Computer aided Breast lesions classification system using digitized fine needle aspirate images

Finding breast cancer early and accurately is very important for helping patients live longer. We use the Wisconsin Diagnostic Breast Cancer (WDBC) dataset to look at two popular machine learning classifiers: Support Vector Machines (SVM) and k-nearest neighbors (KNN). Our work combines theory with practice by including things like data preprocessing exploratory feature evaluation, model configuration, hyperparameter, and validation through cross-validation techniques. We use common evaluation metrics like accuracy, precision, recall, F1-score, and Area Under the Curve (AUC) to judge performance. When we did experiments with both a 70:30 train-test split and 10-fold cross-validation, we found that SVM did a little better (~96.3% accuracy) than compared to KNN, which has an accuracy of about 96.5% and AUC values of about 0.96 and 0.968, respectively. SVM is better at dealing with high-dimensional data and making maximum-margin decision boundaries, but KNN is still appealing because it is simple and works well on datasets of moderate size. Even though both methods have their pros and cons, they also have some problems that are talked about, as well as ways they could be improved and how they could be used in the clinic. This paper evaluates and compares the theoretical and practical performance of SVM and KNN using the WDBC dataset. We include preprocessing, careful evaluation metrics, and hyperparameter optimization to provide an empirical comparison of both models. The study also discusses important trade-offs and real-world factors for using these algorithms in medical screening contexts.

The study presents a method for iterative parameter tuning of tree ensemble-based models using Bayesian hyperparameter tuning for states prediction, using breast cancer as an example. The proposed method utilizes three different datasets, including the Wisconsin Diagnostic Breast Cancer (WDBC) dataset, the Surveillance, Epidemiology, and End Results (SEER) breast cancer dataset, and the Breast Cancer Coimbra dataset (BCCD), and implements tree ensemble-based models, specifically AdaBoost, Gentle-Boost, LogitBoost, Bag, and RUSBoost, for breast cancer prediction. Bayesian optimization was used to tune the hyperparameters of the models iteratively, and the performance of the models was evaluated using several metrics, including accuracy, precision, recall, and f1-score. Our results show that the proposed method significantly improves the performance of tree ensemble-based models, resulting in higher accuracy, precision, recall, and f1-score. Compared to other state-of-the-art models, the proposed method is more efficient. It achieved perfect scores of 100% for Accuracy, Precision, Recall, and F1-Score on the WDBC dataset. On the SEER BC dataset, the method achieved an accuracy of 95.9%, a precision of 97.6%, a recall of 94.2%, and an F1-Score of 95.9%. For the BCCD dataset, the method achieved an accuracy of 94.7%, a precision of 90%, a recall of 100%, and an F1-Score of 94.7%. The outcomes of this study have important implications for medical professionals, as early detection of breast cancer can significantly increase the chances of survival. Overall, this study provides a valuable contribution to the field of breast cancer prediction using machine learning.

Breast cancer death rates are higher than any other cancer in American women. Machine learning-based predictive models promise earlier detection techniques for breast cancer diagnosis. However, making an evaluation for models that efficiently diagnose cancer is still challenging. In this work, we proposed data exploratory techniques (DET) and developed four different predictive models to improve breast cancer diagnostic accuracy. Prior to models, four-layered essential DET, e.g., feature distribution, correlation, elimination, and hyperparameter optimization, were deep-dived to identify the robust feature classification into malignant and benign classes. These proposed techniques and classifiers were implemented on the Wisconsin Diagnostic Breast Cancer (WDBC) and Breast Cancer Coimbra Dataset (BCCD) datasets. Standard performance metrics, including confusion matrices and K-fold cross-validation techniques, were applied to assess each classifier’s efficiency and training time. The models’ diagnostic capability improved with our DET, i.e., polynomial SVM gained 99.3%, LR with 98.06%, KNN acquired 97.35%, and EC achieved 97.61% accuracy with the WDBC dataset. We also compared our significant results with previous studies in terms of accuracy. The implementation procedure and findings can guide physicians to adopt an effective model for a practical understanding and prognosis of breast cancer tumors.

The purpose of the proposed study was to provide a solution to the Wisconsin Diagnostic Breast Cancer (WDBC) classification problem. The WDBC dataset, provided by the University of Wisconsin Hospital, was derived from a group of images using Fine Needle Aspiration (biopsies) of the breast. An ensemble of Support Vector Machines (SVM's) was employed in this study. Support vectors with linear, polynomial and RBF kernel functions were trained using a fraction of the WDBC dataset as a training set. The five top performing models were recruited into the ensemble. The classification was then carried out using the majority opinion of the ensemble. The SVM ensemble successfully classified more than 99 percent of the testing data and in the process yielded 100 percent benign tumor prediction accuracy. Keywords: Support Vector Machines; Breast Cancer; Machine Learning, Ensemble, Classifiers

Support vector machine (SVM) is one of the most effective classification methods for cancer detection. The efficiency and quality of a SVM classifier depends strongly on several important features and a set of proper parameters. Here, a series of classification analyses, with one set of photoacoustic data from ovarian tissues ex vivo and a widely used breast cancer dataset- the Wisconsin Diagnostic Breast Cancer (WDBC), revealed the different accuracy of a SVM classification in terms of the number of features used and the parameters selected. A pattern recognition system is proposed by means of SVM-Recursive Feature Elimination (RFE) with the Radial Basis Function (RBF) kernel. To improve the effectiveness and robustness of the system, an optimized tuning ensemble algorithm called as SVM-RFE(C) with correlation filter was implemented to quantify feature and parameter information based on cross validation. The proposed algorithm is first demonstrated outperforming SVM-RFE on WDBC. Then the best accuracy of 94.643% and sensitivity of 94.595% were achieved when using SVM-RFE(C) to test 57 new PAT data from 19 patients. The experiment results show that the classifier constructed with SVM-RFE(C) algorithm is able to learn additional information from new data and has significant potential in ovarian cancer diagnosis.

The association between insulin resistance (IR) and ovarian neoplasm is little known. The present study attempted to investigate the difference in clinicopathological characteristics, metabolic parameters, and IR prevalence between benign and malignant ovarian neoplasms. The cross-sectional study involved 52 non-diabetic women with benign (n=27) and malignant (n=25) diagnoses in a tertiary hospital in Indonesia. Fasting insulin level (FIL), homeostatic model assessment of IR and β-cell dysfunction (HOMA-IR and HOMA-β), fasting IR index (FIRI), and quantitative insulin sensitivity check index (QUICKI) were used as surrogate markers to evaluate IR. Parametric and nonparametric statistical tests were employed to analyze the different parameters between the two groups. Pearson or Spearman's rank test assessed the correlation between markers and clinical variables. Results revealed that patients with benign neoplasms were younger than those with malignant neoplasms (38.63 vs. 47.40 years; P=0.003) and had a higher median body mass index (BMI) than their counterparts (22.98 vs. 18.61 kg/m2; P=0.014). Different characteristics between benign and malignant neoplasm cases were found in menopausal status, ovary side affected, systolic blood pressure, and BMI classes. Endometrial cysts and mucinous carcinoma were the most often diagnosed benign and malignant neoplasms. Malignant neoplasms had a lower median HOMA-β score than benign neoplasms (49.33 vs. 75.79; P=0.011), indicating more severe β-cell dysfunction. No significant difference was observed in the prevalence of IR between benign and malignant ovarian neoplasms for the following values of each marker: FIL (25.9% vs. 12.0%), HOMA-IR (37.0% vs. 28.0%), FIRI (51.9% vs. 48.0%) and QUICKI (81.5% vs. 92.0%). The indicators of FIL, HOMA-IR, HOMA-β, FIRI, and QUICKI correlated with each other and confirmed the reliability of these surrogate markers for measuring IR status in ovarian neoplasms. In brief, benign ovarian neoplasms tended to have more IR when compared with malignant ovarian neoplasms. However, this difference was not statistically significant.

We evaluated ultrasonographic features of superficial epidermoid tumour with a focus on strain elastography (SE) features that will help in the differential diagnosis of epidermoid tumour from other benign and malignant soft-tissue tumours. We retrospectively evaluated ultrasonographic and SE data of 103 surgically confirmed superficial soft-tissue tumours and tumour-like lesions: 29 cases of epidermoid tumour, 46 cases of other benign tumours and 28 cases of malignant tumour. SE and B-mode imaging were performed at the same time. SE characteristics were assigned into four grades (1-4) according to their elasticity. Interobserver agreement for the four SE scores between the two radiologists was analysed using kappa statistics. We classified each SE finding as a hard lesion (SE Score 3-4) or soft lesion (SE Score 1-2) and compared these findings using the χ(2) test to identify whether a significant difference in mass hardness existed among epidermoid tumour, other benign tumour and malignant tumour. Overall interobserver agreement according to the four SE scores was moderate (κ = 0.540), and overall agreement for the hardness [soft (Score 1-2) or hard (Score 3-4)] was almost perfect (κ = 0.825). Malignant tumours showed higher SE scores (3-4, hard nature) than did epidermoid tumour or other benign soft-tissue tumours. There were no differences in SE score between epidermoid tumour and other benign tumours. Superficial epidermoid tumour exhibits a softer nature than does malignant tumour but does not have a different SE pattern from other benign tumours. SE features of epidermoid tumour might be helpful in differentiating from other benign and malignant tumours.

Objective:The study aims to determine whether apparent diffusion coefficient (ADC) can help differentiate benign and malignant bone tumors.Materials and Methods:From January 2012 to February 2013, we prospectively included 26 patients. Of these 15 patients were male and 11 were female; ranging in age from 8 to 76 years (mean age, 34.5 years). Diffusion-weighted magnetic resonance (MR) imaging was obtained with a single-shot echo-planar imaging sequence using a 1.5T MR scanner. We grouped malignant lesions as primary, secondary, and primary tumor with chondroid matrix. The minimum ADC was measured in the tumors and mean minimum ADC values were selected for statistical analysis. ADC values were compared between malignant and benign tumors using the Mann-Whitney U-test and receiver operating curve analysis were done to determine optimal cut-off values.Results:The mean ADC values from the area with lowest ADC values of benign and malignant tumors were 1.99 ± 0.57 × 10−3 mm2/s and 1.02 ± 1.0 × 10−3 mm2/s, respectively. The mean minimum ADC values of benign and malignant tumors were statistically different (P = 0.029). With cut-off value of 1.37 (10−3 mm2/s), sensitivity was 77.8% and specificity was 82.4%, for distinguishing benign and malignant lesion. Benign and secondary malignant tumors showed statistically significant difference (P = 0.002). There was some overlap in ADC values between benign and malignant tumors. The mean minimum ADC values of benign and malignant chondroid tumors were high. Giant cell tumor, non-ossifying fibroma and fibrous dysplasia showed lower ADC values.Conclusion:Although there is some overlap, ADC values of benign and malignant bone tumors seem to be different. Further studies with larger patient groups are needed to find an optimal cut-off ADC value.

Worldwide, cancer is the most frequent cause of passing away for women. Any development in predicting and diagnosing cancer is crucial for a healthy life. As such, vital cancer accuracy in predicting patients' survival parameters and treatment aspects is necessary. Machine learning methods significantly impact breast cancer diagnosis and early diagnosis. This study aims to increase prediction accuracy using a novel statistical feature selection technique. This article examines the classification test accuracy, standard data precision, and the process performance of multiple machine learning (ML) algorithms, such as Random; using the Wisconsin Diagnostic Breast Cancer (WDBC) dataset, the following models were used: Support Vector Machine, Logistic Regression, Decision Tree (C4.5), Forest Naïve Bayes (NB), Linear Regression (LR), k-nearest-neighbors (KNN), and Multilayer Perceptron (MLP). The data set is partitioned to use the machine learning algorithm: 20% is used during the test phase, while 80% is used throughout training. Hyper-parameters that are manually assigned are utilized to modify the classifier. When applied to a subset of data, it showed that combining SVM and model in machine learning reached the maximum accuracy of up to 90%, which was noticeably superior to the other ML model.

Qualitative and quantitative contrast-enhanced ultrasonography for the characterisation of non-palpable testicular tumours

BackgroundRecently, transvaginal color Doppler ultrasound (TVCDU) has been widely used in the diagnosis of pelvic tumors. This study aimed to assess the reliability of the application of TVCDU in the identification of pelvic tumors in women of childbearing age.MethodsThe clinical data of 209 patients with pelvic tumors at childbearing age admitted to our hospital from January 2016 to December 2019 were retrospectively analyzed. The ultrasound signs, color Doppler ultrasound score (CDUS), and resistance index (RI) of benign and malignant pelvic tumors diagnosed by TVCDU were analyzed. The value of transabdominal color Doppler ultrasound (TACDU) and TVCDU in the diagnosis of benign and malignant pelvic tumors was calculated.ResultsThere were 150 cases with benign pelvic tumors and 59 cases with malignant tumors. Most benign tumors had an intact capsule and regular shape, were mainly cystic, and were rarely accompanied by ascites; meanwhile, malignant tumors were mostly non-capsular and irregular in shape, mainly solid or cystic, and were often accompanied by ascites. There were significant differences in the above-mentioned ultrasound signs between benign and malignant tumors (P<0.05). The CDUS score of benign pelvic tumors was significantly lower than that of malignant tumors, and the RI value was significantly higher than that of the malignant tumors (P<0.05). The sensitivity, specificity, and accuracy of TACDU in the diagnosis of benign pelvic and malignant tumors were 83.33%, 84.75%, and 83.73%, respectively; and the sensitivity, specificity, and the accuracy of TVCDU in the diagnosis of benign pelvic and malignant tumors were 95.33%, 88.14%, and 93.30%, respectively; thus TVCDU had a superior performance compared to TACDU, especially in sensitivity and accuracy (P<0.05).ConclusionsThe sensitivity and accuracy of TVCDU in the differential diagnosis of benign pelvic and malignant tumors in women at childbearing age were significantly higher than those of TACDU. The combined application of CDUS and RI can further improve the accuracy in the diagnosis of pelvic tumors.

We retrospectively reviewed the clinical features and outcome of benign and malignant eyelid tumors from 1995 to 2015 in a tertiary medical center. Among 4,521 histologically confirmed eyelid tumors, 4,294 (95.0%) were benign tumors and 227 (5.0%) were malignant tumors. The mean age at diagnosis was significantly higher in patients with malignant lid tumors than those with benign lid tumors (72.5 and 55.4 years, resp., p < 0.001). The most common benign eyelid tumors were intradermal nevus (21.1%), followed by seborrheic keratosis (12.6%) and xanthelasma (11.2%). The most common malignant eyelid tumors were basal cell carcinomas (57.8%), followed by sebaceous gland carcinomas (21.1%) and squamous cell carcinomas (10.1%). There was a relative male predominance (63.4% and 49.2%, resp., p < 0.001) and higher recurrence rate (11.9% and 4.4%, resp., p < 0.001) in malignant lid tumors as compared with those of benign lid tumors. Twenty-two patients (9.7%) received orbital exenteration/enucleation. Eight patients (3.5%) with malignant lid tumors died of disease. Patients with eyelid melanoma were associated with a high mortality rate (25.0%). It is important to differentiate between benign and malignant eyelid tumors, because they may cause cosmetic disfigurement and severe morbidity, especially in those with malignant eyelid tumors.

Objective To investigate the differential diagnosis of benign and malignant parotid tumors by 3.0T MR dynamic enhancement (DCE-MRI) and diffusion-weighted MR imaging (DWI). Methods A retrospective analysis of 56 cases of parotid tumors in our hospital in May-2017 May 2016, and the integration of all patients with MRI and DWI data. The ADC value, the time signal intensity curve (TIC) and the quantitative parameters of the lesions were statistically analyzed, and the parameters of the diagnosis of malignant tumors were calculated by the receiver operating characteristic (ROC) curve. Results 86.67% (13/15) of pleomorphic adenoma in benign tumors showed A curve (Persistent), 13.23% (2/15) showed C curve (Washout), 73.07% (19/26) in adenolymphoma was B curve (Plateau), and the other 26.93% (9/26) was C type. In malignant tumor, 90% (9/10) presented a C curve and 10% (1/10) presented a B curve. The TIC types of benign and malignant parotid tumors were significantly different (P<0.05). There was a significant difference in ADC values between malignant tumors and pleomorphic adenoma, other benign tumors. Conclusion DCE-MRI and diffusion-weighted imaging have high clinical value in the differential diagnosis of benign and malignant tumors in the parotid gland, which may be useful for the diagnosis and treatment plan of parotid gland tumors. Key words: Parotid gland tumor; Dynamic enhancement; Quantitative analysis; Magnetic resonance imaging

Objective To study the value of time intensity curve of contrast enhanced ultrasound in the discrimination of benign and malignant soft tissue tumors. Methods Thirty-seven cases of soft tissue tumors were carried out contrast-enhanced ultrasound examination. The region of interest in the dynamic images were chosen to carry out Gamma-Variate curve fitting through Q-LAB software and made a quantitative analysis. Then they were compared with pathological diagnosis to analyze the ultrasound imaging performance of benign and malignant tumors and their characteristics of micro-irrigation. Results Thirty-seven cases of soft tissue tumors were pathologically diagnosed 21 cases of benign and malignant in 16 cases. The perfusion of benign tumors were mainly uniformity (19/21) or potentialization (2/21). The perfusion of malignant tumors were mainly uniformity (10/16) or nonuniformity (6/16). There were no significant differences in time to peak(TTP) and arrival time of contrast agent(AT) between benign and malignant tumors( P ＞0.05). The upslope rate(A) and the derived peak intensity(DPI) in benign group were significantly less than malignant group ( P ＜0.05). Conclusions Time-intensity curve analysis of contrast-enhanced ultrasound showed that benign and malignant soft tissue tumors were different in perfusion,and it will be of a certain value for the clinical differential diagnosis. Key words: Ultrasonography; Microbubbles; Soft tissue neoplasms

Breast cancer is a disease that causes excessive fear in women around the world. The number of high death rates by breast cancer can be reduced by early detection. This can make breast cancer a disease that is easy to cure. A collection of datasets about breast cancer is used in the process of early detection. Early detection is carried out to analyze the state of the early stages of breast cancer patients. This research paper proposes machine learning methods, namely Generalized Linear Model, Logistic Regression, and Gradient Boosted Decision Tree to enhance the classification performance of Wisconsin Diagnostic Breast Cancer Data. The diagnosis results in two classes of cancer decisions which are malignant and benign by looking at evaluating the accuracy of the data classification test. The result shows that the Generalized Linear Model achieves the accuracy of 99.4%, which is higher than the accuracies of the previous studies for classifying the Wisconsin Diagnostic Breast Cancer dataset.