18 1 Introduction
chip itself i s inexpensive, an incorrect experimental outcome caused by a defective
biochip could lead to an erroneous clinical diagnosis, which is unacceptable and
must be prevented. In addition, owing to the inherent randomness of component
reactions in biochemistry, predictive modeling and accurate control are difficult [3,
74]. All these factors make biochips especially vulnerable to defects and erroneous
microfluidic operations, which are unacceptable for applications such as real-time
DNA sequencing [4] and antigen detection [75] that require high precision. If an
unexpected error occurs during a biochemistry experiment and the requirements of
the bioassay protocol are violated, the outcome of the experiment will be incorrect,
leading to the likelihood of misdiagnosis and incorrect disease prognosis. Even if
errors are detected while tests are being conducted on patient samples, the repetition
of experiments on newer chips leads to an increase in time-to-response for diagnosis
and higher cost, since samples are hard to obtain and reagents are expensive [76].
To overcome these barriers to practical adoption and deployment, a quality-
checking method is needed to target key fluidic operations before chip use [77].
Defective chips need to be identified and subsequently discarded. Furthermore, for
the adoption of PDMS microfluidic technology, quality-control measurements are
required. The service providers must have robust post-fabrication testing methods to
ensure the quality of their product. Only then can researchers trust data generated
using these devices. Recent advances in testing of digital microfluidic biochips [78–
80] cannot be used here because of the many differences in the underlying tech-
nologies. Flow-based biochips manipulate continuous flow in microchannels using
pumps and valves, while digital microfluidic biochips control discrete droplets on
an electrode array using the principle of electrowetting-on-dielectric [1]. The test
stimuli, test control, and observation methods are inherently different for these types
of biochips. The fluid-control mechanism and complex microchannel network struc-
tures in flow-based devices make testing difficult.
Visual inspection under microscopes remains the most common test method [81,
82]. However, it suffers from low throughput, it is labor-intensive, and it requires
expensive scopes, cameras, scanning stages, and complex vision-analysis software.
Even skilled observers cannot systematically scan the entire chip at high resolution.
More importantly, the fault coverage (percentage of detectable faults) obtained using
visual inspection is inadequate: defects can easily escape detection and some defects
are invisible under the microscope even at high magnification. For example:
• Valves cannot close completely. Two potential causes may lead to this defect:
(i) the membrane layer is too thick; (ii) flow channels are too tall.
• Poor bonding between control and flow layers: It could result in a short under pres-
sure between channels and these defects are undetectable through visual inspection
Moreover, virtual inspection would over-test, and therefore lead to unnecessary
yield loss. For example:
• Slight misalignment: When valves are partially misaligned with underlying chan-
nels, the chip could still be functional and must be assessed with functional tests.