However, the objective should be to identify substandard performance EARLY in the roll, to avoid high speed rejects.
That's going to create a ton of low speed rejects. Let's say you're going at 40 kts and a 10 kt headwind comes along. Now, according to the model, airspeed is 25% off. The plane shows 55% more drag than I expect. Is it because of a dragging brake? Wing problem? Wind? Not to mention the dynamic change in engine performance. The NASA system commanded a reject at 15% discrepancy, so you'd be stopped by now.
The best method early in the roll is engine parameters, which is what you do today. And then couple the engine parameters to a model in order to detect problems.