INGENIEUR
INGENIEUR
recommendations to doctors to support their clinical decisions .
Alzheimer ’ s disease ( AD ) is a degenerative neurological condition that results in a decline in cognitive abilities and has no validated cure . Early detection of AD is crucial to provide timely treatment . Mild cognitive impairment ( MCI ) is a stage between normal cognitive ageing and AD . To predict the conversion from MCI to probable AD , a study used a deep learning approach called multimodal recurrent neural network ( Lee et al ., 2019 ). The study established an integrative framework that combined cross-sectional neuroimaging biomarkers at baseline and longitudinal cerebrospinal fluid ( CSF ) and cognitive performance biomarkers from the Alzheimer ’ s Disease Neuroimaging Initiative cohort ( ADNI ). The framework integrated longitudinal multi-domain data . The results indicated that :
1 . The prediction model for MCI conversion to AD achieved up to 75 % accuracy ( area under the curve ( AUC ) = 0.83 ) when using single modalities of data ; and
2 . The best performance was achieved with 81 % accuracy ( AUC = 0.86 ) when incorporating longitudinal multi-domain data . The use of a multi-modal deep learning approach has the potential to identify individuals at risk of developing AD who could benefit from clinical trials or as a stratification approach in these trials .
In the field of geriatric falls , UTAR researchers are collaborating with UM researchers on developing effective fall clustering . Falls are a common issue among older individuals that can lead to serious physical and psychological consequences . The causes of falls are often attributed to multiple risk factors , making clinical evaluations time-consuming and limiting their availability . To address this challenge , Goh et al . ( 2022 ) conducted a study to develop a clusteringbased algorithm for fall risk assessment . The study used data from the Malaysian Elders Longitudinal Research ( MELoR ), which included 1,411 subjects aged 55 years and older . The algorithm was developed through several stages : data pre-processing , feature identification and extraction using t-Distributed Stochastic Neighbour Embedding ( t-SNE ) or Principal Component Analysis ( PCA ), clustering using
K-means , Hierarchical , and Fuzzy C-means clustering , and characterisation interpretation with statistical analysis . After the pre-processing stage , 1,279 subjects and nine variables were selected for clustering . Using t-SNE and K-means clustering , the subjects were grouped into low , intermediate A , intermediate B , and high fall risk groups , corresponding to fall occurrence rates of 13 %, 19 %, 21 %, and 31 % respectively . The key variables identified were slower gait , poorer balance , weaker muscle strength , presence of cardiovascular disorder , poorer cognitive performance , and advancing age . The fall risk clustering algorithm was able to group subjects based on features , making it a useful tool for clinical decision-making , and enhancing access to falls prevention efforts .
Clinical Trials and Clinical Research
Artificial intelligence is being used in clinical trials to improve the efficiency , accuracy , and costeffectiveness of drug development . AI algorithms can be used to identify patients who are eligible for clinical trials , optimise clinical trial designs , and analyse data from clinical trials . A study published in the Journal of Clinical Oncology ( Wen et al ., 2019 ) showed that AI algorithms were able to identify patients who were eligible for clinical trials for cancer drugs with high accuracy , reducing the time and cost of clinical trial recruitment .
Bringing a new drug to market takes 10 to 15 years and costs between USD1.5 and 2 billion , with clinical trials taking up the latter half of this period . If a clinical trial fails , not only is the investment in the trial lost , but also the costs incurred in preclinical development , leading to a loss of between USD800 million and 1.4 billion per failed trial . The high failure rates are due to the suboptimal selection of patient cohorts and poor recruitment techniques , as well as the inability to effectively monitor patients during trials ( Harrer et al ., 2019 ). Only one in ten compounds that enter clinical trials reach the market . With the recent advances in AI , it can be utilised in real-life situations to support human decision-making . By transforming the key components of clinical trial design and implementation , AI has the potential to enhance
54 VOL 95 JULY-SEPTEMBER 2023