned to subtypes A, whereas subtype B tumors mainly contained NF2 wild type sequences. In addition, subtype A tumors showed a lower frequency of MGMT promoter methylation (p= 0.018) and contained almost all pediatric patients of the cohort. Whole-exome sequencing (30 cases) identified numerous mutations in NF2 wild type and mutated tumors. Mutated genes in NF2 wild type tumors were enriched for genes associated with cell cycle and cytoskeleton. RNA sequencing revealed two distinct transcriptional groups with upregulation of proliferative genes in one group and upregulation of cilial genes in the other group. The molecular subtypes displayed subtle, but significant differences in the appearance of histopathological characteristics, such as surfaces, inflammation, and hyalinized vessels. Investigation of clinical parameters is ongoing and will complete the picture of SP-EPN heterogeneity as an important basis for future clinical decision-making.

SURG-10. DEVELOPMENT OF NOVEL TOPICAL FLUORESCENT PROBE FOR INTRAOPERATIVE RAPID DETECTION OF GLIOMA

PURPOSE

Fluorescence imaging is an important surgical adjunct in malignant glioma surgery. 5-aminolevulinic acid (5-ALA) has been proven effective for radical tumor resection and extended progression-free survival in a phase III randomized trial and therefore integrated into surgery for malignant glioma. Importantly, however, some limitations still exist in its use, which include false positivity and false negativity as well as inability of re-administration. In this study, we aimed to develop a novel, spray-type fluorescent probe using hydroxymethyl rhodamine green (HMRG) as a fluorescent scaffold.

METHODS

We have previously established a fluorescent probe library comprised of more than 320 kinds of HMRG probes. They have HMRG as a fluorescent scaffold with various types of dipeptides attached to it. Primary probe screening was performed using the homogenized tumor samples from patients with glioblastoma operated at our institution. Secondary screening followed using the selected probes and fresh tumor samples obtained from patients with glioblastoma operated from 2016 until 2018. Diced electrophoresis gel (DEG) assay, two-dimensional gel electrophoresis followed by a multi-well plate-based fluorometric assay, was performed to identify responsible enzymes for the selected probe. Further experiments with inhibitors, real-time PCR, immunohistochemistry, and western blotting were performed for confirmation.

RESULTS

Proline-arginine-HMRG (PR-HMRG) was selected as a candidate probe based upon the above two-step screenings. It achieved 79.4% accuracy in receiver operating characteristic curve analysis. Calpain-1 was found to be responsible to cleave PR-HMRG probe by DEG-proteome analysis. Calpain-1 protein was expressed at significantly higher level in tumors that were fluoresced by PR-HMRG than in those that were not.

CONCLUSIONS

Our innovative screening method was able to find PR-HMRG as a novel fluorescent probe effective for rapid detection of glioblastoma. A preclinical study is planned to assess the efficacy and safety of the selected probe.

EPCO-01. MOLECULAR PROFILING OF SPINAL CORD EPENDYMOMA

BACKGROUND

Ependymomas are currently classified into 9 subgroups by DNA methylation profiles. Although spinal cord ependymoma (SP-EPN) is distinct from other tumors, diversity within SP-EPN is still unclear. Here, we used transcriptomic and epigenomic profiles to investigate the diversity among Japanese SP-EPN cases.

MATERIALS AND METHODS

We analyzed 57 SP-EPN patients (32 males and 25 females, aged from 18 to 78 years, median: 52), including two cases of neurofibromatosis type 2, five cases of grade 3 (WHO grade). We obtained transcriptome (RNA-seq) and DNA methylation (Infinium Methylation EPIC array) data from fresh frozen specimens of SP-EPN resected at the University of Tokyo Hospital and our collaborative groups.

RESULTS

Three cases had a previous intracranial ependymoma operation. Hierarchical clustering of the DNA methylation data showed that these three cases of intracranial origin as a different cluster from spinal origin. The 45 grade 2 spinal ependymoma showed a relatively homogenous methylation pattern. However, the methylation status of HOX gene cluster regions is compatible with the segment of origin, which reflects the cells of origins are derived after the determination of segment identity. RNA sequencing of 57 cases revealed two subgroups within grade 2. Gene ontology analysis of differentially expressed genes suggested the difference in metabolic state such as rRNA translation and mitochondrial respiration between the two expression subgroups.

CONCLUSION

Epigenetic analysis indicated the accurate body segment origin of SP-EPN. We observed that metabolic states could divide grade 2 spinal cord ependymoma into 2 subgroups and will present the relationship to clinicopathological information.

A Novel Topical Fluorescent Probe for Detection of Glioblastoma

PURPOSE

Five-aminolevulinic acid (5-ALA) is widely used as an intraoperative fluorescent probe for radical resection of high-grade glioma, and thus aids in extending progression-free survival of patients. However, there exist some cases where 5-ALA fails to fluoresce. In some other cases, it may undergo fluorescence quenching but cannot be orally readministered during surgery. This study aimed to develop a novel hydroxymethyl rhodamine green (HMRG)-based fluorescence labeling system that can be repeatedly administered as a topical spray during surgery for the detection of glioblastoma.

EXPERIMENTAL DESIGN

We performed a three-stage probe screening using tumor lysates and fresh tumor tissues with our probe library consisting of a variety of HMRG probes with different dipeptides. We then performed proteome and transcript expression analyses to detect candidate enzymes responsible for cleaving the probe. Moreover, in vitro and ex vivo studies using U87 glioblastoma cell line were conducted to validate the findings.

RESULTS

The probe screening identified proline-arginine-HMRG (PR-HMRG) as the optimal probe that distinguished tumors from peritumoral tissues. Proteome analysis identified calpain-1 (CAPN1) to be responsible for cleaving the probe. CAPN1 was highly expressed in tumor tissues which reacted to the PR-HMRG probe. Knockdown of this enzyme suppressed fluorescence intensity in U87 glioblastoma cells. In situ assay using a mouse U87 xenograft model demonstrated marked contrast of fluorescence with the probe between the tumor and peritumoral tissues.

CONCLUSIONS

The novel fluorescent probe PR-HMRG is effective in detecting glioblastoma when applied topically. Further investigations are warranted to assess the efficacy and safety of its clinical use.

PATH-33. EPIGENOMIC ANALYSIS OF SPINAL EPENDYMOMA

BACKGROUND

Ependymomas commonly occur in the fourth ventricle and the spinal cord. Gross total resection, age and WHO grade are known prognostic factors. Ependymomas are currently classified into 9 distinct subgroups by DNA methylation profile analysis. Spinal cord ependymoma is distinct from other subgroups. To investigate heterogeneity within spinal cord ependymoma, we examined DNA methylation profiles.

MATERIALS AND METHODS

We used Infinium MethylationEPIC array (illumina) to obtain DNA methylation data from frozen specimens of spinal ependymoma resected at the University of Tokyo, Osaka City University, and Tokyo Metropolitan Neurological Hospital. Japan Pediatric Molecular Neuro-Oncology Group provided methylation data for 11 reported cases. Cluster analysis was performed using Cluster3.0.

RESULTS

We analyzed 34 patients, 21 male and 13 female, aged from 18 to 76 years (median 50.5 years), including 2 cases with neurofibromatosis type 2. WHO grade was grade_3 in 2 cases and grade_2 in others. Clustering of the DNA methylation data showed that WHO grade_3 cases tended to be classified into a subgroup distinct from other cases.

CONCLUSION

This is the largest DNA methylation profiling study on spinal cord ependymoma to date. The study may suggest a new subgroup correlated with higher WHO grade.

First case of human neurocoenurosis caused by Taenia serialis: A case report

Human coenurosis is caused by the larval stages of Taenia species, mainly Taenia multiceps and Taenia serialis. T. multiceps has been reported to cause human central nervous system (CNS) infections, but no CNS case caused by T. serialis has been reported. The authors report the first case of human neurocoenurosis caused by T. serialis, which was confirmed by mitochondrial DNA analysis. A 38-year-old man presented with visual disturbance and headache, and subsequent magnetic resonance imaging (MRI) revealed a ring-enhancing cystic lesion in the left occipital lobe. Biopsy was performed, and the resultant histopathological diagnosis was that of low-grade B-cell lymphoma. Chemotherapy was initiated, but a subsequent MRI showed increased ring enhancement. Due to the unexpected clinical course, a surgical resection of the lesion was performed. The lesion was completely removed. Pathological examination showed multiple scolices with hooklets, suckers, and numerous calcareous corpuscles. Therefore, the diagnosis of neurocysticercosis was made. However, mitochondrial DNA analysis showed that the disease was definitively coenurosis caused by T. serialis. Albendazole was administered, with no evidence of recurrence at 12 months following the operation. In this study, we demonstrate that T. serialis can cause CNS infection and that genetic analysis is recommended to establish a definitive diagnosis.

A rare case of brain metastasis from poorly differentiated small bowel adenocarcinoma

Background:

Small bowel adenocarcinoma (SBA) accounts for <2% of all gastrointestinal malignancies. The most common organs of SBA metastases are the abdominal lymph node, liver, and peritoneum. There have been almost no reports of brain metastases of SBA. Dabaja et al. reported 1 case of brain metastasis out of 217 SBA cases, but details of the clinical course of the case were unclear. Our case might be the first report covering the full clinical course, pathological findings, and genetic data. Here, we report a very rare case of brain metastasis from poorly differentiated SBA.

Case Description:

A 54-year-old man who suffered from abdominal pain and melena visited a nearby hospital. This patient had no risk factors for SBA. He underwent partial resection of the jejunum with regional lymphadenectomy and combined resection of the transverse colon. Pathological diagnosis was poorly differentiated adenocarcinoma, pT4N2M0 Stage IIIB (UICC-TNM: 8^th edition). One month after curative surgery, liver metastasis was detected by a computed tomography (CT) scan, and then, palliative chemotherapy was started. During the third-line chemotherapy, a brain tumor on the left cerebellum was detected by the CT scan. Tumor resection was performed, and the histopathological features coincided with the primary jejunum tumor. Based on surgical, radiological, pathological, and genetic findings, this brain tumor was comprehensively diagnosed as a metastasis from poorly differentiated SBA.

Conclusion:

Here, we experienced a very rare case of brain metastasis from poorly differentiated SBA.

BOT-03 INVESTIGATION OF NOVEL SPRAY TYPE FLUORESCENT PROBE FOR GLIOBLASTOMA DETECTION

PURPOSE

5-ALA is commonly used as an intraoperative tool in malignant glioma surgery, which has been proven effective for radical tumor resection and extended progression-free survival. However, there are some limitations in its use, such as false positivity, false negativity, and inability of re-administration. We aim to develop a novel fluorescent labeling system which can be repeatedly administered by spray during surgery, using hydroxymethyl rhodamine green (HMRG) as fluorescent scaffold originally designed at our university for cancer detection.

METHODS

Primary probe screening was performed using the homogenized glioblastoma (GBM) samples with the fluorescent probe library comprised of more than 320 kinds of HMRG fluorescent scaffold combined with various types of dipeptides. Second probe screening was performed using fresh GBM specimens and the selected probes in primary screening. To identify the responsible enzymes, diced electrophoresis gel (DEG) assay was performed. This method utilizes the combination of two dimensional electrophoresis (isoelectric point and molecular weight) and a multiwell-plate-based fluorometric assay to find protein spots with the specified activities.

RESULTS

The prominent probes were selected based upon the above two-step screenings. We identified two enzymes by proteome analysis and experiments using inhibitors, which was further confirmed with real-time PCR and western blotting.

DISCUSSION

This screening methodology is innovative in that it is based on selecting probes from the probe library that respond to clinical samples rather than creating probes from the responsible enzymes. Practical fluorescent probes can be established even for low-grade gliomas, which would be a breakthrough for rapid intraoperative diagnosis in glioma surgery.

CONCLUSION

HMRG-based aminopeptidase fluorescent probes may be effective for GBM detection.

RARE-16. A NOVEL RADIOMICS MODEL DIFFERENTIATING CHORDOMA AND CHONDROSARCOMA

BACKGROUND

Chordoma and chondrosarcoma account for the majority of skull base tumors affecting the petroclival region. Chondrosarcoma has better prognosis than chordoma; heavy particle therapy is often indicated for residual/recurrent chordoma. Preoperative, precise diagnosis of the tumor would be desirable, as it can potentially impact on the choice of a surgical approach and the aggressiveness of surgery.

METHODS

We conducted a radiomics study to create a machine learning model distinguishing chondrosarcoma from chordoma. We collected DICOM T2-weighted images and T1-weighted images with gadolinium (GdT1) enhancement in the consective patients of chordoma or chondrosarcoma who underwent surgery at The University of Tokyo Hospital from September of 2012 to January of 2019. We selected patients with uniform MRI images. VOI (volume of interest) was set using Monaco (https://www.elekta.com/software-solutions/treatment-management/external-beam-planning/monaco.html). Not only sematic features but also agnostic features were calculated. The original images and 8 wavelet transformed images were calculated for texture agnostic features such as Gray-Level Co-occurrence Matrix (GLCM). Features were selected by recursive feature elimination (RFE). The final model evaluation was performed by average area under the curve (AUC).

RESULTS

The study population included 17 chordomas and 22 chondrosarcomas in a total of 39 patients. 476 features were obtained per image sequence. The number of features per case was 476 × 2 = 952 The most accurate machine learning model was created using the extracted three features from only T2. The best AUC was 0.77 ± 0.11 in logistic regression (dataset was divided randomly into halves, average value of AUC calculated six times).

CONCLUSIONS

This novel machine learning model can differentiate chordoma and chondrosarcoma reasonably well. A validation study with a larger number of patients is warranted.

NIMG-11. VISUALIZATION OF JUDGMENT BASIS OF CNN TO GRADING GLIOMA

BACKGROUND

convolutional neural network (CNN) model using MRI imaging are likely to be effective for grading glioma. However, interpretation of the judgment basis of CNN is difficult. The purpose of this study is twofold. One is to create a high accuracy machine learning model that grading of glioma (Glioma grade II, between III and IV). The other is to visualize the judgment basis of the model.

METHODS

We targeted cases that were imaged at our Hospital during the period from August 2014 to January 2018. Five types of MRI are used. The five types are two types of DWI (b1000DWI and b2000DWI, respectively, with a value of 1000 and 2000), apparent diffusion coefficient (ADC), fractional anisotropy (FA), and mean kurtosis (MK). The images were input to CNN wtihout ROI. Seven types of CNN were prepared, and 100 epochs of learning were performed.

RESULTS

55 cases were included. There were 14 grade II, 12 grade III, and 29 grade IV). Of these, 44 cases up to July 2017 in chronological order were used as a dataset for learning, and 11 cases from August 2017 to January 2018 were taken as independent test dataset. The best results were obtained using MK as input, with a percentage of correct cases of test data of 0.82. Subsequently, we applied Grad-CAM to the model. Grad-CAM is visualized technique that shows where CNN focused on. The model focused on the tumor part of the image. In addition, in the cases of glioma and meningioma coexisting, grading was performed focusing on the glioma area not on meningioma area.

CONCLUSION

When grading glioma, it was considered that CNN decide glioma’s grade focusing on tumor part of the image, as well as humans.

DDIS-02. DEVELOPMENT OF NOVEL SPRAY-TYPE FLUORESCENT PROBES FOR GLIOBLASTOMA DETECTION

PURPOSE: ALA is commonly used as an intraoperative tool in malignant glioma surgery, which has been proven effective for radical tumor resection and extended progression-free survival. However, there are some limitations in its use, such as false positivity, false negativity, and inability of re-administration. We aim to develop a novel fluorescent labeling system which can be repeatedly administered by spray during surgery, using hydroxymethyl rhodamine green (HMRG) as fluorescent scaffold originally designed at our university for cancer detection. [Methods]Primary probe screening was performed using the homogenized glioblastoma (GBM) samples with the fluorescent probe library comprised of more than 320 kinds of HMRG fluorescent scaffold combined with various types of dipeptides. Second probe screening was performed using fresh GBM specimens and the selected probes in primary screening. To identify the responsible enzymes, diced electrophoresis gel (DEG) assay was performed. This method utilizes the combination of two dimensional electrophoresis (isoelectric point and molecular weight) and a multiwell-plate-based fluorometric assay to find protein spots with the specified activities. [Results] The prominent probes were selected based upon the above two-step screenings. We identified two enzymes by proteome analysis and experiments using inhibitors, which was further confirmed with real-time PCR and western blotting. [Discussion] This screening methodology is innovative in that it is based on selecting probes from the probe library that respond to clinical samples rather than creating probes from the responsible enzymes. Practical fluorescent probes can be established even for low-grade gliomas, which would be a breakthrough for rapid intraoperative diagnosis in glioma surgery. [Conclusion] HMRG-based aminopeptidase fluorescent probes may be effective for GBM detection.

COMP-02. MINING-GUIDED MACHINE LEARNING ANALYSES SUPPORTS GRASPING THE LATEST TRENDS ON NEURO-ONCOLOGY

The systems that can objectively predict the future trends of a particular research field are always anticipated while conducting medical research. Such systems also provide a considerable aid to researchers while determining and acquiring appropriate research budgets. This study intended to establish a novel and versatile algorithm that can predict the latest trends in neuro-oncology. Seventy-nine neuro-oncological research fields were selected using computational sorting methods, such as text-mining analyses, along with 30 journals that represent the recent trends in the neuro-oncology field. Further, the annual impact (AI) for each year with respect to each journal and field (number of articles published in the journal × the impact factor of the journal) was calculated as a novel concept. Subsequently, the AI index (AII) for the year was defined as the sum of the AIs for the aforementioned 30 journals. With respect to the aforementioned neuro-oncological research fields, the AII trends from 2008 to 2017 were subjected to machine learning predicting analyses. The prediction accuracy of the latest trends in neuro-oncology was validated using actual data obtained from previous studies. In particular, the linear prediction model achieved a relatively good accuracy. The most notable and latest predicted fields in neuro-oncology included some interesting emerging fields, such as microenvironment and anti-mitosis, as well as the already renowned fields, such as immunology and epigenetics. Furthermore, we retrospectively attempted an analysis of the fields different from neuro-oncology. Interestingly, as of 2008, the future emergence of the CRISPR-Cas9 gene editing system has been predicted using this system. Overall, the presented algorithm displays potential to be an effective and versatile tool for the prediction of future trends in a particular medical field.