CLASSIFICATION MODEL FOR BREAST CANCER MAMMOGRAMS

Machine learning has been the topic of interest in research related to early detection of breast cancer based on mammogram images. In this study, we compare the performance results from three (3) types of machine learning techniques: 1) Naïve Bayes (NB), 2) Neural Network (NN) and 3) Support Vector Machine (SVM) with 2000 digital mammogram images to choose the best technique that could model the relationship between the features extracted and the state of the breast (‘Normal’ or ‘Cancer’). Grey Level Co-occurrence Matrix (GLCM) which represents the two dimensions of the level variation gray in the image is used in the feature extraction process. Six (6) attributes consist of contrast, variance, standard deviation, kurtosis, mean and smoothness were computed as feature extracted and used as the inputs for the classification process. The data has been randomized and the experiment has been repeated for ten (10) times to check for the consistencies of the performance of all techniques. 70% of the data were used as the training data and another 30% used as testing data. The result after ten (10) experiments show that, Support Vector Machine (SVM) gives the most consistent results in correctly classifying the state of the breast as ‘Normal’ or ‘Cancer’, with the accuracy of 99.4%, in training and 98.76% in testing. The SVM classification model has outperformed NN and NB model in the study, and it shows that SVM is a good choice for determining the state of the breast at the early stage. ABSTRAK: Pembelajaran mesin telah menjadi topik yang diminati dalam penyelidikan yang berkaitan dengan pengesanan awal kanser payudara berdasarkan imej mamogram. Dalam kajian ini, kami membandingkan hasil prestasi dari tiga (3) jenis teknik pembelajaran mesin: 1) Naïve Bayes (NB), 2) Neural Network (NN) dan 3) Support Vector Machine (SVM) dengan 2000 imej digital mammogram hingga teknik terbaik yang dapat memodelkan hubungan antara ciri yang diekstraksi dan keadaan payudara ('Normal' atau 'Cancer') dapat diperoleh. Grey Level Co-occurrence Matrix (GLCM) yang mewakili dua dimensi variasi tahap kelabu pada gambar digunakan dalam proses pengekstrakan ciri. Enam (6) atribut terdiri dari kontras, varians, sisihan piawai, kurtosis, min dan kehalusan dihitung sebagai fitur yang diekstrak dan digunakan sebagai input untuk proses klasifikasi. Eksperimen telah diulang selama sepuluh (10) kali untuk memeriksa kesesuaian prestasi semua teknik. 70% data digunakan sebagai data latihan dan 30% lagi digunakan sebagai data ujian. Hasil setelah sepuluh (10) eksperimen menunjukkan bahawa, Support Vector Machine (SVM) memberikan hasil yang paling konsisten dalam mengklasifikasikan keadaan payudara dengan betul sebagai 'Normal' atau 'Kanser', dengan akurasi 99.4%, dalam latihan dan 98.76% dalam ujian. Model klasifikasi SVM telah mengungguli model NN dan NB dalam kajian ini, dan ia menunjukkan bahawa SVM adalah pilihan yang baik untuk menentukan keadaan payudara pada peringkat awal.

To Design a Mammogram Edge Detection Algorithm Using an Artificial Neural Network (ANN)

An artificial neural network (ANN) is used to resolve problems related to complex scenarios and logical thinking. Nowadays, a cause for concern is the mortality rate among women due to cancer. Generally, women to around 45 years old are the most vulnerable to this disease. Early detection is the only hope for the patient to survive, otherwise it may reach an unrecoverable stage. Currently, there are numerous techniques available for the diagnosis of such diseases out of which mammography is the most trustworthy method for detecting early stage cancer. The analysis of these mammogram images is always difficult to analyze due to low contrast and non-uniform background. The mammogram images are scanned, digitized for processing, nut that further reduces the contrast between region of interest (ROI) and the background. Furthermore, presence of noise, glands, and muscles leads to background contrast variations. The boundaries of the suspected tumor area are always fuzzy and improper. The aim of this article is to develop a robust edge detection technique which works optimally on mammogram images to segment a tumor area.

Comparative Analysis of Proposed Artificial Neural Network (ANN) Algorithm With Other Techniques

The mortality rate among women is increasing progressively due to cancer. Generally, women around 45 years old are vulnerable from this disease. Early detection is hope for patients to survive otherwise it may reach to unrecoverable stage. Currently, there are numerous techniques available for diagnosis of such a disease out of which mammography is the most trustworthy method for detecting early cancer stage. The analysis of these mammogram images are difficult to analyze due to low contrast and nonuniform background. The mammogram images are scanned and digitized for processing that further reduces the contrast between Region of Interest and background. Presence of noise, glands and muscles leads to background contrast variations. Boundaries of suspected tumor area are fuzzy & improper. Aim of paper is to develop robust edge detection technique which works optimally on mammogram images to segment tumor area. Output results of proposed technique on different mammogram images of MIAS database are presented and compared with existing techniques in terms of both Qualitative & Quantitative parameters.

Breast Cancer Detection Using Mammogram Images with Improved Multi-Fractal Dimension Approach and Feature Fusion

Breast cancer detection using mammogram images at an early stage is an important step in disease diagnostics. We propose a new method for the classification of benign or malignant breast cancer from mammogram images. Hybrid thresholding and the machine learning method are used to derive the region of interest (ROI). The derived ROI is then separated into five different blocks. The wavelet transform is applied to suppress noise from each produced block based on BayesShrink soft thresholding by capturing high and low frequencies within different sub-bands. An improved fractal dimension (FD) approach, called multi-FD (M-FD), is proposed to extract multiple features from each denoised block. The number of features extracted is then reduced by a genetic algorithm. Five classifiers are trained and used with the artificial neural network (ANN) to classify the extracted features from each block. Lastly, the fusion process is performed on the results of five blocks to obtain the final decision. The proposed approach is tested and evaluated on four benchmark mammogram image datasets (MIAS, DDSM, INbreast, and BCDR). We present the results of single- and double-dataset evaluations. Only one dataset is used for training and testing in the single-dataset evaluation, whereas two datasets (one for training, and one for testing) are used in the double-dataset evaluation. The experiment results show that the proposed method yields better results on the INbreast dataset in the single-dataset evaluation, whilst better results are obtained on the remaining datasets in the double-dataset evaluation. The proposed approach outperforms other state-of-the-art models on the Mini-MIAS dataset.

A machine and human reader study on AI diagnosis model safety under attacks of adversarial images

While active efforts are advancing medical artificial intelligence (AI) model development and clinical translation, safety issues of the AI models emerge, but little research has been done. We perform a study to investigate the behaviors of an AI diagnosis model under adversarial images generated by Generative Adversarial Network (GAN) models and to evaluate the effects on human experts when visually identifying potential adversarial images. Our GAN model makes intentional modifications to the diagnosis-sensitive contents of mammogram images in deep learning-based computer-aided diagnosis (CAD) of breast cancer. In our experiments the adversarial samples fool the AI-CAD model to output a wrong diagnosis on 69.1% of the cases that are initially correctly classified by the AI-CAD model. Five breast imaging radiologists visually identify 29%-71% of the adversarial samples. Our study suggests an imperative need for continuing research on medical AI model’s safety issues and for developing potential defensive solutions against adversarial attacks.