In principle, yes: using different effective sensitivities or exposures across the sensor can help capture a wider brightness range. But in practice it is not as simple as assigning ISO per pixel and getting a better image for free.

A real example is Fujifilm’s EXR “DR” mode, where groups of pixels are exposed differently and then combined. This can increase dynamic range, but it comes with trade-offs such as reduced resolution. Similar sensor designs can also use pixels of different sensitivity, but they require extra circuitry per pixel, which reduces light-gathering area and can hurt noise performance.

There are also workflow costs: the sensor data must be merged or tone-mapped during processing, which increases camera processing demands and complicates RAW support. Standard RAW converters are built around more conventional sensor data, so unusual per-pixel exposure schemes need special handling.

So the idea is useful and has been implemented in limited forms, but the downsides are complexity, possible noise penalties, reduced resolution, and software support challenges.