Detection And Diagnosis Of Breast Cancer With MRI And Breast Ultrasound: A Review

In this era, Breast cancer is a deadly disease among the women all over the world with high mortality rate. It is spreading rapidly because of the unconsciousness and the inactivity of the women. Most of the women don’t understand that they have breast cancer in the early stage. By the reported of World Health Organization (WHO), like in much of the world, breast cancer is the most common cancer amongst Bangladesh's women, with 32. 8 percent of female cancer patients suffering from this strain of the disease. The death rate can be decreased by the early detection and diagnosis of breast cancer. In this case, mammography is the best detection and diagnosis tools for screening mammograms. Although mammography is cost effectiveness and higher availability but it is not easy to interpret the mammograms because of the architectural distortion, false positive result and also the misinterpretation of the radiologists etc. To interpret mammograms and to improve the accuracy rate of the diagnosis, Computer Aided Diagnosis (CAD) is the best tool which can help the radiologists to interpret mammograms with a high accuracy rate. There are many techniques for the implementation of mammography such as Magnetic Resonance Imaging (MRI), Digital Breast Tomosynthesis, Full Field Digital Mammography (FFDM), breast Ultrasound etc. In this paper, we reviewed the MRI and breast Ultrasound technique in details.

In order to determining the probability of the malignity of lesion on contrast-enhanced MRI images Deurloo et al. prepared a computer a system. For the purpose of supporting the computer system they conducted 100 lesions (in 78 patients). An expert radiologist gave the narrative definition (benign, probably benign, indeterminate, suspicious or highly suggestive malignancy) of the lesions on MRI to them. By using the logistic regression exploration into an assembled model, the result from the computer system and the radiologist were assembled. On an individual collection of 72 clinically and mammographically occult lesions in 60 patients the model was examined. The computer system and the assembled model gave the Az values for the radiologist’s studies and the values were 0. 86, 0. 85 and 0. 91 gradually. To enhance the entire achievement for clinically and mammographically occult lesions the supplementary data from a radiologist and a computer has the possibility which was explained by the test. Chan et al. accomplished on malignant and benign alignment on T1-weighted three-dimensional destroyed slope-echo series with and without intravenously injected contradict. He applied a dataset of 121 cases (77 malignant and 44 benign). For the hand-segmented and computer separated outline, the alignment accuracy in terms of Az applying a graphical classifier was 0. 80 and 0. 86, gradually. Morphology, improvement dynamics and the field of improvement-changing over the lesion were removed from both the hand-outlined and instinctive computer separated lesion outline that was associated to characteristic. In order to separating between malignant and benign breast masses Nattkemper et al. performed a classifier by applying MRI data from 74 cases (49 malignant and 25 benign). From progressive contradict -improvement MRI time-field data, they removed outline type and wash-out-type characteristics which was determinate by radiologists. They gained an Az of 0. 88 by applying a support vector machine classifier.

Applying MRI before and after neoadjuvant chemotherapy DeMarini et al. analyzed 15 patients with 16 newly diagnosed locally ahead breast cancers. The appearance or default of important improvement, improvement profiles and maximum sizes were filed in CAD assessments. All tumors displayed CAD-assessed important improvement earlier to chemotherapy. With minimal remaining malignancy at pathology, the outcomes showed a decrement in wash-out improvement in patients following chemotherapy. Although all had remaining disease at pathology, seven of the 16 tumors showed no remaining important improvement after chemotherapy. The radiologist should not replace the cautious assessment of tumors though CAD for breast MRI may fulfill in this patient demography. In order to illustrate mammographic architectural characteristics, L. W. Numes and colleagues applied a decision tree approach. The lesion margin’s characterization was the initial definitive of malignancy, they observed. For the purpose of characterizing margin asperity, Alan Penn and his colleagues researched the mass boundary’s fractal dimension. A statistically important enhancement in characterization was invented assembling the fractal exploration with the earlier architectural characteristics. The first commercially obtainable software Cadstream®) was produced by Confirma® in 2004. In order to acquire greater ability this software is dedicated automated software schematic to automate the image processing and exploration performances generally accomplished by the radiologist. This software was performed by Lehman et al. in 33 ambiguous lesions that were detected on MRI and biopsied under MR leading. All malignant lesions were exactly identified by the software and displayed important improvement with a susceptibility of 100%. With the comparison of the radiologist’s exploration, the false positive rates for the CAD software were reduced by 25% at a 50% threshold, by 33% at an 80% threshold and 50% at a 100% threshold for improvement. A high accuracy with improved specificity without reducing susceptibility was discovered by CAD MR Software.

For distinguishing breast masses as malignant or benign on three dimensional ultrasound scans applying 102 biopsied masses Sahiner et al. performed a CAD system. In an observer testing the three-dimensional ultrasound volumes are performed by four radiologists. The CAD system gained a field under ROC curve, Az, of 0. 92 for testing system based on the three-dimensional segmentation process. The four radiologists organized the Az values from 0. 84 to 0. 92. For distinguishing malignant and benign breast masses on three-dimensional ultrasound volumes in this testing the accuracy of the classifier schematic was analogous to that of expert breast radiologists.

Drukker et al. improved a two-stage CAD system that first discovered the lesions on ultrasound images and then classified the lesion by differentiating the cancer from all other lesions. For training four hundreds cases were applied and for testing 485 cases were applied. For the purpose of characterizing the true lesions and false-positives the Az was 0. 94 and 0. 91 for the training and testing datasets, gradually. The az for the purpose of classifying cancer from all other detections was 0. 87 and 0. 81 for the training and testing data sets, gradually. An associated three-dimensional ultrasound and mammographic CAD system was performed by Sahiner et al. that applied 67 patients and accomplished an observer testing with five radiologists. The associated CAD system’s Az was 0. 91. By applying associated CAD the radiologists significantly enhance (P = 0. 03) their average performance to an Az of 0. 95. A CAD system assembling ultrasound and mammographic data was performed by Drukker et al. A dataset of 100 lesions from 97 patients with obtainable mammograms and ultrasound images was tested by this method. Comparing with the mammography system alone (Az = 0. 77) and the ultrasound system alone (Az = 0. 88), the Az for the attached system was 0. 92, which shown up a statistically important enhancement.

For characterizing breast masses as malignant or benign on ultrasound images Joo et al. , Chang et al. , Moon et al. , and Song et al. indicated additional CAD systems. . The CAD systems were evaluated with the datasets of various sizes (between 54 and 584 cases) and based on artificial intelligence techniques (support vector devices and artificial neural networks). These CAD systems gained the classification accuracy Az ranged from 0. 86 to 0. 95.

For distinguishing masses as benign, undefined or malignant several features were applied by A. T. Stavros and colleagues. Ultrasound is a digital modality in which at the time of experiment to enhance diagnostic accuracy we finally could apply CAD techniques. Given that sonographic explanation is a method which is subjective and dissociation of benign and malignant solid breast masses that definite criteria might approve, especially when assembled with similarly mammographic data CAD techniques used to sonographic images should also promote radiologist performance. In the use of computer-extracted features derived from co-occurrence matrices of breast-lesion images promising outcomes were displayed by Garra and colleagues.

Medipattern® (Toronto, Canada) prepared B-CAD which is the first commercial software. Therefore, performing teams have also prepared other home-made systems in imaging experiment. The principles from ordinary physics are generally the equal. The sonographic character must be quantified into computerized sonographic characters to recount a mass or a cystic lesion in the computer software. Characteristic classes could be determinate to analyze the mass, shape and margin. Various criteria which are morphological can be removed: shape, roundness, outline analysis, convexity, density, spiculation of the margins etc.