Abstract:A new paradigm is beginning to emerge in Radiology with the advent of increased computational capabilities and algorithms. This has led to the ability of real time learning by computer systems of different lesion types to help the radiologist in defining disease. For example, using a deep learning network, we developed and tested a multiparametric deep learning (MPDL) network for segmentation and classification using multiparametric magnetic resonance imaging (mpMRI) radiological images. The MPDL network was constructed from stacked sparse autoencoders with inputs from mpMRI. Evaluation of MPDL consisted of cross-validation, sensitivity, and specificity. Dice similarity between MPDL and post-DCE lesions were evaluated. We demonstrate high sensitivity and specificity for differentiation of malignant from benign lesions of 90% and 85% respectively with an AUC of 0.93. The Integrated MPDL method accurately segmented and classified different breast tissue from multiparametric breast MRI using deep leaning tissue signatures.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to improve the accuracy of automatic detection and classification of breast lesions (benign or malignant) in the detection of radiobiomarkers for breast cancer by using Multiparametric Deep Learning (MPDL) technology. Specifically, the researchers developed an MPDL network based on Multiparametric MRI (mpMRI), aiming to improve the automated detection and classification of breast lesions by defining tissue signatures. This method can not only distinguish benign and malignant lesions more accurately, but also build a personalized tissue signature database for each patient, thus providing support for clinical decision - making. ### Main Objectives: 1. **Develop MPDL Method**: Use Multiparametric MRI data to construct an MPDL network through Stacked Sparse Autoencoders (SSAE) to achieve the segmentation and classification of breast tissue. 2. **Verify MPDL Performance**: Verify the performance of the MPDL method on an independent radiology data set to ensure its robustness at different magnetic field intensities and imaging directions. 3. **Improve Diagnostic Accuracy**: By combining pharmacokinetic modeling parameters and Diffusion - Weighted Imaging/Apparent Diffusion Coefficient (ADC) indicators, make the MPDL model achieve diagnostic performance comparable to that of radiologists in distinguishing between benign and malignant breast lesions. ### Key Technical Points: - **Multiparametric MRI Data**: Including T1 - weighted, T2 - weighted, Diffusion - Weighted Imaging (DWI), ADC maps and Dynamic Contrast - Enhanced (DCE) imaging. - **Stacked Sparse Autoencoders (SSAE)**: An unsupervised learning method for constructing low - dimensional representations, which can extract features of high - dimensional data sets from original images. - **Tissue Feature Vectors**: Define feature vectors for normal tissue, adipose tissue, lesion tissue and background noise. - **Dice Similarity Index (DS)**: Used to evaluate the degree of overlap between MPDL segmentation results and the true values. - **Classification Performance Evaluation**: Evaluate the classification performance of the model through sensitivity, specificity and the area under the Receiver Operating Characteristic curve (ROC) (AUC). ### Experimental Design: - **Data Set**: MRI data of 189 female patients, among which 96 cases are malignant lesions, 39 cases are benign lesions, and 4 cases are normal (without lesions). - **Independent Validation Set**: 50 cases from an independent test data set of the University of California, San Francisco (UCSF), and these data are from patients with locally advanced breast cancer who received neoadjuvant chemotherapy. - **Evaluation Indicators**: Dice similarity index, sensitivity, specificity, AUC, etc. ### Results: - **Segmentation Performance**: The MPDL model performs excellently in segmenting different tissue types. The Dice similarity index for malignant lesions is 0.87 ± 0.05, and that for benign lesions is 0.85 ± 0.07. - **Classification Performance**: When distinguishing between benign and malignant lesions, the MPDL model has a sensitivity of 90%, a specificity of 85%, and an AUC of 0.93. - **Independent Verification**: On the UCSF independent verification data set, the volume difference between the segmentation results of the MPDL model and the true values is only 4.4% ± 3.9%, showing good robustness and generalization ability. ### Significance: - **Clinical Application Prospects**: The MPDL method can be used as an auxiliary tool for radiologists to improve the accuracy of breast lesion detection and classification, which is helpful for achieving precision radiology medicine. - **Technological Progress**: By integrating advanced computing technologies and artificial intelligence methods, in the future, reading rooms will be able to better process complex medical image data, improving work efficiency and diagnostic accuracy.

Multiparametric Deep Learning Tissue Signatures for a Radiological Biomarker of Breast Cancer: Preliminary Results

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

Multiparametric Deep Learning and Radiomics for Tumor Grading and Treatment Response Assessment of Brain Cancer: Preliminary Results

Radiomics and Machine Learning with Multiparametric Breast MRI for Improved Diagnostic Accuracy in Breast Cancer Diagnosis

Multiparametric Deep Learning Tissue Signatures for Muscular Dystrophy: Preliminary Results

Deep Learning for identifying radiogenomic associations in breast cancer

Radiomic and deep learning characterization of breast parenchyma on full field digital mammograms and specimen radiographs: a pilot study of a potential cancer field effect

Determining breast cancer biomarker status and associated morphological features using deep learning

Fully Automatic Classification of Breast Lesions on Multi-Parameter MRI Using a Radiomics Model with Minimal Number of Stable, Interpretable Features

Deep Learning-based Prediction of Breast Cancer Tumor and Immune Phenotypes from Histopathology

Deep learning radiomics based on multimodal imaging for distinguishing benign and malignant breast tumours

Artificial Intelligence-Based Classification of Breast Lesions Imaged With a Multiparametric Breast MRI Protocol With Ultrafast DCE-MRI, T2, and DWI

Prior Clinico‐Radiological Features Informed Multi‐Modal MR Images Convolution Neural Network: A novel deep learning framework for prediction of lymphovascular invasion in breast cancer

Optimizing and Visualizing Deep Learning for Benign/Malignant Classification in Breast Tumors

Detecting Cancerous Tissue in Mammograms Using Deep Neural Networks

Breast Cancer Classification on Multiparametric MRI – Increased Performance of Boosting Ensemble Methods

Deep Learning Radiomics of Preoperative Breast MRI for Prediction of Axillary Lymph Node Metastasis in Breast Cancer

Machine Learning-Based Analysis of MR Multiparametric Radiomics for the Subtype Classification of Breast Cancer.

Study on the classification of benign and malignant breast lesions using a multi-sequence breast MRI fusion radiomics and deep learning model

Preoperative Prediction of Ki-67 Status in Breast Cancer with Multiparametric MRI Using Transfer Learning

Deep BI-RADS Network for Improved Cancer Detection from Mammograms