Research Article
Optimizing Neuropsychological Assessments for Cognitive, Behavioral, and Functional Impairment Classification: A Machine Learning Study
Table 9
Top 10 features most frequently found as best predictors across all 10 rounds and all 100 iterations using the FDR feature reduction.
| | Level of impairment | Features | |
| | CDR = 1 versus CDR = 0 | (1) LDELTOTAL (LM) | 71% | | (2) TOTALMOD (ADAS) | 10% | | (3) LIMMTOTAL (LM) | 4% | | (4) FAQTOTAL (FAQ) | 4% | | (5) Q4 (ADAS) | 4% | | (6) AVTOT5 (AVLT) | 3% | | (7) AVTOT4 (AVLT) | 1% | | (8) Q1 (ADAS) | 0.8% | | (9) AVDEL30MIN (AVLT) | 0.6% | | (10) TOTAL11 (ADAS) | 0.5% |
| | CDR = 0.5 versus CDR = 0 | (1) LDELTOTAL (LM) | 91% | | (2) Q4 (ADAS-Cog) | 22% | | (3) LIMMTOTAL (LM) | 15% | | (4) TOTALMOD (ADAS-Cog) | 12% | | (5) GDHOPE (GDS) | 6% | | (6) MMD (MMSE) | 2% | | (7) MMSCORE (MMSE) | 0.3% | | (8) AVTOT4 (AVLT) | 0.1% | | (9) CATVEGESC (Semantic Fluency Test) | 0.1% | | (10) TOTAL11 (ADAS) | 0.1% |
| | CDR = 1 versus CDR = 0.5 | (1) FAQTOTAL (FAQ) | 31% | | (2) TOTALMOD (ADAS-Cog) | 22% | | (3) AVTOT5 (AVLT) | 10% | (4) FAQFORM (FAQ)
(5) Q1 (ADAS-Cog) | 6%
6% | | (6) FAQREM (FAQ) | 6% | | (7) TOTAL11 (ADAS) | 5% | | (8) CLOCKSCOR (CLOCK Test) | 4% | | (9) CATVEGESC (Semantic Fluency Test) | 4% | | (10) Q8 (ADAS) | 4% |
|
|
10 rounds of the nested CV and across 100 iterations.
|
