Assuntos

RESUMO

BACKGROUND AND OBJECTIVE: Cell segmentation in bright-field histological slides is a crucial topic in medical image analysis. Having access to accurate segmentation allows researchers to examine the relationship between cellular morphology and clinical observations. Unfortunately, most segmentation methods known today are limited to nuclei and cannot segment the cytoplasm. METHODS: We present a new network architecture Cyto R-CNN that is able to accurately segment whole cells (with both the nucleus and the cytoplasm) in bright-field images. We also present a new dataset CytoNuke, consisting of multiple thousand manual annotations of head and neck squamous cell carcinoma cells. Utilizing this dataset, we compared the performance of Cyto R-CNN to other popular cell segmentation algorithms, including QuPath's built-in algorithm, StarDist, Cellpose and a multi-scale Attention Deeplabv3+. To evaluate segmentation performance, we calculated AP50, AP75 and measured 17 morphological and staining-related features for all detected cells. We compared these measurements to the gold standard of manual segmentation using the Kolmogorov-Smirnov test. RESULTS: Cyto R-CNN achieved an AP50 of 58.65% and an AP75 of 11.56% in whole-cell segmentation, outperforming all other methods (QuPath 19.46/0.91%; StarDist 45.33/2.32%; Cellpose 31.85/5.61%, Deeplabv3+ 3.97/1.01%). Cell features derived from Cyto R-CNN showed the best agreement to the gold standard (D¯=0.15) outperforming QuPath (D¯=0.22), StarDist (D¯=0.25), Cellpose (D¯=0.23) and Deeplabv3+ (D¯=0.33). CONCLUSION: Our newly proposed Cyto R-CNN architecture outperforms current algorithms in whole-cell segmentation while providing more reliable cell measurements than any other model. This could improve digital pathology workflows, potentially leading to improved diagnosis. Moreover, our published dataset can be used to develop further models in the future.

Assuntos

RESUMO

Segmentation of tumor regions in H &E-stained slides is an important task for a pathologist while diagnosing different types of cancer, including oral squamous cell carcinoma (OSCC). Histological image segmentation is often constrained by the availability of labeled training data since labeling histological images is a highly skilled, complex, and time-consuming task. Thus, data augmentation strategies become essential to train convolutional neural networks models to overcome the overfitting problem when only a few training samples are available. This paper proposes a new data augmentation strategy, named Random Composition Augmentation (RCAug), to train fully convolutional networks (FCN) to segment OSCC tumor regions in H &E-stained histological images. Given the input image and their corresponding label, a pipeline with a random composition of geometric, distortion, color transfer, and generative image transformations is executed on the fly. Experimental evaluations were performed using an FCN-based method to segment OSCC regions through a set of different data augmentation transformations. By using RCAug, we improved the FCN-based segmentation method from 0.51 to 0.81 of intersection-over-union (IOU) in a whole slide image dataset and from 0.65 to 0.69 of IOU in a tissue microarray images dataset.

Assuntos

Assuntos

RESUMO

OBJECTIVE: Oral Squamous Cell Carcinoma (OSCC) is conventionally treated by surgical resection, and positive surgical margins strongly increase local recurrence and decrease survival. This study aimed to evaluate whether a Three-Dimensional Segmentation (3DS) image of OSCC confers advantage over Multiplanar Reconstruction (MPR) of OSCC using images of computed tomography scan in surgical planning of tumor resection. METHODS: Twenty-six patients with locally advanced OSCC had tumor morphology and dimensions evaluated by MPR images, 3DS images, and Surgical Pathology Specimen (SPS) analyses (gold standard). OSCC resection was performed with curative intent using only MPR images. RESULTS: OSCC morphology was more accurately assessed by 3DS than by MPR images. Similar OSCC volumes and dimensions were obtained when MPR images, 3DS images and SPS measurements were considered. Nevertheless, there was a strong correlation between the OSCC longest axis measured by 3DS and SPS analyses (ICC = 0.82; 95% CI 0.59‒0.92), whereas only a moderate correlation was observed between the longest axis of OSCC measured by MPR images and SPS analyses (ICC = 0.51; 95% CI 0.09‒0.78). Taking only SPS with positive margins into account, MPR images and 3DS images underestimated the tumor's longest axis in eight out of 11 (72.7%) and 5 out of the 11 (45.5%) cases, respectively. CONCLUSION: Our data present preliminary evidence that 3DS model represents a useful tool for surgical planning of OSCC resection, but confirmation in a larger cohort of patients is required. LEVEL OF EVIDENCE: Laboratory study.

Assuntos

RESUMO

OBJECTIVE: To evaluate the prognostic impact of clinical and pathological variables and patterns of recurrence in patients with locally advanced cervical cancer with pelvic lymph node involvement (stage IIIC1 according to the 2018 FIGO Staging System). METHODS: We retrospectively analyzed 62 patients with locally advanced cervical cancer treated with curative intent with radiotherapy associated with chemotherapy in AC Camargo Cancer Center from January 2007 to December 2018. RESULTS: Lymph node involvement was assessed by CT, MRI and positron emission tomography (PET)/CT in 28 (45.2%), 20 (32.3%) and 14 (22.6%) patients, respectively. The median tumor size was 5.0 cm and 72.6% of cases were squamous cell carcinomas. The median number of positive pelvic lymph nodes was three, and the median size of lymph nodes was 24 mm. Twenty-two (35.5%) patients had recurrence and 50% had only one site of recurrence. The sites of recurrence were pelvic, para-aortic and distant in 12 (19.4%), 6 (9.7%) and 16 (25.8%) patients, respectively. The 3 year overall and disease-free survival were 70.8% and 64.6%, respectively. Patients with adenocarcinoma had worse disease-free survival (HR 2.38; 95% CI 1.01 to 5.60; p=0.047) and overall survival (HR 2.99; 95% CI 1.14 to 7.75; p=0.025) compared with squamous cell carcinoma. In multivariate analysis, metastatic pelvic lymph node size of >2.5 cm (HR 4.38; 95% CI 1.65 to 11.6; p=0.003) and incomplete response to radiotherapy (HR 5.14; 95% CI 1.60 to 16.4; p=0.006) maintained the negative impact for overall survival. CONCLUSIONS: We found that pelvic lymph node size and incomplete response to radiotherapy negatively impact overall survival in patients with advanced cervical cancer with pelvic lymph node involvement. This finding may help to stratify risk in this group of patients.

Assuntos

RESUMO

PURPOSE: To evaluate prevalence and diagnostic performance of three colposcopic images to diagnose squamous and glandular cervical precursor neoplasias. METHODS: Cross-sectional study, conducted through analysis of stored digital colposcopic images. To evaluate the diagnostic performance of three images, herein named grouped glands, aceto-white villi, and atypical vessels, for detection of adenocarcinoma in situ (AIS) and cervical squamous intraepithelial neoplasias (CIN) grades 2 and 3, calculations of sensitivity, specificity, accuracy, positive likelihood ratio, receiver operating characteristic (ROC) curve, and area under the curve (AUC) were made, with their respective 95% confidence intervals. RESULTS: Grouped glands, aceto-white villi, and atypical vessels images had: prevalence of 21.3, 53.8, and 33.8% in patients with AIS, and 16.2, 19.5, and 9.3% in those with CIN 2 and 3; for the diagnosis of AIS, sensitivity of 21.3, 53.8, and 33.8%, specificity of 89.8, 95.2, and 94.9%, accuracy of 76.6, 87.2, and 83.1%, positive likelihood ratio of 2.1, 11.2, and 6.6, and AUC of 0.55, 0.74, and 0.64; for the diagnosis of CIN 2 and 3, sensitivity of 16.2, 19.5, and 9.3%, specificity of 89.8, 95.2, and 94.9%, accuracy of 39.4, 43.4, and 36.3%, positive likelihood ratio of 1.6, 4.1, and 1, 8, and AUC of 0.53, 0.57, and 0.52, respectively. CONCLUSION: Prevalence and accuracy of the three images were higher for the diagnosis of glandular than squamous cervical precursor neoplasias. Sensitivity, specificity, positive likelihood, and AUC of aceto-white villi and atypical vessels images were higher for the diagnosis of glandular than squamous cervical precursor neoplasias.

Assuntos

RESUMO

Primary squamous cell carcinoma of the colon is extremely rare. The etiology is poorly understood, and currently, there are different hypotheses about the origin of this malignant neoplasm. Here, we report a case of an 87-year-old male with a moderately-differentiated nonkeratinizing squamous cell carcinoma of the colon.

Assuntos

Assuntos

Assuntos