Mitosis Count Research Articles

Mitosis Detection for Invasive Breast Cancer Grading in Histopathological Images.

Histopathological grading of cancer not only offers an insight to the patients' prognosis but also helps in making individual treatment plans. Mitosis counts in histopathological slides play a crucial role for invasive breast cancer grading using the Nottingham grading system. Pathologists perform this grading by manual examinations of a few thousand images for each patient. Hence, finding the mitotic figures from these images is a tedious job and also prone to observer variability due to variations in the appearances of the mitotic cells. We propose a fast and accurate approach for automatic mitosis detection from histopathological images. We employ area morphological scale space for cell segmentation. The scale space is constructed in a novel manner by restricting the scales with the maximization of relative-entropy between the cells and the background. This results in precise cell segmentation. The segmented cells are classified in mitotic and non-mitotic category using the random forest classifier. Experiments show at least 12% improvement in F1 score on more than 450 histopathological images at 40× magnification.

Validation of mitosis counting by automated phosphohistone H3 (PHH3) digital image analysis in a breast carcinoma tissue microarray

Mitosis counting in H&E stained sections is the most informative constituent of the Nottingham histological grade in breast carcinoma prognosis. Phosphohistone H3 (PHH3) immunohistochemistry (IHC) is a highly specific marker of mitoses, with practical application in identifying mitoses in poorly fixed or distorted tissue and is of prognostic significance in breast carcinoma. Our aim was to assess methods of PHH3 IHC mitosis counting in a tissue microarray (TMA) of 2 mm cores from 36 resected breast carcinomas. Mitoses in H&E and PHH3 stained slides were manually scored by pathologist consensus and expressed as counts/2 mm. PHH3 stained cores were also evaluated by automated digital image analysis (DIA). Results were compared using Spearman correlation. A strong and significant correlation was observed between manual PHH3 and manual H&E mitotic counts (correlation = 0.81; p < 0.0001) and between automated PHH3 DIA and manual H&E mitotic counts (correlation = 0.79; p < 0.0001). More mitoses were identified with PHH3 IHC than with H&E. Manual and DIA PHH3 counts were strongly and significantly correlated (correlation = 0.83; p < 0.0001) and of similar absolute values. PHH3 DIA is a valid alternative to manual counting with potential application in breast cancer reporting and prognostication.

Immunostaining of phospho-histone H3 and Ki-67 improves reproducibility of recurrence risk assessment of gastrointestinal stromal tumors.

Gastrointestinal stromal tumors (GISTs) are the most common non-epithelial tumors in the digestive tract. Beyond surgery, therapeutic management depends on risk of recurrence. Risk evaluation of GIST takes into account location and size of the tumor, whether or not the tumor was intact or ruptured and mitotic index. The mitotic index lacks in intra- and interobserver reproducibility. In this study, we evaluated on 61 GISTs, the reproducibility of mitotic counting using classical hematoxylin-eosin-saffron (HES) staining, and its correlation with the mitotic count obtained through immunohistochemical staining for phospho-histone H3 (PHH3) and the proliferation index based upon Ki-67 immunostaining. Mitotic counts by HES and PHH3 staining and Ki-67 proliferation index were evaluated twice by three independent observers taking into account interpretation times per tumor for each technique. HES-based and PHH3-based mitotic counts and Ki-67 proliferation index correlated well and presented good intra- and interobserver reproducibility. PHH3 staining resulted in a slight but statistically significant difference of about two more mitotic figures per 5 mm(2) than the HES-based count, which might have put some borderline tumors in a different risk category. Immunohistochemical staining for PHH3 and Ki-67 allowed more rapid interpretation than mitotic counts based upon HES staining, but only PHH3 staining allows counting of mitoses. Immunostaining using anti-PHH3 and anti-Ki-67 antibodies will eventually provide improved recurrence risk stratification of GIST and may become effective ancillary tools in deciding on optimal therapeutic management.

Clinical pathological features and prognosis analysis of gastrointestinal stromal tumor: a series of 558 cases

To investigate and analysis the clinical and pathological characteristic of gastrointestinal stromal tumor (GIST) patients, and to clarify the factors that effect on prognosis. The clinical and pathological features and follow-up of GIST patients who received surgery in Peking Union Medical College Hospital from May 2002 to December 2013 were analyzed retrospectively. The prognosis was evaluated by univariate and multivariate analysis. Kaplan-Meier unvariate analysis and Log-rank test were used to compare the survival rates. Multivariate factors for survival were analyzed by Cox proportional hazards regression model. A total of 558 GIST patients were collected, including 284 males and 272 females. The high incidence was in the elderly and age of 50 to 70 years. Most of the primary tumors are located in stomach (303 cases), followed by the small intestine (118 cases). Surgical procedures included R0 resection in 517 cases, R1 resection in 4 cases, R2 or palliative resection in 37 cases. The recurrence risk was very low in 102 cases, low in 156 cases, moderate in 67 cases and high in 233 cases. Of all the patients, 495 cases completed the follow-up, the follow-up rate was 88.7%. Five year survival rate was 87.4%. Patients who took targeted therapy with moderate and high risk of recurrence had a better prognosis compared with not taking the drug. Univariate analysis revealed that the factors impacting the prognosis were age, tumor size, tumor site and mitotic count. Multivariate analysis showed that tumor size (P=0.01, RR=1.562, 95% CI: 1.452 to 15.664), location (P=0.01, RR=1.552, 95% CI:1.324 to 12.225), mitotic figures (P<0.01, RR=1.415, 95% CI: 2.126 to 7.968) and tumor rupture (P=0.01, RR=1.578, 95% CI: 1.543 to 15.892) were independent prognostic factors. R0 resection combined with targeted therapy is the best treatment of GIST. Tumor size, location, mitosis count and tumor rupture are independent prognostic factors of GIST patients.

Automatic detection of breast cancer mitotic cells based on the combination of textural, statistical and innovative mathematical features

Automatic grading systems based on histopathological slide images are applied to various types of cancers. To date, cancer scientists and researchers have conducted many experiments to find and evaluate new and innovative automatic cancer grading systems to accelerate their therapeutic diagnoses and ultimately to enable more efficient prognoses. The previously proposed automatic or computer-aided systems for breast cancer grading, including specializing mitosis counting, suffer from various shortcomings. The most important one is their low efficiency along with high complexity due to the huge amount of features. In this paper, three types of features with more flexibility and less complexity are employed. These features are: completed local binary pattern (CLBP) as textural features, statistical moment entropy (SME) and stiffness matrix (SM) as a mathematical model which includes geometric, morphometric and shape-based features. In the proposed automatic mitosis detection method, these three types of features are fused with each other. The SM feature comprises of characteristics which are to be extracted for reliable discrimination of mitosis objects from non-mitosis ones. The evaluations are applied over histology datasets A and H provided by the Mitos-ICPR2012 contest sponsors. Employing both a nonlinear radial basis function (RBF) kernel for support vector machine (SVM) and also random forest classifiers, leads to the best efficiencies among the other competitive methods which have been proposed in the past. The results are in the form of F-measure criterion which is a basis for bioinformatics assessments and evaluation.

Prognostic factors in solitary fibrous tumors of the pleura

BackgroundSolitary fibrous tumors of the pleura (SFTP) are rare neoplasms originating from submesothelial mesenchymal cells with fibroblastic differentiation. The clinical behavior of SFTPs is mostly benign; however, up to 20% of patients develop local recurrence and/or distant metastasis. Although different risk-stratification models have been described, definitive criteria to predict a malignant clinical course of SFTP are still lacking. MethodsIn a retrospective analysis at a single-institution, 25 patients with histologically proven SFTP were identified. Clinicopathologic and survival data were collected and pathologic sections reviewed. Different markers and risk-stratification models were correlated with disease- and overall-free survival by Kaplan–Meier analysis. ResultsOf 25 SFTP, 8 tumors (32%) were classified as malignant according to the World Health Organization criteria. Three patients (12%) developed recurrence. Cohort median follow-up was 28 mo, and median overall survival was 160 mo. Comparison of proliferation markers showed higher mitosis count per high-power field and MIB-1 labeling index (MIB) in malignant compared with nonmalignant SFTP. MIB was identified as a predictor for disease-free survival. Applying the previously reported classifications to categorize SFTP according to the probability to show malignant behavior, significant differences in disease-free survival were also present in our cohort. ConclusionsIn the present analysis of rare SFTP, previously proposed staging systems were applicable for prediction of disease-free survival. Independently of treatment, MIB was the only sole predictive marker. A prospective multi-institutional database could be helpful in establishing detailed predictive criteria in patients diagnosed with SFTP.

Mitosis detection in breast cancer pathology images by combining handcrafted and convolutional neural network features.

Breast cancer (BCa) grading plays an important role in predicting disease aggressiveness and patient outcome. A key component of BCa grade is the mitotic count, which involves quantifying the number of cells in the process of dividing (i.e., undergoing mitosis) at a specific point in time. Currently, mitosis counting is done manually by a pathologist looking at multiple high power fields (HPFs) on a glass slide under a microscope, an extremely laborious and time consuming process. The development of computerized systems for automated detection of mitotic nuclei, while highly desirable, is confounded by the highly variable shape and appearance of mitoses. Existing methods use either handcrafted features that capture certain morphological, statistical, or textural attributes of mitoses or features learned with convolutional neural networks (CNN). Although handcrafted features are inspired by the domain and the particular application, the data-driven CNN models tend to be domain agnostic and attempt to learn additional feature bases that cannot be represented through any of the handcrafted features. On the other hand, CNN is computationally more complex and needs a large number of labeled training instances. Since handcrafted features attempt to model domain pertinent attributes and CNN approaches are largely supervised feature generation methods, there is an appeal in attempting to combine these two distinct classes of feature generation strategies to create an integrated set of attributes that can potentially outperform either class of feature extraction strategies individually. We present a cascaded approach for mitosis detection that intelligently combines a CNN model and handcrafted features (morphology, color, and texture features). By employing a light CNN model, the proposed approach is far less demanding computationally, and the cascaded strategy of combining handcrafted features and CNN-derived features enables the possibility of maximizing the performance by leveraging the disconnected feature sets. Evaluation on the public ICPR12 mitosis dataset that has 226 mitoses annotated on 35 HPFs ([Formula: see text] magnification) by several pathologists and 15 testing HPFs yielded an [Formula: see text]-measure of 0.7345. Our approach is accurate, fast, and requires fewer computing resources compared to existent methods, making this feasible for clinical use.

Evaluation of high-risk clinicopathological indicators in gastrointestinal stromal tumors for prognosis and imatinib treatment outcome.

BackgroundAlthough the clinical benefit of imatinib adjuvant therapy for high-risk patients with gastrointestinal stromal tumor (GIST) has been proven, the recurrence rate still remains high. This study aimed to sub-divide high-risk GIST patients with some “very high-risk” factors for more precise prognostic indicator, and possible association with efficiency of imatinib adjuvant therapy.MethodsClinicopathological data were confirmed by pathological diagnosis and clinical records. Recurrence-free survivals (RFS) were evaluated in 370 GIST patients (212 cases as test cohort and 158 cases as validation cohort) and 48 high-risk GISTs with imatinib adjuvant therapy after R0 resection.ResultsMitosis count > 10/50 high-power fields (HPF) and serosal invasion are independent prognostic factors for RFS of GIST patients. Mitosis count > 10/50HPF and serosal invasion can sub-divide high-risk GIST patients effectively and significantly improve the area under the curve (AUC) of receiver operating characteristics (ROC) curve for prognostic indicator both in test and validation cohort. Patients with serosal invasion after R0 resection showed a poorer prognosis with imatinib adjuvant therapy.ConclusionsSub-division of high-risk GIST patients helps to more precisely predicting the prognosis. Serosal invasion may be an adverse predictive factor in high-risk patients and imatinib treatment outcome.

An interesting case of liposarcoma of the breast

Background: Primary liposarcomas of the breast are an extremely rare pathological entity. We report a case of liposarcoma in a 51-year-old woman who was treated at our Breast Cancer Center. Case report: The patient presented with a palpable, asymptomatic mass located in the lower inner quadrant of her right breast. She had no concomitant diseases at time of presentation beside obesity (BMI 31 kg/m2), her previous surgeries included cholecystectomy and she was a nonsmoker. At mammography, the lesion had smooth margins. Breast ultrasound showed an irregular structure of complex echogenicity measuring 2,4 × 2,0 × 1,6 cm (BI-RADS 5). Axillary lymph nodes were sonographically unremarkable. Ultrasound guided minimal-invasive core biopsy revealed a biphasic tumor of the phyllodes type with suspicious stroma. We conducted a breast conserving surgery; histopathological workup described a malignant phyllodes tumor of 21 mm diameter with a specific heterologous component identified as well differentiated liposarcoma; mitosis count was 21/10 HPF. Because of close margins (min. 1 mm) the patient received a subsequent wide excision which showed no further malignant lesion. The patient received no further therapy and is after one year disease-free. Conclusions: Well differentiated liposarcoma of the breast is a rare neoplasm. Precise classification of the tumor may represent a diagnostic challenge.

Unraveling tumor grading and genomic landscape in lung neuroendocrine tumors.

Currently, grading in lung neuroendocrine tumors (NETs) is inherently defined by the histological classification based on cell features, mitosis count, and necrosis, for which typical carcinoids (TC) are low-grade malignant tumors with long life expectation, atypical carcinoids (AC) intermediate-grade malignant tumors with more aggressive clinical behavior, and large cell NE carcinomas (LCNEC) and small cell lung carcinomas (SCLC) high-grade malignant tumors with dismal prognosis. While Ki-67 antigen labeling index, highlighting the proportion of proliferating tumor cells, has largely been used in digestive NETs for assessing prognosis and assisting therapy decisions, the same marker does not play an established role in the diagnosis, grading, and prognosis of lung NETs. Next generation sequencing techniques (NGS), thanks to their astonishing ability to process in a shorter timeframe up to billions of DNA strands, are radically revolutionizing our approach to diagnosis and therapy of tumors, including lung cancer. When applied to single genes, panels of genes, exome, or the whole genome by using either frozen or paraffin tissues, NGS techniques increase our understanding of cancer, thus realizing the bases of precision medicine. Data are emerging that TC and AC are mainly altered in chromatin remodeling genes, whereas LCNEC and SCLC are also mutated in cell cycle checkpoint and cell differentiation regulators. A common denominator to all lung NETs is a deregulation of cell proliferation, which represents a biological rationale for morphologic (mitoses and necrosis) and molecular (Ki-67 antigen) parameters to successfully serve as predictors of tumor behavior (i.e., identification of pathological entities with clinical correlation). It is envisaged that a novel grading system in lung NETs based on the combined assessment of mitoses, necrosis, and Ki-67 LI may offer a better stratification of prognostic classes, realizing a bridge between molecular alterations, morphological features, and clinical behavior.

Silent GH pituitary tumor: Diagnostic and therapeutic challenges

Silent GH pituitary tumors are characterized by the absence of clinical features of acromegaly, normal to slightly elevated GH and/or IGF-1 levels, as well as immunohistochemical expression of GH. The diagnostic and the therapeutic challenges of these “silent” GH tumors are illustrated in this case report, supported by a literature review. A 20-year-old woman presented with visual disturbances related to an invasive macroadenoma but without clinical and biological signs of GH hypersecretion. After two surgeries, a residual tumor remained in the right cavernous sinus. According to the recent classifications, the histopathological diagnosis was a sparsely GH-PRL atypical adenoma or invasive and proliferative (Ki-67 index: 4%) and p53 positive (1%) grade 2b tumor, with high expression (>75% of the cells) of somatostatin receptors type 2A and 5. From this case and the review of the literature, an invasive macroadenoma in young women requires: the preoperative determination of plasma GH and IGF-1, the immunohistochemical detection in the tumor of GH, PRL, somatostatin receptor expression and the evaluation of the proliferation (mitoses count, Ki-67 and p53 indexes). The suspicion of an aggressive behavior needs a particular follow-up. In the case of tumor remnant, a postoperative treatment such as radiotherapy and/or somatostatin analogs must be considered.

Hormonal pretreatment preserves liver regenerative capacity and minimizes inflammation after partial hepatectomy

Introduction. The treatment of brain dead donors with combined hormonal resuscitation protocols, including methylprednisolone (MP) and triiodothyronine (T3), among others, was developed to increase the viability and function of transplanted organs, primarily heart and lung. Even when it has regarded successful results in term of donors and organs recovery, its effects over specific parameters in organs like the liver are unknown.Material and methods. Male Sprague-Dawley rats were pretreated with MP (0.34 mg/kg) and/or T3 (0.05 mg/kg) or their vehicles, and then subjected to partial hepatectomy of 70%. Three experimental groups and their respective controls were conformed: a. T3; b. NaOH; c. MP; d. vMP; e. MP+T3 and f. vMP+NaOH. The groups were evaluated at 0, 16, 24, 72 and 120 h post surgery. The effects of this protocol on regeneration, liver mass recovery, liver injury, oxidative stress and liver function were analyzed.Results. MP+T3 pretreatment does not deleteriously affect liver regeneration after partial hepatectomy, as shown in the curve of total mass recovery, Ki67 staining and mitosis counting, and does not alter liver function. In addition, the treatment modestly decreases oxidative stress and liver injury, as evidenced by transaminases levels, histological analysis and oxidized proteins content.Conclusion. These preclinical results indicate that MP+T3 is harmless for liver tissue regeneration post hepatectomy and additionally exhibits anti-inflammatory and antioxidant effects; therefore, it would not be contraindicated for the treatment of multiorgan donors in brain death and particularly, if the occurrence of small for size syndrome is suspected.

Correlation of 18F-fluorodeoxyglucose uptake with histopathological prognostic factors in breast carcinoma

This study investigated the prognostic value of [18F]-fluorodeoxyglucose (18F-FDG) uptake in breast carcinomas by comparing 18F-FDG PET/computed tomography (CT) images with histopathological and immunohistochemical prognostic factors. This study included 136 women and four men with positive biopsy breast carcinomas who underwent 18F-FDG PET/CT imaging for initial staging. Maximum standardized uptake values (SUVmax) and tumour-to-background SUVmax ratios were calculated and compared with histopathological and immunohistochemical tumour characteristics, patient properties and axillary lymph node involvement. Calculations of SUVmax for men were performed separately. For the tumours in women, the mean SUVmax was 10.06±6.91 and the median SUVmax was 9.05 (0.7-35.0). Primary tumour 18F-FDG uptake and tumour-to-background SUVmax ratios were correlated with tumour size (P<0.001), histological type (P<0.001), histological grade (P=0.004), pleomorphism (P=0.010), mitosis count (P<0.001), lymphatic invasion (P=0.009), necrosis (P=0.005), oestrogen negativity (P=0.004), high Ki-67 level (P<0.001), axillary lymph node involvement (P<0.001) and triple negativity (P=0.002). High Ki-67 level (odds ratio=16; 95% confidence interval=1.6-160; P=0.016) and tumour size (odds ratio=4; 95% confidence interval=1.5-11; P=0.007) were determining factors for high 18F-FDG uptake values. Other clinicopathological and immunohistopathological parameters including progesterone receptor (P=0.211), CerbB2 overexpression (P=0.170), perineural invasion (P=0.053), intratumoural calcification (P=0.438), desmoplasia (P=0.112), tubular formation (P=0.768) and age (P=0.675) were not significantly correlated with 18F-FDG uptake. No significant relationship was observed between the tumour/contralateral breast SUVmax ratio and mitotic count, oestrogen receptor status or triple negativity. F-FDG uptake may serve as a prognostic indicator for biological behaviour in breast tumours.

Can Proliferation Biomarkers Reliably Predict Recurrence in World Health Organization 2003 Defined Endometrial Stromal Sarcoma, Low Grade?

An estimated 1500–3000 invasive Endometrial Stromal Sarcomas (ESS) cases annually occur worldwide. Before 2003, ESS was divided as low and high grade ESS based on mitotic activity. In 2003 the WHO changed the names, excluded mitoses and made nuclear atypia and necrosis the essential diagnostic criteria to distinguish ESS, Low Grade (ESS-LG, recurrence-free survival >90%) and Undifferentiated Endometrial Sarcoma (UES, poor prognosis). We have evaluated in WHO2003 defined ESS-LG whether proliferation biomarkers predict recurrence. Using survival analysis, the prognostic value of classical mitosis counts (Mitotic Activity Index, MAI) in haematoxyllin-eosin (H&E) sections, and immunohistochemical proliferation biomarkers (Ki-67 and PhosphoHistone-3 (PPH3)) were examined in 24 invasive endometrial stromal sarcomas. Three of 24 (12.5%) ESS-LG recurred. The MAI, PPH3 and Ki-67 were all prognostic (P = 0.001, 0.002 and 0.03). MAI values were >3 in the recurrent cases, but never exceeded 10 (the classical threshold for low and high grade). Non-recurrent cases had 0≤MAI≤3. PPH3 and Ki67 counts can be easier to perform than MAI and therefore helpful in the diagnosis of ESS, Low Grade. In conclusion, in this small study of WHO2003 defined ESS-LG, high levels of proliferation as measured by MAI, PPH3 and Ki-67 are predictive of recurrence. Larger studies are required to confirm these results.

Overexpression of High Mobility Group A1 Protein in Human Uveal Melanomas: Implication for Prognosis

There is increasing evidence that the high mobility group A1 (HMGA1) protein, which functions as a transcriptional master regulator, plays critical roles in tumor progression. We evaluated HMGA1 expression in 89 primary uveal melanomas (UM) by immunohistochemistry to determine the clinicopathological and prognostic value of HMGA1 in UM after adjusting for other prognostic variables. Nuclear expression of HMGA1 was detected in 44% UMs. High expression levels of HMGA1 were more frequent in UMs with high levels of epithelioid cell pattern, mitoses count, and Ki67 labeling index (P = 0.025, P<0.0001, P = 0.0018; respectively), and HMGA1 expression levels were directly correlated with Ki67 labeling indexes and mitoses counts (R = 0.31, P <0.0001; R = 0.27, P<0.0068; respectively). High expression of HMGA1 was also independently associated with an increased risk of distant metastases as determined using the Cox proportional hazards regression model (multivariate hazard ratio: 3.44; 95% confidence interval: 1.56–7.60; log rank P = 0.0022). Moreover, high HMGA1 expression was associated with shorter UM-specific survival (multivariate hazard ratio: 2.41; 95% confidence interval: 1.10–5.53; log rank P = 0.041). These findings suggest that high levels of HMGA1 are associated with adverse clinical outcomes in UM patients and that further evaluation of HMGA1 as a potential therapeutic target in UM is warranted.

Are we counting mitoses correctly?

The number of mitotic figures in a predefined area is essential in pathologic evaluation for most tumors. This information sometimes provides clues in differentiating neoplastic lesions from nonneoplastic ones and sometimes in defining and grading of the tumors as well as prognosticating expected lifetime of the patient. As a generally accepted concept, scanning a certain number of consecutive nonoverlapping areas that are rich in viable tumor cells is required. Invasion fronts or the periphery of the tumors is preferred for counting mitosis. The target area to be counted for mitotic activity for various tumors is standardized as the number of mitosis in an established number of high-power fields. However, suggested mitotic counts, which constitute the basis of these studies, were obtained via the old microscopes, which usually had narrower visual fields than the state-of-the-art microscopes. Because the visual fields of the present microscopes provide larger areas compared with the older ones, corrections in mitosis counting are needed to make them compatible with the criteria, which had been put forward in the original reference studies.

Automated mitosis detection using texture, SIFT features and HMAX biologically inspired approach

Context: According to Nottingham grading system, mitosis count in breast cancer histopathology is one of three components required for cancer grading and prognosis. Manual counting of mitosis is tedious and subject to considerable inter- and intra-reader variations. Aims: The aim is to investigate the various texture features and Hierarchical Model and X (HMAX) biologically inspired approach for mitosis detection using machine-learning techniques. Materials and Methods: We propose an approach that assists pathologists in automated mitosis detection and counting. The proposed method, which is based on the most favorable texture features combination, examines the separability between different channels of color space. Blue-ratio channel provides more discriminative information for mitosis detection in histopathological images. Co-occurrence features, run-length features, and Scale-invariant feature transform (SIFT) features were extracted and used in the classification of mitosis. Finally, a classification is performed to put the candidate patch either in the mitosis class or in the non-mitosis class. Three different classifiers have been evaluated: Decision tree, linear kernel Support Vector Machine (SVM), and non-linear kernel SVM. We also evaluate the performance of the proposed framework using the modified biologically inspired model of HMAX and compare the results with other feature extraction methods such as dense SIFT. Results: The proposed method has been tested on Mitosis detection in breast cancer histological images (MITOS) dataset provided for an International Conference on Pattern Recognition (ICPR) 2012 contest. The proposed framework achieved 76% recall, 75% precision and 76% F-measure. Conclusions: Different frameworks for classification have been evaluated for mitosis detection. In future work, instead of regions, we intend to compute features on the results of mitosis contour segmentation and use them to improve detection and classification rate.

Unique histopathologic features of brain metastases from hepatocellular carcinoma.

169 Background: In the era of anti-angiogenic therapy as treatment for advanced hepatocellular carcinoma (HCC), the incidence and importance of brain metastases are increasing. We aimed to study their histopathologic features. Methods: We searched for all patients who were diagnosed to have HCC with brain metastasis from 1999 to 2010 at National Taiwan University Hospital, Taipei, Taiwan. Patients who had HCC with lung metastasis were also included for comparison. Patients with available tissues of both primary and metastatic tumors were enrolled in this study. Tumor slides from paired primary and metastatic HCCs were stained by H and E, and immunohistochemically stained for CK7, p53, Ki67, vimentin, Hes1, and c-Met. The expressions of CK7, p53, and vimentin were graded according to percentages of positive staining, but those of Hes1 and c-Met were recorded as an H score, which was defined as intensity (0, 1, 2, or 3) × percentages of positive staining. Results: A total of 14 patients had available tumor tissues of both primary and metastatic brain tumors. Another 21 patients had tumor tissues of both primary and metastatic lung tumors. The metastatic brain tumors, comparing to the metastatic lung tumors, had significantly more bizarre dilated vessels (86% vs. 14%, p &lt; 0.001), hyaline globules (50% vs. 5%, p = 0.003), higher Hes1 H scores (mean, 245 vs. 131, p = 0.001), and lower c-Met H scores (mean, 15.4 vs. 38.1, p = 0.046). Tumor necrosis also tended to be more common among metastatic brain tumors (93% vs. 62%, p = 0.056). On the contrary, the above differences were not identified between the primary tumors which later developed brain metastasis and those which later developed lung metastasis. When disease progressed from primary liver to brain metastasis, mitosis counts (p = 0.034) and bizarre dilated vessels (p = 0.020) significantly increased, and necrosis (p = 0.059) tended to be more common. Conclusions: Metastatic brain tumors from HCC had unique histopathologic features compared to primary liver tumors or lung metastases. The increased Hes1 expression and decreased c-Met expression in HCC brain metastasis should be further explored. (This study was supported by the grant of NSC101-2314-B-002 -141, 100CAP1020-2 &amp; NTUH.101-N1965.)

Mitosis detection using generic features and an ensemble of cascade adaboosts

Context: Mitosis count is one of the factors that pathologists use to assess the risk of metastasis and survival of the patients, which are affected by the breast cancer. Aims: We investigate an application of a set of generic features and an ensemble of cascade adaboosts to the automated mitosis detection. Calculation of the features rely minimally on object-level descriptions and thus require minimal segmentation. Materials and Methods: The proposed work was developed and tested on International Conference on Pattern Recognition (ICPR) 2012 mitosis detection contest data. Statistical Analysis Used: We plotted receiver operating characteristics curves of true positive versus false positive rates; calculated recall, precision, F-measure, and region overlap ratio measures. Results: We tested our features with two different classifier configurations: 1) An ensemble of single adaboosts, 2) an ensemble of cascade adaboosts. On the ICPR 2012 mitosis detection contest evaluation, the cascade ensemble scored 54, 62.7, and 58, whereas the non-cascade version scored 68, 28.1, and 39.7 for the recall, precision, and F-measure measures, respectively. Mostly used features in the adaboost classifier rules were a shape-based feature, which counted granularity and a color-based feature, which relied on Red, Green, and Blue channel statistics. Conclusions: The features, which express the granular structure and color variations, are found useful for mitosis detection. The ensemble of adaboosts performs better than the individual adaboost classifiers. Moreover, the ensemble of cascaded adaboosts was better than the ensemble of single adaboosts for mitosis detection.

Automated mitosis detection in histopathology using morphological and multi-channel statistics features

Context: According to Nottingham grading system, mitosis count plays a critical role in cancer diagnosis and grading. Manual counting of mitosis is tedious and subject to considerable inter- and intra-reader variations. Aims: The aim is to improve the accuracy of mitosis detection by selecting the color channels that better capture the statistical and morphological features, which classify mitosis from other objects. Materials and Methods: We propose a framework that includes comprehensive analysis of statistics and morphological features in selected channels of various color spaces that assist pathologists in mitosis detection. In candidate detection phase, we perform Laplacian of Gaussian, thresholding, morphology and active contour model on blue-ratio image to detect and segment candidates. In candidate classification phase, we extract a total of 143 features including morphological, first order and second order (texture) statistics features for each candidate in selected channels and finally classify using decision tree classifier. Results and Discussion: The proposed method has been evaluated on Mitosis Detection in Breast Cancer Histological Images (MITOS) dataset provided for an International Conference on Pattern Recognition 2012 contest and achieved 74% and 71% detection rate, 70% and 56% precision and 72% and 63% F-Measure on Aperio and Hamamatsu images, respectively. Conclusions and Future Work: The proposed multi-channel features computation scheme uses fixed image scale and extracts nuclei features in selected channels of various color spaces. This simple but robust model has proven to be highly efficient in capturing multi-channels statistical features for mitosis detection, during the MITOS international benchmark. Indeed, the mitosis detection of critical importance in cancer diagnosis is a very challenging visual task. In future work, we plan to use color deconvolution as preprocessing and Hough transform or local extrema based candidate detection in order to reduce the number of candidates in mitosis and non-mitosis classes.