HDformer: A Higher Dimensional Transformer for Diabetes Detection Utilizing Long Range Vascular Signals

Diabetes mellitus is a worldwide concern, and early detection can help to prevent serious complications. Low-cost, non-invasive detection methods, which take cardiovascular signals into deep learning models, have emerged. However, limited accuracy constrains their clinical usage. In this paper, we present a new Transformer-based architecture, Higher Dimensional Transformer (HDformer), which takes long-range photoplethysmography (PPG) signals to detect diabetes. The long-range PPG contains broader and deeper signal contextual information compared to the less-than-one-minute PPG signals commonly utilized in existing research. To increase the capability and efficiency of processing the long range data, we propose a new attention module Time Square Attention (TSA), reducing the volume of the tokens by more than 10x, while retaining the local/global dependencies. It converts the 1-dimensional inputs into 2-dimensional representations and groups adjacent points into a single 2D token, using the 2D Transformer models as the backbone of the encoder. It generates the dynamic patch sizes into a gated mixture-of-experts (MoE) network as decoder, which optimizes the learning on different attention areas. Extensive experimentations show that HDformer results in the state-of-the-art performance (sensitivity 98.4, accuracy 97.3, specificity 92.8, and AUC 0.929) on the standard MIMIC-III dataset, surpassing existing studies. This work is the first time to take long-range, non-invasive PPG signals via Transformer for diabetes detection, achieving a more scalable and convenient solution compared to traditional invasive approaches. The proposed HDformer can also be scaled to analyze general long-range biomedical waveforms. A wearable prototype finger-ring is designed as a proof of concept.

Performer: A Novel PPG to ECG Reconstruction Transformer For a Digital Biomarker of Cardiovascular Disease Detection

Cardiovascular diseases (CVDs) have become the top one cause of death; three-quarters of these deaths occur in lower-income communities. Electrocardiography (ECG), an electrical measurement capturing the cardiac activities, is a gold-standard to diagnose CVDs. However, ECG is infeasible for continuous cardiac monitoring due to its requirement for user participation. Meanwhile, photoplethysmography (PPG) is easy to collect, but the limited accuracy constrains its clinical usage. In this research, a novel Transformer-based architecture, Performer, is invented to reconstruct ECG from PPG and to create a novel digital biomarker, PPG along with its reconstructed ECG, as multiple modalities for CVD detection. This architecture, for the first time, performs Transformer sequence to sequence translation on biomedical waveforms, while also utilizing the advantages of the easily accessible PPG and the well-studied base of ECG. Shifted Patch-based Attention (SPA) is created to maximize the signal features by fetching the various sequence lengths as hierarchical stages into the training while also capturing cross-patch connections through the shifted patch mechanism. This architecture generates a state-of-the-art performance of 0.29 RMSE for reconstructing ECG from PPG, achieving an average of 95.9% diagnosis for CVDs on the MIMIC III dataset and 75.9% for diabetes on the PPG-BP dataset. Performer, along with its novel digital biomarker, offers a low-cost and non-invasive solution for continuous cardiac monitoring, only requiring the easily extractable PPG data to reconstruct the not-as-accessible ECG data. As a prove of concept, an earring wearable, named PEARL (prototype), is designed to scale up the point-of-care (POC) healthcare system.