Digital CMOS ISFET architectures and algorithmic methods for point-of-care diagnostics

Abstract

Over the past decade, the surge of infectious diseases outbreaks across the globe is redefining how healthcare is provided and delivered to patients, with a clear trend towards distributed diagnosis at the Point-of-Care (PoC). In this context, Ion-Sensitive Field Effect Transistors (ISFETs) fabricated on standard CMOS technology have emerged as a promising solution to achieve a precise, deliverable and inexpensive platform that could be deployed worldwide to provide a rapid diagnosis of infectious diseases. This thesis presents advancements for the future of ISFET-based PoC diagnostic platforms, proposing and implementing a set of hardware and software methodologies to overcome its main challenges and enhance its sensing capabilities.

The first part of this thesis focuses on novel hardware architectures that enable direct integration with computational capabilities while providing pixel programmability and adaptability required to overcome pressing challenges on ISFET-based PoC platforms. This section explores oscillator-based ISFET architectures, a set of sensing front-ends that encodes the chemical information on the duty cycle of a PWM signal. Two initial architectures are proposed and fabricated in AMS 0.35um, confirming multiple degrees of programmability and potential for multi-sensing. One of these architectures is optimised to create a dual-sensing pixel capable of sensing both temperature and chemical information on the same spatial point while modulating this information simultaneously on a single waveform. This dual-sensing capability, verified in silico using TSMC 0.18um process, is vital for DNA-based diagnosis where protocols such as LAMP or PCR require precise thermal control.

The COVID-19 pandemic highlighted the need for a deliverable diagnosis that perform nucleic acid amplification tests at the PoC, requiring minimal footprint by integrating sensing and computational capabilities. In response to this challenge, a paradigm shift is proposed, advocating for integrating all elements of the portable diagnostic platform under a single piece of silicon, realising a ``Diagnosis-on-a-Chip". This approach is enabled by a novel Digital ISFET Pixel that integrates both ADC and memory with sensing elements on each pixel, enhancing its parallelism. Furthermore, this architecture removes the need for external instrumentation or memories and facilitates its integration with computational capabilities on-chip, such as the proposed ARM Cortex M3 system.

These computational capabilities need to be complemented with software methods that enable sensing enhancement and new applications using ISFET arrays. The second part of this thesis is devoted to these methods. Leveraging the programmability capabilities available on oscillator-based architectures, various digital signal processing algorithms are implemented to overcome the most urgent ISFET non-idealities, such as trapped charge, drift and chemical noise. These methods enable fast trapped charge cancellation and enhanced dynamic range through real-time drift compensation, achieving over 36 hours of continuous monitoring without pixel saturation.

Furthermore, the recent development of data-driven models and software methods open a wide range of opportunities for ISFET sensing and beyond. In the last section of this thesis, two examples of these opportunities are explored: the optimisation of image compression algorithms on chemical images generated by an ultra-high frame-rate ISFET array; and a proposed paradigm shift on surface Electromyography (sEMG) signals, moving from data-harvesting to information-focused sensing. These examples represent an initial step forward on a journey towards a new generation of miniaturised, precise and efficient sensors for PoC diagnostics.