During the diagnostic process, clinicians leverage multimodal information, such as the chief complaint, medical images and laboratory test results. Deep-learning models for aiding diagnosis have yet to meet this requirement of leveraging multimodal information. Here we report a transformer-based representation-learning model as a clinical diagnostic aid that processes multimodal input in a unified manner. Rather than learning modality-specific features, the model leverages embedding layers to convert images and unstructured and structured text into visual tokens and text tokens, and uses bidirectional blocks with intramodal and intermodal attention to learn holistic representations of radiographs, the unstructured chief complaint and clinical history, and structured clinical information such as laboratory test results and patient demographic information. The unified model outperformed an image-only model and non-unified multimodal diagnosis models in the identification of pulmonary disease (by 12% and 9%, respectively) and in the prediction of adverse clinical outcomes in patients with COVID-19 (by 29% and 7%, respectively). Unified multimodal transformer-based models may help streamline the triaging of patients and facilitate the clinical decision-making process.

Recent studies have integrated convolutions into transformers to introduce inductive bias and improve generalization performance. However, the static nature of conventional convolution prevents it from dynamically adapting to input variations, resulting in a representation discrepancy between convolution and self-attention as self-attention calculates attention matrices dynamically. Furthermore, when stacking token mixers that consist of convolution and self-attention to form a deep network, the static nature of convolution hinders the fusion of features previously generated by self-attention into convolution kernels. These two limitations result in a suboptimal representation capacity of the constructed networks. To find a solution, we propose a lightweight dual dynamic token mixer (D-Mixer) to simultaneously learn global and local dynamics, that is, mechanisms that compute weights for aggregating global contexts and local details in an input-dependent manner. D-Mixer works by applying an efficient global attention module and an input-dependent depthwise convolution separately on evenly split feature segments, endowing the network with strong inductive bias and an enlarged effective receptive field. We use D-Mixer as the basic building block to design TransXNet, a novel hybrid CNN-transformer vision backbone network that delivers compelling performance. In the ImageNet-1K image classification task, TransXNet-T surpasses Swin-T by 0.3% in top-1 accuracy while requiring less than half of the computational cost. Furthermore, TransXNet-S and TransXNet-B exhibit excellent model scalability, achieving top-1 accuracy of 83.8% and 84.6%, respectively, with reasonable computational costs. In addition, our proposed network architecture demonstrates strong generalization capabilities in various dense prediction tasks, outperforming other state-of-the-art networks while having lower computational costs. Code is publicly available at https://github.com/LMMMEng/TransXNet.

Instance level detection and segmentation of thoracic diseases or abnormalities are crucial for automatic diagnosis in chest X-ray images. Leveraging on constant structure and disease relations extracted from domain knowledge, we propose a structure-aware relation network (SAR-Net) extending Mask R-CNN. The SAR-Net consists of three relation modules: 1. the anatomical structure relation module encoding spatial relations between diseases and anatomical parts. 2. the contextual relation module aggregating clues based on query-key pair of disease RoI and lung fields. 3. the disease relation module propagating co-occurrence and causal relations into disease proposals. Towards making a practical system, we also provide ChestX-Det, a chest X-Ray dataset with instance-level annotations (boxes and masks). ChestX-Det is a subset of the public dataset NIH ChestX-ray14. It contains ~3500 images of 13 common disease categories labeled by three board-certified radiologists. We evaluate our SAR-Net on it and another dataset DR-Private. Experimental results show that it can enhance the strong baseline of Mask R-CNN with significant improvements. The ChestX-Det is released at https://github.com/Deepwise-AILab/ChestX-Det-Dataset .

Purpose: Personalized treatments such as targeted therapy and immunotherapy have revolutionized the predominantly therapeutic paradigm for non-small cell lung cancer (NSCLC). However, these treatment decisions require the determination of targetable genomic and molecular alterations through invasive genetic or immunohistochemistry (IHC) tests. Numerous previous studies have demonstrated that artificial intelligence can accurately predict the single-gene status of tumors based on radiologic imaging, but few studies have achieved the simultaneous evaluation of multiple genes to reflect more realistic clinical scenarios. Methods: We proposed a multi-label multi-task deep learning (MMDL) system for non-invasively predicting actionable NSCLC mutations and PD-L1 expression utilizing routinely acquired computed tomography (CT) images. This radiogenomic system integrated transformer-based deep learning features and radiomic features of CT volumes from 1096 NSCLC patients based on next-generation sequencing (NGS) and IHC tests. Results: For each task cohort, we randomly split the corresponding dataset into training (80%), validation (10%), and testing (10%) subsets. The area under the receiver operating characteristic curves (AUCs) of the MMDL system achieved 0.862 (95% confidence interval (CI), 0.758–0.969) for discrimination of a panel of 8 mutated genes, including EGFR, ALK, ERBB2, BRAF, MET, ROS1, RET and KRAS, 0.856 (95% CI, 0.663–0.948) for identification of a 10-molecular status panel (previous 8 genes plus TP53 and PD-L1); and 0.868 (95% CI, 0.641–0.972) for classifying EGFR / PD-L1 subtype, respectively. Conclusions: To the best of our knowledge, this study is the first deep learning system to simultaneously analyze 10 molecular expressions, which might be utilized as an assistive tool in conjunction with or in lieu of ancillary testing to support precision treatment options.