Singe sensor on foot sole predicts gait freezing in Parkinson’s disease

A single plantar pressure sensor device, when placed on the least affected side (LAS), helps predict freezing of gait (FOG) in patients with Parkinson’s disease (PD), according to a recent study.

“The LAS model would have similar FOG prediction performance to the bilateral model at the cost of slightly more false positives,” the researchers said. “Given the advantages of single sensor systems, the increased false positive rate may be acceptable to people with PD.”

The present analysis used data plantar pressure data from 11 men with PD who reported experiencing freezing at least once per week. Patients were asked to wear pressure-sensing insoles and to walk through a freeze-inducing path for up to 30 times.

Data were then retrieved and processed to arrive at FOG prediction models, comparing among algorithms that used data from the LAS, most affected side (MAS), or both limbs. Models were trained using many different combinations of time- and frequency-domain features, such as the number of weight shifts, freeze index, and deviations from the centre of pressure.

Among all models, the highest sensitivity was at 79.5 percent, achieved by a LAS model that incorporated five features. Of note, LAS models consistently had the highest sensitivity outcomes for five, 10, 15, and 25 features. Bilateral models with 20 and 30 features yielded the good sensitivity values at 74.6 percent and 66.7 percent, respectively. [Front Neurol 2022;13:831063]

Meanwhile, the most specific model used bilateral sensors with 30 features (88.0 percent). A LAS and MAS model, each using 30 features, also yielded high specificity, with corresponding values of 87.5 percent and 83.9 percent.

Overall, the high-performing models using five, 10, or 15 features identified 90.2 percent to 94.9 percent of all FOG episodes. Specifically, a LAS model with five features emerged as the most accurate model, identifying 94.9 percent of freezing. Of note, increasing the number of features incorporated led to a marked drop in the percentage of FOG predicted.

In terms of prediction timing, bilateral and LAS models using five features performed best, both identifying a freeze 0.8 seconds before the episodes. In addition, models that used five, 10, or 15 features could predict an FOG from 0.4 to 0.8 seconds before the fact. In comparison, MAS models, while generally slower to catch FOG episodes, had the lowest average false positives.

“In a system that uses a single plantar pressure sensor, the decision to instrument the LAS or MAS may be person specific,” the researchers said. “For someone who tends to recover independently from FOG, instrumenting the MAS may be preferable, since there would be fewer false positives.”

On the other hand, patients for whom freezes typically lead to loss of balance and falling, equipping the LAS with the sensing device might be better in order to prevent FOGs in the first place.

“In practice, using a single-limb plantar-pressure based FOG prediction system could enhance wearability and compliance, since fewer sensors would need to be worn,” the researchers said.