Diagnosing Parkinson’s Disease

First symptoms of Parkinson's
KeySense - Diagnosis of Parkinson's Disease
® KeySense is a registered trademark of Parkinson’s Foundation (Australia) © Copyright 2023 Parkinson’s Foundation (Australia)

Parkinson’s FOUNDATION


How KeySense Works

There has been a need for more accurate and objective diagnoses of Parkinson’s Disease (PD), particularly in its early stages where the observable symptoms may be subtle and imprecise. In order to be effec- tive, such a new assessment tech- nique needs to - Be significantly more accurate than current non- specialist clinician diagnoses. Provide an objective, quantifiable and repeatable diagnostic test for PD. Be able to diagnose PD where there are just mild symptoms present. Be able to detect it significantly earlier in the disease progression. Not require specialised equipment, training or referral. Take place in the patient’s home or office environment as they type normally on a computer. Also be capable of providing ongoing monitoring of medication, treatments & disease progression.

KeySense Methodology

This involves the use of typing – by analyzing the rhythm and cadence of keystrokes as someone types on a keyboard. There is a wealth of information embedded in those keystroke. Way back in the mid-19th century, when the tele- graph was in wide use, telegraph opera- tors could identify other operators based on their rhythm as they tapped on the telegraph key. Much later, during World War II, a methodology known as the ‘Fist of the Sender’ was used to identify the sender of the telegraph by using the rhythm, pace and cadence of their Morse code. Even without decoding the messages, this allowed German troop movements to be monitored by the Allied Powers, as particular telegraph operators
were attached to a specific army unit. These days, as someone types on a computer keyboard, the characteristics of their finger movement can be measured down to millisecond accuracy.

KeySense Development

Based on 4 years of peer-reviewed scientific research, with nearly 500 worldwide participants and 9,000 typing samples. Capture & recording of keystroke char- acteristics as someone types normally on a computer keyboard. Detection of 4 separate types of PD feature, especially two of the cardinal features for clinical diagnosis - bradyki- nesia & tremor. Use of sophisticated artificial intelli- gence (AI) machine learning models. As implemented, the technique provides objective, quantifiable, repeatable measurements and results.

The Outcome

Provides a tool for both detection & ongoing monitoring of disease status and progression. The patient can take the assessment reports to their medical practi- tioner (GP) or specialist. KeySense has a high and repeatable accu- racy, with an 86% detection rate for even early, mild symptoms of Parkinson’s, and just 5% false positives. This compares with GP accuracies of less than 75% across all disease severities, and much lower than that for early, mild cases.

The User Assessment Process

The user types 350 words of text (two thirds of a page) on their own computer and KeySense records a myriad of timings about their finger movement. This data is then fed through a series of machine learning models which iden- tify the probabilities of specific move- ment features being within – or outside – normal expected ranges. Specifically, KeySense measures inconsistency, drop- off and sidedness of movement and tremor, as well as combining these into an overall score. The results are accompanied by descrip- tive details which are also emailed to the user.

Privacy and Anonymity

The KeySense assessment does not need any personal details to be provided and the participants remain anonymous at all times.
How KeySense Works


Adams, W. R. (2020). The detection of movement-related disease by changes in keystroke characteristics while typing. Charles Sturt University, Australia Adams, W. R. (2019). Keyboard typing for the detection of early Parkinson’s Disease. In V. R. Preedy & C. Martin (Eds.), The Neuroscience Of Parkinson’s Disease. Academic Press. Adams, W. R. (2018). The detection of hand tremor through the characteristics of finger movement while typing. BioRxiv. Min, O., Wei, Z., Nian, Z., & Su, X. (2020). An Application of LSTM Prediction Model Based on Keystroke Data. In 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence. New York, NY, USA: Association for Computing Machinery. Peachap, A. B., Tchiotsop, D., Louis-Dorr, V., & Wolf, D. (2020). Detection of early Parkinson’s disease with wavelet features using finger typing movements on a keyboard. SN Applied Sciences, 2(10), 1–8.