Breast cancer screening, primarily conducted through mammography, is often supplemented with ultrasound for women with dense breast tissue. However, existing deep learning models analyze each modality independently, missing opportunities to integrate information across imaging modalities and time. In this study, we present Multi-modal Transformer (MMT), a neural network that utilizes mammography and ultrasound synergistically, to identify patients who currently have cancer and estimate the risk of future cancer for patients who are currently cancer-free. MMT aggregates multi-modal data through self-attention and tracks temporal tissue changes by comparing current exams to prior imaging. Trained on 1.3 million exams, MMT achieves an AUROC of 0.943 in detecting existing cancers, surpassing strong uni-modal baselines. For 5-year risk prediction, MMT attains an AUROC of 0.826, outperforming prior mammography-based risk models. Our research highlights the value of multi-modal and longitudinal imaging in cancer diagnosis and risk stratification.
Medical data poses a daunting challenge for AI algorithms: it exists in many different modalities, experiences frequent distribution shifts, and suffers from a scarcity of examples and labels. Recent advances, including transformers and self-supervised learning, promise a more universal approach that can be applied flexibly across these diverse conditions. To measure and drive progress in this direction, we present BenchMD: a benchmark that tests how modality-agnostic methods, including architectures and training techniques (e.g. self-supervised learning, ImageNet pretraining), perform on a diverse array of clinically-relevant medical tasks. BenchMD combines 19 publicly available datasets for 7 medical modalities, including 1D sensor data, 2D images, and 3D volumetric scans. Our benchmark reflects real-world data constraints by evaluating methods across a range of dataset sizes, including challenging few-shot settings that incentivize the use of pretraining. Finally, we evaluate performance on out-of-distribution data collected at different hospitals than the training data, representing naturally-occurring distribution shifts that frequently degrade the performance of medical AI models. Our baseline results demonstrate that no modality-agnostic technique achieves strong performance across all modalities, leaving ample room for improvement on the benchmark. Code is released at https://github.com/rajpurkarlab/BenchMD .
3D imaging enables a more accurate diagnosis by providing spatial information about organ anatomy. However, using 3D images to train AI models is computationally challenging because they consist of tens or hundreds of times more pixels than their 2D counterparts. To train with high-resolution 3D images, convolutional neural networks typically resort to downsampling them or projecting them to two dimensions. In this work, we propose an effective alternative, a novel neural network architecture that enables computationally efficient classification of 3D medical images in their full resolution. Compared to off-the-shelf convolutional neural networks, 3D-GMIC uses 77.98%-90.05% less GPU memory and 91.23%-96.02% less computation. While our network is trained only with image-level labels, without segmentation labels, it explains its classification predictions by providing pixel-level saliency maps. On a dataset collected at NYU Langone Health, including 85,526 patients with full-field 2D mammography (FFDM), synthetic 2D mammography, and 3D mammography (DBT), our model, the 3D Globally-Aware Multiple Instance Classifier (3D-GMIC), achieves a breast-wise AUC of 0.831 (95% CI: 0.769-0.887) in classifying breasts with malignant findings using DBT images. As DBT and 2D mammography capture different information, averaging predictions on 2D and 3D mammography together leads to a diverse ensemble with an improved breast-wise AUC of 0.841 (95% CI: 0.768-0.895). Our model generalizes well to an external dataset from Duke University Hospital, achieving an image-wise AUC of 0.848 (95% CI: 0.798-0.896) in classifying DBT images with malignant findings.
Deep neural networks (DNNs) show promise in image-based medical diagnosis, but cannot be fully trusted since their performance can be severely degraded by dataset shifts to which human perception remains invariant. If we can better understand the differences between human and machine perception, we can potentially characterize and mitigate this effect. We therefore propose a framework for comparing human and machine perception in medical diagnosis. The two are compared with respect to their sensitivity to the removal of clinically meaningful information, and to the regions of an image deemed most suspicious. Drawing inspiration from the natural image domain, we frame both comparisons in terms of perturbation robustness. The novelty of our framework is that separate analyses are performed for subgroups with clinically meaningful differences. We argue that this is necessary in order to avert Simpson's paradox and draw correct conclusions. We demonstrate our framework with a case study in breast cancer screening, and reveal significant differences between radiologists and DNNs. We compare the two with respect to their robustness to Gaussian low-pass filtering, performing a subgroup analysis on microcalcifications and soft tissue lesions. For microcalcifications, DNNs use a separate set of high frequency components than radiologists, some of which lie outside the image regions considered most suspicious by radiologists. These features run the risk of being spurious, but if not, could represent potential new biomarkers. For soft tissue lesions, the divergence between radiologists and DNNs is even starker, with DNNs relying heavily on spurious high frequency components ignored by radiologists. Importantly, this deviation in soft tissue lesions was only observable through subgroup analysis, which highlights the importance of incorporating medical domain knowledge into our comparison framework.
Breast cancer is the most common cancer in women, and hundreds of thousands of unnecessary biopsies are done around the world at a tremendous cost. It is crucial to reduce the rate of biopsies that turn out to be benign tissue. In this study, we build deep neural networks (DNNs) to classify biopsied lesions as being either malignant or benign, with the goal of using these networks as second readers serving radiologists to further reduce the number of false positive findings. We enhance the performance of DNNs that are trained to learn from small image patches by integrating global context provided in the form of saliency maps learned from the entire image into their reasoning, similar to how radiologists consider global context when evaluating areas of interest. Our experiments are conducted on a dataset of 229,426 screening mammography exams from 141,473 patients. We achieve an AUC of 0.8 on a test set consisting of 464 benign and 136 malignant lesions.
Deep neural networks (DNNs) show promise in breast cancer screening, but their robustness to input perturbations must be better understood before they can be clinically implemented. There exists extensive literature on this subject in the context of natural images that can potentially be built upon. However, it cannot be assumed that conclusions about robustness will transfer from natural images to mammogram images, due to significant differences between the two image modalities. In order to determine whether conclusions will transfer, we measure the sensitivity of a radiologist-level screening mammogram image classifier to four commonly studied input perturbations that natural image classifiers are sensitive to. We find that mammogram image classifiers are also sensitive to these perturbations, which suggests that we can build on the existing literature. We also perform a detailed analysis on the effects of low-pass filtering, and find that it degrades the visibility of clinically meaningful features called microcalcifications. Since low-pass filtering removes semantically meaningful information that is predictive of breast cancer, we argue that it is undesirable for mammogram image classifiers to be invariant to it. This is in contrast to natural images, where we do not want DNNs to be sensitive to low-pass filtering due to its tendency to remove information that is human-incomprehensible.
Medical images differ from natural images in significantly higher resolutions and smaller regions of interest. Because of these differences, neural network architectures that work well for natural images might not be applicable to medical image analysis. In this work, we extend the globally-aware multiple instance classifier, a framework we proposed to address these unique properties of medical images. This model first uses a low-capacity, yet memory-efficient, network on the whole image to identify the most informative regions. It then applies another higher-capacity network to collect details from chosen regions. Finally, it employs a fusion module that aggregates global and local information to make a final prediction. While existing methods often require lesion segmentation during training, our model is trained with only image-level labels and can generate pixel-level saliency maps indicating possible malignant findings. We apply the model to screening mammography interpretation: predicting the presence or absence of benign and malignant lesions. On the NYU Breast Cancer Screening Dataset, consisting of more than one million images, our model achieves an AUC of 0.93 in classifying breasts with malignant findings, outperforming ResNet-34 and Faster R-CNN. Compared to ResNet-34, our model is 4.1x faster for inference while using 78.4% less GPU memory. Furthermore, we demonstrate, in a reader study, that our model surpasses radiologist-level AUC by a margin of 0.11. The proposed model is available online: https://github.com/nyukat/GMIC.
We present a deep convolutional neural network for breast cancer screening exam classification, trained and evaluated on over 200,000 exams (over 1,000,000 images). Our network achieves an AUC of 0.895 in predicting whether there is a cancer in the breast, when tested on the screening population. We attribute the high accuracy of our model to a two-stage training procedure, which allows us to use a very high-capacity patch-level network to learn from pixel-level labels alongside a network learning from macroscopic breast-level labels. To validate our model, we conducted a reader study with 14 readers, each reading 720 screening mammogram exams, and find our model to be as accurate as experienced radiologists when presented with the same data. Finally, we show that a hybrid model, averaging probability of malignancy predicted by a radiologist with a prediction of our neural network, is more accurate than either of the two separately. To better understand our results, we conduct a thorough analysis of our network's performance on different subpopulations of the screening population, model design, training procedure, errors, and properties of its internal representations.
Advances in deep learning for natural images have prompted a surge of interest in applying similar techniques to medical images. The majority of the initial attempts focused on replacing the input of a deep convolutional neural network with a medical image, which does not take into consideration the fundamental differences between these two types of images. Specifically, fine details are necessary for detection in medical images, unlike in natural images where coarse structures matter most. This difference makes it inadequate to use the existing network architectures developed for natural images, because they work on heavily downscaled images to reduce the memory requirements. This hides details necessary to make accurate predictions. Additionally, a single exam in medical imaging often comes with a set of views which must be fused in order to reach a correct conclusion. In our work, we propose to use a multi-view deep convolutional neural network that handles a set of high-resolution medical images. We evaluate it on large-scale mammography-based breast cancer screening (BI-RADS prediction) using 886,000 images. We focus on investigating the impact of the training set size and image size on the prediction accuracy. Our results highlight that performance increases with the size of training set, and that the best performance can only be achieved using the original resolution. In the reader study, performed on a random subset of the test set, we confirmed the efficacy of our model, which achieved performance comparable to a committee of radiologists when presented with the same data.