Tunable phosphorene modulators – accelerating medical diagnosis with ultra-efficient photonic platforms

The rapid integration of deep learning in medical image diagnosis is challenged by the energy inefficiency and data bottlenecks of conventional electronic computing. Photonic Neural Networks (PNNs) offer a promising, ultra-parallel computing paradigm to overcome these limitations. In this work, we present a high-performance electro-optic modulator realized by synergizing the exceptional properties of a phosphorene-based van der Waals heterostructure with the enhanced light-matter interaction of a microring resonator. This design achieves efficient, low-power modulation with a high modulation efficiency of 0.25 V·cm. We utilized this modulator to construct a prototype all-fiber PNN based on a time-division multiplexing architecture. To improve computational precision, we employed a Ring-Assisted Mach–Zehnder Interferometer (RAMZI) structure, effectively extending the linear operating region for precise weight matrix representation. Experimental validation confirms the PNN’s robust and accurate performance in complex tasks, including handwritten digit classification, retinal B-scan analysis, and multiphase liver CT image diagnosis. This study introduces a high-efficiency electro-optic modulation scheme and lays the foundation for highly integrated, low-power, fiber-based PNNs for accelerated and efficient medical diagnostics.