Kieran Baker

Mr. Kieran Baker

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 February 2024 in Nature Machine Intelligence
Verified
Journal Article
Open access
PDF file
HCHarry Coppock
Harry Coppock
GNGeorge Nicholson
George Nicholson
IKIvan Kiskin
Ivan Kiskin
VKVasiliki Koutra
Vasiliki Koutra
KBKieran Baker
Kieran Baker
CHChris Holmes
Chris Holmes
Recent work has reported that respiratory audio-trained AI classifiers can accurately predict SARS-CoV-2 infection status. However, it has not yet been determined whether such model performance is driven by latent audio biomarkers with true causal links to SARS-CoV-2 infection or by confounding effects, such as recruitment bias, present in observational studies. Here we undertake a large-scale study of audio-based AI classifiers as part of the UK government’s pandemic response. We collect a dataset of audio recordings from 67,842 individuals, with linked metadata, of whom 23,514 had positive polymerase chain reaction tests for SARS-CoV-2. In an unadjusted analysis, similar to that in previous works, AI classifiers predict SARS-CoV-2 infection status with high accuracy (ROC–AUC = 0.846 [0.838–0.854]). However, after matching on measured confounders, such as self-reported symptoms, performance is much weaker (ROC–AUC = 0.619 [0.594–0.644]). Upon quantifying the utility of audio-based classifiers in practical settings, we find them to be outperformed by predictions on the basis of user-reported symptoms. We make best-practice recommendations for handling recruitment bias, and for assessing audio-based classifiers by their utility in relevant practical settings. Our work provides insights into the value of AI audio analysis and the importance of study design and treatment of confounders in AI-enabled diagnostics.

11 June 2025 in Biochemical Journal
Verified
Journal Article
Open access
CUChianna Umamahesan
Chianna Umamahesan
APAleksandra Pilcicka
Aleksandra Pilcicka
JYJennifer Yick
Jennifer Yick
KBKieran Baker
Kieran Baker
MSMelvyn Smith
Melvyn Smith
ACAndré Charlett
André Charlett
Constipation is a forerunner to Parkinson’s disease (PD) diagnosis, worsening thereafter. We explore the relationship of intestinal barrier dysfunction to constipation and whether intestinal fungal load is an aggravating factor. Fungal load was quantified by real-time PCR, using ITS1F-ITS2 primer set, on microbial DNA extract from stool in 68 participants with PD, 102 without. Fungal load was 60% higher per decade after age 60 years, with no PD status interaction with age. After age adjustment, it was associated inversely with dietary renal acid load. It was unrelated to the presence of constipation or barrier dysfunction. Neither consumption of antimicrobials nor of other targeted exogenous substances was associated. Enzyme-linked immunosorbent assays measured barrier dysfunction markers, faecal alpha-1 antitrypsin (AAT), zonulin and serum intestinal fatty acid-binding protein (I-FABP). Barrier dysfunction was associated with constipation and slower radiographic colonic transit. Functional constipation was 28% more frequent with a doubling of AAT concentration. More colonic-transit test markers were retained in the transverse colon, the higher the AAT and zonulin concentrations, anatomically spotlighting abnormality for the entire colon. In contrast, the concentration of the small intestinal barrier marker I-FABP was associated with looser stool consistency, which is consistent with secondary microbial overgrowth. By showing a relationship of intestinal barrier dysfunction to constipation, this study supports the hypothesis that dysfunction may be consequential. Dysfunction may be a necessary, but not sufficient, precursor to PD, in allowing inflammaging. Since ageing is the clearest risk for PD, a gut pathogen escalating in abundance from the sixth decade, integral to fungal load, and whose reproduction and virulence is favoured by alkalinity, tallies.

12 October 2023 in Journal of Clinical Medicine
Verified
Journal Article
Open access
PDF file
WWWenjing Wang
Wenjing Wang
KBKieran Baker
Kieran Baker
CUChianna Umamahesan
Chianna Umamahesan
SGSteven Gilmour
Steven Gilmour
ACAndré Charlett
André Charlett
RDRJ Dobbs
RJ Dobbs
We question whether bradyphrenia, slowing of cognitive processing not explained by depression or a global cognitive assessment, is a nosological entity in idiopathic parkinsonism (IP). The time taken to break contact of an index finger with a touch-sensitive plate was measured, with and without a warning in the alerting signal as to which side the imperative would indicate, in 77 people diagnosed with IP and in 124 people without an IP diagnosis. The ability to utilise a warning, measured by the difference between loge-transformed reaction times (unwarned minus warned), was termed ‘cognitive efficiency’. It was approximately normally distributed. A questionnaire on self- and partner perception of proband’s bradyphrenia was applied. A multivariable model showed that those prescribed levodopa were less cognitively efficient (mean −5.2 (CI −9.5, −1.0)% per 300 mg/day, p = 0.02), but those prescribed the anti-muscarinic trihexyphenidyl were more efficient (14.7 (0.2, 31.3)% per 4 mg/day, p < 0.05) and those prescribed monoamine oxidase-B inhibitor (MAOBI) tended to be more efficient (8.3 (0.0, 17.4)%, p = 0.07). The variance in efficiency was greater within IP (F-test, p = 0.01 adjusted for any demographic covariates: coefficient of variation, with and without IP, 0.68 and 0.46, respectively), but not so after adjustment for anti-parkinsonian medication (p = 0.13: coefficient of variation 0.62). The within-participant follow-up time, a median of 4.8 (interquartile range 3.1, 5.5) years (101 participants), did not influence efficiency, irrespective of IP status. Perception of bradyphrenia did not usefully predict efficiency. We conclude that both bradyphrenia and ‘tachyphrenia’ in IP appear to have iatrogenic components, of clinically important size, related to the dose of antiparkinsonian medication. Levodopa is the most commonly prescribed first-line medication: co-prescribing a MAOBI may circumvent its associated bradyphrenia. The previously reported greater efficiency associated with (low-dose) anti-muscarinic was confirmed.