Analog machine learning hardware platforms, such as those using wave physics, present potential for edge artificial intelligence (AI) applications due to in-sensor computing architecture, offering superior energy efficiency compared to digital circuits. While the diffractive neural network has been implemented in optical systems, its deployment on integrated acoustic systems has not been achieved due to the challenges associated with hardware optimization. In this paper, we propose the Diffractive Acoustic Neural Network (DANN), a novel approach that applies diffractive neural network algorithms to surface acoustic wave (SAW) systems for in-sensor multibiomarker diagnosis. To address optimization challenges, we introduce a novel training methodology that combines Finite Element Analysis (FEA) with gradient descent. We validate our method on Major Depressive Disorder (MDD) and prostate cancer, achieving accuracies of $74.07 \%$ and $86.0 \%$, respectively, nearly reaching the accuracy levels of clinical diagnoses. By comparing the co-training method with the traditional gradient descent training method and direct training on the FEA model, the co-training method demonstrates its advantages in balancing training efficiency and accuracy. Furthermore, a comparison of power consumption is conducted between the traditional method and the in-sensor computing system, indicating $66 \%$ energy savings attributed to its high level of integration.

Transcriptome sequencing has become essential in clinical tumor research, providing in-depth insights into the biology and functionality of tumor cells. However, the vast amount of data generated and the complex relationships between gene expressions make it challenging to effectively identify clinically relevant information. In this study, we developed a method called Gene Swin Transformer to address these challenges. This approach converts transcriptomic data into Synthetic Image Elements (SIEs). We utilized data from 12 datasets, including GSE17536-GSE103479 datasets (n = 1771) and The Cancer Genome Atlas (n = 459), to generate SIEs. These elements were then classified based on survival time using deep learning algorithms to predict colorectal cancer prognosis and build a reliable prognostic model. We trained and evaluated four deep learning models-BeiT, ResNet, Swin Transformer, and ViT Transformer-and compared their performance. The enhanced Swin-T model outperformed the other models, achieving weighted precision, recall, and F1 scores of 0.708, 0.692, and 0.705, respectively, along with area under the curve values of 80.2%, 72.7%, and 76.9% across three datasets. This model demonstrated the strongest prognostic prediction capabilities among those evaluated. Additionally, the PEX10 gene was identified as a key prognostic marker through both visual attention matrix analysis and bioinformatics methods. Our study demonstrates that the Gene Swin model effectively transforms Ribonucleic Acid (RNA) sequencing data into SIEs, enabling prognosis prediction through attention-based algorithms. This approach supports the development of a data-driven, unified, and automated model, offering a robust tool for classification and prediction tasks using RNA sequencing data. This advancement presents a novel clinical strategy for cancer treatment and prognosis forecasting.

Bio-inspired event sensors are gaining popularity at the edge, such as in robots and wearable electronics. This trend necessitates learning vast amounts of sensory data on the edge, often in few-shot or even zero-shot scenarios, posing challenges in both software and hardware. This paper presents a novel software-hardware co-design to address these issues. Software-wise, we develop an SNN-ANN model, where the SNN encoder is a liquid state machine (LSM) that naturally processes events and significantly reduces learning complexity at the edge due to fixed random weights. The lightweight trainable ANN projection heads are optimized through contrastive learning, enabling zero-shot learning of multimodal events. Hardware-wise, we propose a hybrid analog (RRAM)-digital (CMOS) accelerator - LSMR. The analog in-memory computing core physically implements the LSM by leveraging RRAM stochasticity to generate fixed random weights. The digital core utilizes innovative reconfigurable systolic arrays to accelerate the contrastive learning of ANN projection heads. Extensive experimental outcomes from six neuromorphic datasets, encompassing visual, tactile, and auditory modalities, demonstrate that LSMR considerably improves energy efficiency by a range of 1.65× to 23.70×, in comparison to state-of-the-art edge devices. Simultaneously, it reduces training complexity by a range of 152.83× to 20,587.77× across various edge learning tasks.