A comprehensive hierarchical classification based on multi-features of breast DCE-MRI for cancer diagnosis

Deep transfer learning based hierarchical CAD system designs for SFM images

Present work involves rigorous experimentation for classification of mammographic masses by employing four deep transfer learning models using hierarchical framework. Experimental work is carried on 518 SFM images of DDSM dataset with 208, 150 and 160 images of probably benign, suspicious- malignant and highly malignant classes, respectively. ResNet50 model is used for generating segmented mass images. For hierarchical classification framework, at node 1, the segmented mass image is classified as belonging to probably benign (BIRAD-3) class or suspicious abnormality (BIRAD-4 and BIRAD-5) class. At node 2, the segmented mass image belonging to suspicious abnormality class is further classified as suspicious malignant (BIRAD-4) class or highly malignant (BIRAD-5) class. Deep transfer learning based hierarchical CAD systems experimented in the present work include VGG16/VGG19/ GoogleNet/ResNet50 models. It was noted that deep transfer learning model VGG19 at node 1 and VGG16 at node 2, yielded highest classification accuracy of 93 % and 90 %, respectively, therefore, a deep transfer learning based hybrid hierarchical CAD system was developed by employing VGG19 at node 1 and VGG16 at node 2. This model yields overall classification accuracy of 88 %. Further, hybrid hierarchical CAD system was designed using VGG19/ANFC-LH classifier at node 1, and VGG16/ANFC-LH classifier at node 2 yielding the highest classification accuracy of 92%. The promising result yielded by hybrid hierarchical CAD system design indicates its usefulness for step-wise classification of mammographic masses.

Improving signal-to-noise ratio by maximal convolution of longitudinal and transverse magnetization components in MRI: application to the breast cancer detection

PURPOSE

The extraction of information from images provided by medical imaging systems may be employed to obtain the specific objectives in the various fields. The quantity of signal to noise ratio (SNR) plays a crucial role in displaying the image details. The higher the SNR value, the more the information is available.

METHODS

In this study, a new function has been formulated using the appropriate suggestions on convolutional combination of the longitudinal and transverse magnetization components related to the relaxation times of T1 and T2 in MRI, where by introducing the distinct index on the maximum value of this function, the new maps are constructed toward the best SNR. Proposed functions were analytically simulated using Matlab software and evaluated with respect to various relaxation times. This proposed method can be applied to any medical images. For instance, the T1- and T2-weighted images of the breast indicated in the reference [35] were selected for modelling and construction of the full width at x maximum (FWxM) map at the different values of x-parameter from 0.01 to 0.955 at 0.035 and 0.015 intervals. The range of x-parameter is between zero and one. To determine the maximum value of the derived SNR, these intervals have been first chosen arbitrarily. However, the smaller this interval, the more precise the value of the x-parameter at which the signal to noise is maximum.

RESULTS

The results showed that at an index value of x = 0.325, the new map of FWxM (0.325) will be constructed with a maximum derived SNR of 22.7 compared to the SNR values of T1- and T2-maps by 14.53 and 17.47, respectively.

CONCLUSION

By convolving two orthogonal magnetization vectors, the qualified images with higher new SNR were created, which included the image with the best SNR. In other words, to optimize the adoption of MRI technique and enable the possibility of wider use, an optimal and cost-effective examination has been suggested. Our proposal aims to shorten the MRI examination to further reduce interpretation times while maintaining primary sensitivity.

SIGNIFICANCE

Our findings may help to quantitatively identify the primary sources of each type of solid and sequential cancer.

MRI-Based Breast Cancer Classification and Localization by Multiparametric Feature Extraction and Combination Using Deep Learning

BACKGROUND

Deep learning (DL) have been reported feasible in breast MRI. However, the effectiveness of DL method in mpMRI combinations for breast cancer detection has not been well investigated.

PURPOSE

To implement a DL method for breast cancer classification and detection using feature extraction and combination from multiple sequences.

STUDY TYPE

Retrospective.

POPULATION

A total of 569 local cases as internal cohort (50.2 ± 11.2 years; 100% female), divided among training (218), validation (73) and testing (278); 125 cases from a public dataset as the external cohort (53.6 ± 11.5 years; 100% female).

FIELD STRENGTH/SEQUENCE

T1-weighted imaging and dynamic contrast-enhanced MRI (DCE-MRI) with gradient echo sequences, T2-weighted imaging (T2WI) with spin-echo sequences, diffusion-weighted imaging with single-shot echo-planar sequence and at 1.5-T.

ASSESSMENT

A convolutional neural network and long short-term memory cascaded network was implemented for lesion classification with histopathology as the ground truth for malignant and benign categories and contralateral breasts as healthy category in internal/external cohorts. BI-RADS categories were assessed by three independent radiologists as comparison, and class activation map was employed for lesion localization in internal cohort. The classification and localization performances were assessed with DCE-MRI and non-DCE sequences, respectively.

STATISTICAL TESTS

Sensitivity, specificity, area under the curve (AUC), DeLong test, and Cohen's kappa for lesion classification. Sensitivity and mean squared error for localization. A P-value <0.05 was considered statistically significant.

RESULTS

With the optimized mpMRI combinations, the lesion classification achieved an AUC = 0.98/0.91, sensitivity = 0.96/0.83 in the internal/external cohorts, respectively. Without DCE-MRI, the DL-based method was superior to radiologists' readings (AUC 0.96 vs. 0.90). The lesion localization achieved sensitivities of 0.97/0.93 with DCE-MRI/T2WI alone, respectively.

DATA CONCLUSION

The DL method achieved high accuracy for lesion detection in the internal/external cohorts. The classification performance with a contrast agent-free combination is comparable to DCE-MRI alone and the radiologists' reading in AUC and sensitivity.

EVIDENCE LEVEL

3.

TECHNICAL EFFICACY

Stage 2.

Emerging methods for prostate cancer imaging: evaluating cancer structure and metabolic alterations more clearly

Imaging plays a fundamental role in all aspects of the cancer management pathway. However, conventional imaging techniques are largely reliant on morphological and size descriptors that have well-known limitations, particularly when considering targeted-therapy response monitoring. Thus, new imaging methods have been developed to characterise cancer and are now routinely implemented, such as diffusion-weighted imaging, dynamic contrast enhancement, positron emission technology (PET) and magnetic resonance spectroscopy. However, despite the improvement these techniques have enabled, limitations still remain. Novel imaging methods are now emerging, intent on further interrogating cancers. These techniques are at different stages of maturity along the biomarker pathway and aim to further evaluate the cancer microstructure (vascular, extracellular and restricted diffusion for cytometry in tumours) magnetic resonance imaging (MRI), luminal water fraction imaging] as well as the metabolic alterations associated with cancers (novel PET tracers, hyperpolarised MRI). Finally, the use of machine learning has shown powerful potential applications. By using prostate cancer as an exemplar, this Review aims to showcase these potentially potent imaging techniques and what stage we are at in their application to conventional clinical practice.