Convolutional neural network for breast cancer diagnosis using diffuse optical tomography

Abstract

We have developed a computer-aided diagnosis system based on a convolutional neural network that aims to classify breast mass lesions in optical tomographic images obtained using a diffuse optical tomography system, which is suitable for repeated measurements in mass screening. Sixty-three optical tomographic images were collected from women with dense breasts, and a dataset of 1260 2D gray scale images sliced from these 3D images was built. After image preprocessing and normalization, we tested the network on this dataset and obtained 0.80 specificity, 0.95 sensitivity, 90.2% accuracy, and 0.94 area under the receiver operating characteristic curve (AUC). Furthermore, a data augmentation method was implemented to alleviate the imbalance between benign and malignant samples in the dataset. The sensitivity, specificity, accuracy, and AUC of the classification on the augmented dataset were 0.88, 0.96, 93.3%, and 0.95, respectively.

Making Radiomics More Reproducible across Scanner and Imaging Protocol Variations: A Review of Harmonization Methods

“…Most machine learning (ML) and DL algorithms have been used on reconstructed images in the existing medical imaging workflow. Instead, the abilities of ML and DL could be leveraged to process the underlying raw sensor-level data to access its hidden nuances [ 53 , 54 , 55 ]. A study conducted by Lee et al [ 56 ] investigated the performance of a CNN for classifying raw CT data in the sinogram-space to identify the body region and detect intracranial hemorrhage.…”

“…Many studies [ 1 , 28 , 29 , 30 , 31 ] have also explored the discriminative power of radiomic features. However, the reproducibility of a radiomic feature does not guarantee its discriminative power [ 32 , 33 ], and thus the two aspects of reproducibility and discriminative power cannot be treated in isolation. For instance, a feature may have excellent reproducibility across scanner and protocol variations but have no discriminative power for the problem of interest.…”

Two-stage classification strategy for breast cancer diagnosis using ultrasound-guided diffuse optical tomography and deep learning

“…29 Recently, machine learning (ML) has been applied to DOT in bulk tissue optical property estimation, 30,31 image reconstruction, [32][33][34][35][36][37][38][39] and breast cancer diagnosis. [40][41][42][43] Using simulated breast lesions, Di Sciacca et al 42 applied logistic regression, support vector machines (SVMs), and a fully connected network to reconstruct optical properties and achieved a best accuracy of 78%. Xu et al 43 applied a convolutional neural network (CNN) to 1260 2D optical tomographic images sliced from 3D images collected from 63 women with dense breasts and achieved a 0.95 area under the receiver operating characteristic (ROC) curve (AUC).…”

Clinical photoacoustic/ultrasound dual-modal imaging: Current status and future trends

“…In these studies, the system designs have in common that an optimized imaging system achieved a higher spatial and temporal resolution, better penetration in tissue with reduced artifacts. Consequently, the advancements in PAT have enabled a wide applications ranging from small animal studies to clinical imaging, including imaging of breast ( Becker et al, 2018 ; Xu et al, 2019 ; Yang et al, 2020 ), thyroid ( Dima and Ntziachristos, 2016 ; Sinha et al, 2017 ; Roll et al, 2019 ), skin ( Petri et al, 2016 ; Dahlstrand et al, 2020 ), tumors ( Li et al, 2015 ; Yamada et al, 2020 ; Karmacharya et al, 2021 ; Knorring and Mogensen, 2021 ; Wang C et al, 2021 ), cardiovascular ( Taruttis et al, 2013 ; Karlas et al, 2021a ), functional neuroimaging ( Wang et al, 2003 ; Wu et al, 2019a ), eyes ( Liu and Zhang, 2016 ) and others ( Nagae et al, 2018 ; Yang et al, 2018 ; Liang et al, 2021 ; Yan et al, 2021 ). Therefore, PAT imaging has broader clinical translational potential than other forms of pure optical imaging, indicating its ability to provide potent structural, functional, and molecular information in vivo ( Yao and Wang, 2018 ; Wu M et al, 2021 ).…”