The role of input imaging combination and ADC threshold on segmentation of acute ischemic stroke lesion using U-Net.

BACKGROUND: To evaluate the effect of the weighting of input imaging combo and ADC threshold on the performance of the U-Net and to find an optimized input imaging combo and ADC threshold in segmenting acute ischemic stroke (AIS) lesion. METHODS: This study retrospectively enrolled a total of 212 patients having AIS. Four combos, including ADC-ADC-ADC (AAA), DWI-ADC-ADC (DAA), DWI-DWI-ADC (DDA), and DWI-DWI-DWI (DDD), were used as input images, respectively. Three ADC thresholds including 0.6, 0.8 and 1.8 × 10-3 mm2/s were applied. Dice similarity coefficient (DSC) was used to evaluate the segmentation performance of U-Nets. Nonparametric Kruskal-Wallis test with Tukey-Kramer post-hoc tests were used for comparison. A p < .05 was considered statistically significant. RESULTS: The DSC significantly varied among different combos of images and different ADC thresholds. Hybrid U-Nets outperformed uniform U-Nets at ADC thresholds of 0.6 × 10-3 mm2/s and 0.8 × 10-3 mm2/s (p < .001). The U-Net with imaging combo of DDD had segmentation performance similar to hybrid U-Nets at an ADC threshold of 1.8 × 10-3 mm2/s (p = .062 to 1). The U-Net using the imaging combo of DAA at the ADC threshold of 0.6 × 10-3 mm2/s achieved the highest DSC in the segmentation of AIS lesion. CONCLUSIONS: The segmentation performance of U-Net for AIS varies among the input imaging combos and ADC thresholds. The U-Net is optimized by choosing the imaging combo of DAA at an ADC threshold of 0.6 × 10-3 mm2/s in segmentating AIS lesion with highest DSC. KEY POINTS: • Segmentation performance of U-Net for AIS differs among input imaging combos. • Segmentation performance of U-Net for AIS differs among ADC thresholds. • U-Net is optimized using DAA with ADC = 0.6 × 10-3 mm2/s.

Improving interobserver agreement and performance of deep learning models for segmenting acute ischemic stroke by combining DWI with optimized ADC thresholds.

OBJECTIVES: To examine the role of ADC threshold on agreement across observers and deep learning models (DLMs) plus segmentation performance of DLMs for acute ischemic stroke (AIS). METHODS: Twelve DLMs, which were trained on DWI-ADC-ADC combination from 76 patients with AIS using 6 different ADC thresholds with ground truth manually contoured by 2 observers, were tested by additional 67 patients in the same hospital and another 78 patients in another hospital. Agreement between observers and DLMs were evaluated by Bland-Altman plot and intraclass correlation coefficient (ICC). The similarity between ground truth (GT) defined by observers and between automatic segmentation performed by DLMs was evaluated by Dice similarity coefficient (DSC). Group comparison was performed using the Mann-Whitney U test. The relationship between the DSC and ADC threshold as well as AIS lesion size was evaluated by linear regression analysis. A p < .05 was considered statistically significant. RESULTS: Excellent interobserver agreement and intraobserver repeatability in the manual segmentation (all ICC > 0.98, p < .001) were achieved. The 95% limit of agreement was reduced from 11.23 cm2 for GT on DWI to 0.59 cm2 for prediction at an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI. The segmentation performance of DLMs was improved with an overall DSC from 0.738 ± 0.214 on DWI to 0.971 ± 0.021 on an ADC threshold of 0.6 × 10-3 mm2/s combined with DWI. CONCLUSIONS: Combining an ADC threshold of 0.6 × 10-3 mm2/s with DWI reduces interobserver and inter-DLM difference and achieves best segmentation performance of AIS lesions using DLMs. KEY POINTS: • Higher Dice similarity coefficient (DSC) in predicting acute ischemic stroke lesions was achieved by ADC thresholds combined with DWI than by DWI alone (all p < .05). • DSC had a negative association with the ADC threshold in most sizes, both hospitals, and both observers (most p < .05) and a positive association with the stroke size in all ADC thresholds, both hospitals, and both observers (all p < .001). • An ADC threshold of 0.6 × 10-3 mm2/s eliminated the difference of DSC at any stroke size between observers or between hospitals (p = .07 to > .99).

Treatment monitoring of colorectal cancer by integrated analysis of plasma concentration and sequencing of circulating tumor DNA.

Circulating cell-free DNA (cfDNA) analysis is an important tool for cancer monitoring. The patient-specific mutations identified in colorectal cancer (CRC) tissues are usually used to design the cfDNA analysis. Despite high specificity in predicting relapse, the sensitivity in most studies is around 40-50%. To improve this weakness, we designed a cfDNA panel according to the CRC genomic landscape and recurrent-specific mutations. The pathological variants in cfDNA samples from 60 CRC patients were studied by a next-generation sequencing (NGS) method incorporating the dual molecular barcode. Interestingly, patients in the disease positive group had a significantly higher cfDNA concentration than those in the disease negative group. Based on receiver operating characteristic analysis, the cfDNA concentration of 7 ng/mL was selected into the analytical workflow. The sensitivity in determining the disease status was 72.4%, which represented a considerable improvement on prior studies, and the specificity remained high at 80.6%. Compared to standard imaging and laboratory studies, earlier detection of residual disease and clinical benefits were shown on two cases by this cfDNA assay. We conclude this integrative framework of cfDNA analytical pipeline with a satisfactory sensitivity and specificity could be used in postoperative CRC surveillance.

Computed tomography with histological correlation for evaluating tumor regression of rectal carcinoma after preoperative chemoradiation therapy.

BACKGROUND/AIMS: Preoperative chemoradiation therapy (CRT) is standard procedure for locally advanced rectal cancer. The correlation of tumor response evaluated using CT according to response evaluation criteria in solid tumors (RECIST) with the histo-logical tumor regression grade (TRG) is not well-documented. METHODOLOGY: Ninety-one patients with rectal cancer underwent CT examinations before and after preoperative CRT and following surgery. Clinical tumor staging and tumor response assessed according to RECIST were done on paired CT scans. Pathological tumor staging and TRGs were reviewed in resected specimens.Post-CRT CT findings and histological findings were compared. Survival analysis for 73 patients was done. RESULTS: TRG was positively correlated with the CT-assessed tumor response (r=0.276, p=0.009). Thickened fibrotic areas and muscle disarray caused by fibrosis were more frequently seen in cases of patients over-diagnosed as having residual tumors. The ycT status was positively correlated with ypT status (r=0.44, p<0.001;accuracy=61.5%). Downstaging of cT status was cor-related with a lower TRG (p=0.001). CONCLUSIONS: Fibrosis emerges after neoadjuvant therapy and is usual-ly accompanied by tumor reduction on CT scans of rec-tal cancer patients following preoperative CRT. There-fore, tumor response assessed using CT according to RECIST may serve as a supplementary tool for preoperative planning other than tumor restaging.

Targeting protumor factor chitinase-3-like-1 secreted by Rab37 vesicles for cancer immunotherapy.

Background: Chitinase 3-like-1 (CHI3L1) is a secretion glycoprotein associated with the immunosuppressive tumor microenvironment (TME). The secretory mode of CHI3L1 makes it a promising target for cancer treatment. We have previously reported that Rab37 small GTPase mediates secretion of IL-6 in macrophages to promote cancer progression, whereas the roles of Rab37 in the intracellular trafficking and exocytosis of CHI3L1 are unclear. Methods: We examined the concentration of CHI3L1 in the culture medium of splenocytes and bone marrow derived macrophages (BMDMs) from wild-type or Rab37 knockout mice, and macrophage or T cell lines expressing wild type, active GTP-bound or inactive GDP-bound Rab37. Vesicle isolation, total internal reflection fluorescence microscopy, and real-time confocal microscopy were conducted. We developed polyclonal neutralizing-CHI3L1 antibodies (nCHI3L1 Abs) to validate the therapeutic efficacy in orthotopic lung, pancreas and colon cancer allograft models. Multiplex fluorescence immunohistochemistry was performed to detect the protein level of Rab37 and CHI3L1, and localization of the tumor-infiltrating immune cells in allografts from mice or tumor specimens from cancer patients. Results: We demonstrate a novel secretion mode of CHI3L1 mediated by the small GTPase Rab37 in T cells and macrophages. Rab37 mediated CHI3L1 intracellular vesicle trafficking and exocytosis in a GTP-dependent manner, which is abolished in the splenocytes and BMDMs from Rab37 knockout mice and attenuated in macrophage or T cell lines expressing the inactive Rab37. The secreted CHI3L1 activated AKT, ß-catenin and NF-κB signal pathways in cancer cells and macrophages to foster a protumor TME characterized by activating M2 macrophages and increasing the population of regulatory T cells. Our developed nCHI3L1 Abs showed the dual properties of reducing tumor growth/metastases and eliciting an immunostimulatory TME in syngeneic orthotopic lung, pancreas and colon tumor models. Clinically, high plasma level or intratumoral expression of CHI3L1 correlated with poor survival in 161 lung cancer, 155 pancreatic cancer and 180 colon cancer patients. Conclusions: These results provide the first evidence that Rab37 mediates CHI3L1 secretion in immune cells and highlight nCHI3L1 Abs that can simultaneously target both cancer cells and tumor microenvironment.

Improving performance of deep learning models using 3.5D U-Net via majority voting for tooth segmentation on cone beam computed tomography.

Deep learning allows automatic segmentation of teeth on cone beam computed tomography (CBCT). However, the segmentation performance of deep learning varies among different training strategies. Our aim was to propose a 3.5D U-Net to improve the performance of the U-Net in segmenting teeth on CBCT. This study retrospectively enrolled 24 patients who received CBCT. Five U-Nets, including 2Da U-Net, 2Dc U-Net, 2Ds U-Net, 2.5Da U-Net, 3D U-Net, were trained to segment the teeth. Four additional U-Nets, including 2.5Dv U-Net, 3.5Dv5 U-Net, 3.5Dv4 U-Net, and 3.5Dv3 U-Net, were obtained using majority voting. Mathematical morphology operations including erosion and dilation (E&D) were applied to remove diminutive noise speckles. Segmentation performance was evaluated by fourfold cross validation using Dice similarity coefficient (DSC), accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV). Kruskal-Wallis test with post hoc analysis using Bonferroni correction was used for group comparison. P < 0.05 was considered statistically significant. Performance of U-Nets significantly varies among different training strategies for teeth segmentation on CBCT (P < 0.05). The 3.5Dv5 U-Net and 2.5Dv U-Net showed DSC and PPV significantly higher than any of five originally trained U-Nets (all P < 0.05). E&D significantly improved the DSC, accuracy, specificity, and PPV (all P < 0.005). The 3.5Dv5 U-Net achieved highest DSC and accuracy among all U-Nets. The segmentation performance of the U-Net can be improved by majority voting and E&D. Overall speaking, the 3.5Dv5 U-Net achieved the best segmentation performance among all U-Nets.

Dynamic change of tetraspanin CD151 membrane protein expression in colorectal cancer patients.

PURPOSE: To measure the CD151 expression in colorectal cancer (CRC). METHODS: CD151 expression was assessed in 179 CRC patients and 39 patients with hepatic liver metastasis. RESULTS: High CD151 expression was observed in 48% of patients with early-stage CRC versus only 33% of patients with metastatic colon cancer. A higher level of tumor invasion status correlated with a decrease in CD151 expression. Metastatic stage and advanced tumor stage correlated with a decreased CD151 expression. Twenty-seven out of the 39-paired samples had high CD151 expression in liver metastasis sites. CONCLUSIONS: CD151 expression is decreased in patients with metastatic CRC.

Clinicopathological features of mismatch repair protein expression patterns in colorectal cancer.

Microsatellite instability (MSI) is reflective of a deficient mismatch repair (dMMR) system, which is mostly associated with the methylation of mismatch repair (MMR) genes and BRAF mutations in sporadic colorectal cancers (CRCs). We performed a retrospective study to analyze the clinicopathological features of dMMR CRCs and their association with the BRAF V600E mutation. The incidence of dMMR CRCs in our cohort was 7.4 % (118/1603). Immunohistochemistry (IHC) revealed four common dMMR IHC patterns in 116 dMMR CRCs from 110 patients. dMMR type 1 (MLH1-/PMS2-) CRCs were the most frequent pattern, usually showing typical proximal location and MSI histology. The BRAF V600E mutation was almost exclusively observed in dMMR type 1 (32 of 72) and dMMR type 2 (PMS- only, 7 of 18) CRCs (p = 0.001). Patients with dMMR type 3 (MSH2-/MSH6-) CRCs were usually diagnosed at younger ages (p < 0.001) and had the strongest family history of Lynch syndrome-associated tumors (p = 0.002). dMMR type 3 CRCs frequently presented at advanced stages (p = 0.005) with perineural invasion (p = 0.021). We also found a significant positive association of dMMR type 1 and type 3 with advanced stages of CRC, whereas dMMR types 2 and 4 (MSH6- only) were usually diagnosed at early stages of CRC (p < 0.001). In conclusion, BRAF V600E mutations almost exclusively occurred in dMMR type 1 and 2 CRCs. Patterns of MMR protein expression display distinct associations with tumor staging and age at diagnosis.

Clinical Utility of a Cell-Free DNA Assay in Patients With Colorectal Cancer.

The analysis of cell-free DNA (cfDNA) is rapidly emerging as a powerful approach to guide the clinical care of cancer patients. Several comprehensive cfDNA assays designed to detect mutations across several genes are now available. Here, we analyzed the use of a cfDNA panel in colorectal cancer (CRC) patients. Twenty-eight CRC patients with relapse or metastatic disease and 31 patients with no evidence of disease (NED) were enrolled. Genomic alterations in cfDNA were analyzed by the Oncomine™ Pan-Cancer Cell-Free Assay that detects hotspot mutations, small indels, copy number changes, and gene fusions across 52 genes. In the NED group, genomic alterations in cfDNA were detected in 12/31 patients (38.7%). The detection of alterations was more common in patients who were ≥60 years old, and the most common genomic alteration was a TP53 mutation. Fifty percent of the TP53 mutations were frequently or very frequently found in human cancers. Among 28 patients with relapse or metastatic disease, 22 (78.6%) had genomic alterations in cfDNA. The alterations were detected most frequently in TP53 (n = 10), followed by KRAS (n = 9). Actionable targets for CRC, including ERBB2 amplification and BRAF mutations, could be identified by this cfDNA assay. Compared with mutational profiling routinely analyzed using tumor samples, several additional targets with currently available therapies, including IDH1, IDH2, and PDGFRA mutations, were discovered. The cfDNA assay could identify potentially actionable targets for CRC. Identifying how to filter out cancer-like genomic alterations not derived from tumors remains a challenge.

Quantifying lumbar vertebral perfusion by a Tofts model on DCE-MRI using segmental versus aortic arterial input function.

The purpose of this study was to investigate the influence of arterial input function (AIF) selection on the quantification of vertebral perfusion using axial dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). In this study, axial DCE-MRI was performed on 2 vertebrae in each of eight healthy volunteers (mean age, 36.9 years; 5 men) using a 1.5-T scanner. The pharmacokinetic parameters Ktrans, ve, and vp, derived using a Tofts model on axial DCE-MRI of the lumbar vertebrae, were evaluated using various AIFs: the population-based aortic AIF (AIF_PA), a patient-specific aortic AIF (AIF_A) and a patient-specific segmental arterial AIF (AIF_SA). Additionally, peaks and delay times were changed to simulate the effects of various AIFs on the calculation of perfusion parameters. Nonparametric analyses including the Wilcoxon signed rank test and the Kruskal-Wallis test with a Dunn-Bonferroni post hoc analysis were performed. In simulation, Ktrans and ve increased as the peak in the AIF decreased, but vp increased when delay time in the AIF increased. In humans, the estimated Ktrans and ve were significantly smaller using AIF_A compared to AIF_SA no matter the computation style (pixel-wise or region-of-interest based). Both these perfusion parameters were significantly greater using AIF_SA compared to AIF_A.

Adequate preoperative biliary drainage is determinative to decrease postoperative infectious complications after pancreaticoduodenectomy.

BACKGROUND AND PURPOSE: To assess the effectiveness of preoperative biliary drainage (PBD) by preoperative serum bilirubin level in patients with periampullary lesions receiving pancreaticoduodenectomy. PATIENTS AND METHODS: Between Jan. 1995 to May 2005, 240 consecutive cases received pancreaticoduodenectomy at the National Cheng Kung University Hospital, Taiwan and were included retrospectively. Factors possibly affecting postoperative morbidity and mortality were analyzed. RESULTS: One hundred and forty-three patients (59.6%) underwent preoperative biliary drainage (the PBD group) and 97 patients without drainage (the non-PBD group). The total postoperative morbidity rate was 49.6% and postoperative mortality was 2.9%. There was no difference in total postoperative morbidity and mortality between groups, but higher incidence of sepsis/bacteremia in the PBD patients (p = 0.03), and more cardiovascular events (p = 0.05) in the non-PBD patients. More bile leakage developed in the non-PBD patients, but only with marginal significance (p = 0.09). In the PBD group, patients with preoperative serum bilirubin level > or = 5 mg/dL had higher likelihood to acquire an infectious complication, (OR: 2.70; CI: 1.21-6.04), and surgical site infectious (OR: 2.70; CI: 1.21-6.04), intraabdominal abscess (OR: 2.74; CI: 0.94-8.03), and wound infection (OR: 2.44; CI: 0.97-6.16). CONCLUSION: Preoperative biliary drainage increased postoperative infectious complications but it also decreased cardiovascular events. However, adequate preoperative biliary drainage is the key to decrease infectious complications.

Intratumor Heterogeneity of MYO18A and FBXW7 Variants Impact the Clinical Outcome of Stage III Colorectal Cancer.

Many studies failed to demonstrate benefit from the addition of targeted agents to current standard adjuvant FOLFOX chemotherapy in stage III colorectal cancer (CRC) patients. Intratumor heterogeneity may foster the resistant subclones and leads to cancer recurrence. Here, we built a cancer evolution model and applied machine learning analysis to identify potential therapeutic targets. Among 78 CRC cases, whole-genome (WGS) and deep targeted sequencing data generated from paired blood and primary tumor were used for phylogenetic tree reconstruction. Genetic alterations in the PI3K/AKT, and RTK oncogenic signaling pathways were commonly detected in founding clones. The dominant subclones frequently exhibited dysregulations in the TP53, FBXW7/NOTCH1 tumor suppression, and DNA repair pathways. Fourteen genetic mutations were simultaneously selected by random forest and LASSO methods. The logistic regression model had better accuracy (79%), precision (70%), and recall (65%) and area under the curve (AUC) (82%) for cancer recurrence prediction. Three genes, including MYO18A in the founding clone, FBXW7, and ATM in the dominant subclone, affected the prognosis were selected simultaneously by different feature sets. The in vitro studies, HCT-116 cells transfected with MYO18A siRNA demonstrated a significant reduction in cell migration activity by 20-40%. These results indicate that MYO18A plays a crucial role in the migration of human CRC cells. The cancer evolution model revealed the critical mutations in the founding and dominant subclones. They can be used to predict clinical outcomes and the development of novel therapeutic targets for stage III CRC.

Regulation of CD151 by hypoxia controls cell adhesion and metastasis in colorectal cancer.

PURPOSE: The first step of metastasis is the detachment of cancer cells from the surrounding matrix and neighboring cells; however, how cancer cells accomplish this process remains unclear. Thus, we aimed to investigate the underlying mechanism that controls the early event of metastasis. EXPERIMENTAL DESIGN: One hundred and thirty-seven paired colorectal carcinoma and normal colon tissues were examined by immunohistochemical staining and Western blot for the expression of CD151, a member of the tetraspanin family that plays important roles in cell adhesion and motility. The effect of CD151 on cancer cell adhesion was investigated under normoxia and hypoxia conditions. RESULTS: The level of CD151 was down-regulated in colon cancer compared with the paired normal counterparts. Expression of CD151 was negatively regulated by hypoxia inducible factor-1-dependent hypoxic stress. Suppression of CD151 by hypoxia caused the detachment of cancer cells from the surrounding matrix and neighboring cells whereas restoration of CD151 expression during reoxygenation facilitated the adhesion capacity. Clinical examination further showed that metastasized cancer cells expressed a greater level of CD151 compared with that of primary tumor. CONCLUSION: Regulation of CD151 by oxygen tension may play an important role in cancer metastasis by regulating the detachment from the primary site and homing in the secondary site.

Comprehensive assessment of HER2 alteration in a colorectal cancer cohort: from next-generation sequencing to clinical significance.

PURPOSE: Human epidermal growth factor receptor 2 (HER2) is an emerging therapeutic target in colorectal cancer (CRC). Currently, immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) have been used to determine HER2-positive CRCs; however, the clinical utility of next-generation sequencing (NGS)-based techniques for determining HER2 status in CRC has been limited. Here, we detail our experience regarding the assessment of HER2 alterations in a CRC cohort. MATERIALS AND METHODS: We prospectively enrolled 73 CRC patients who underwent surgery and received adjuvant oxaliplatin treatment. We then examined HER2 alterations using the Oncomine Comprehensive Assay version 1, as well as clinical outcomes, in this cohort. RESULTS: Using the NGS-based assay, HER2 copy number gains in 12 of 73 CRCs were determined to range from 2.74 to 92.62. Of these 12 tumors, 6 had HER2 high-level copy number gain (92.6, 57.9, 57.0, 52.0, 35.2, and 8.42) and were all defined as HER2-positive CRC using HERACLES Diagnostic Criteria. Nevertheless, other 6 patients with low-level copy number gain (ranging from 2.74 to 3.04) and the remaining 61 patients without increase in HER2 copy number were all HER2-negative. Among the 6 HER2-positive CRCs, KRAS and PIK3CA mutations were detected in 1 (17%; G13D) and 2 (33.3%; 1 Q546R and 1 H1047R) patients, respectively. Moreover, 2 of the 6 (33.3%) HER2-positive patients had recurrent disease, while one patient had a partial response after anti-HER2 therapy. CONCLUSION: NGS-based tools could assist in the simultaneous detection of HER2 and other genomic alterations in patients with CRC. Only CRCs with HER2 high-level copy number gain were HER2-postive by current diagnostic criteria.

The prognostic significance of RON and MET receptor coexpression in patients with colorectal cancer.

PURPOSE: Although Recepteur d'Origine Nantais (RON), a member of the MET receptor tyrosine kinase family, is overexpressed and constitutively active in some primary tumors and tumor cell lines, its expression pattern and clinical significance in colorectal cancer are not well documented. METHODS: By using immunohistochemical staining, we examined RON and MET expression in 135 colorectal cancer specimens and investigated the association of the immunoreactivity of both receptors with colorectal cancer clinical parameters and prognosis. RESULTS: We found moderate to strong expression in 99 cases (73 percent) for RON and 97 cases (72 percent) for MET. Univariate analysis showed that increased immunoreactivity of RON or MET was associated with shorter patient survival and that moderate to strong coexpression of both receptors was associated with a significantly worse prognosis. Multivariate Cox analysis showed that the risk of tumor recurrence for patients with high-RON/high-MET expression was approximately 11 times greater than for patients with low-RON/low-MET expression (P = 0.001). In addition, RON and MET expression levels were positively correlated (P

Serrated adenocarcinoma morphology in colorectal mucinous adenocarcinoma is associated with improved patient survival.

Colorectal mucinous adenocarcinoma (MAC) and serrated adenocarcinoma (SAC) share many characteristics, including right-side colon location, frequent mucin production, and various molecular features. This study examined the frequency of SAC morphology in MACs. We assessed the correlation of SAC morphology with clinicopathological parameters, molecular characteristics, and patient prognosis. Eighty-eight colorectal MACs were collected and reviewed for SAC morphology according to Makinen's criteria. We sequenced KRAS and BRAF, assessed CpG island methylator phenotype (CIMP) frequency, and analyzed DNA mismatch repair enzyme levels using immunohistochemistry in tumor samples. SAC morphology was observed in 38% of MACs, and was associated with proximal location (P=0.001), BRAF mutation (P=0.042), CIMP-positive status (P=0.023), and contiguous traditional serrated adenoma (P=0.019). Multivariate analysis revealed that MACs without both SAC morphology and CIMP-positive status exhibited 3.955 times greater risk of cancer relapse than MACs having both characteristics or either one (P=0.035). Our results show that two MAC groups with distinct features can be identified using Makinen's criteria, and suggest a favorable prognostic role for the serrated neoplastic pathway in colorectal MAC.