Table 4 Prediction accuracy performance of using different data modalities for predicting EDSS>4. In each evaluation metric, the top-3 highest scores are highlighted

MRI T1-pre

0.6462 \(\pm\) 0.0352

0.2074 ± 0.0145

0.5089 ± 0.0397

0.7679 ± 0.0209

0.6567 ± 0.0300

MRI T1-post

0.6437 \(\pm\) 0.0389

0.2027 ± 0.0180

0.5501 ± 0.0390

0.6536 ± 0.0252

0.6697 ± 0.0199

MRI T2

0.7736 ± 0.0268

0.2245 ± 0.0198

0.6834 ± 0.0223

0.7409 ± 0.0398

0.7467 ± 0.0390

MRI FLAIR

0.7945 ± 0.2798

0.3306 ± 0.0309

0.7689 ± 0.0261

0.7423 ± 0.0265

0.7423 ± 0.0399

MRI PD

0.5430 ± 0.0401

0.0998 ± 0.0321

0.7536 ± 0.0218

0.4862 ± 0.0300

0.5046 ± 0.0399

Clinical Notes

0.7048 ± 0.0365

0.5201 ± 0.0293

0.4632 ± 0.0320

0.8956 ± 0.0235

0.4958 ± 0.0301

Structured EHR

0.6589 ± 0.0193

0.3651 ± 0.0265

0.7015 ± 0.0263

0.6587 ± 0.0366

0.6984 ± 0.0265

MRIs & Notes

0.7988 ± 0.0465

0.6321 ± 0.0299

0.7024 ± 0.0536

0.7792 ± 0.0563

0.7963 ± 0.0422

MRIs & EHR

0.7836 ± 0.0531

0.4265 ± 0.0323

0.6789 ± 0.0411

0.6875 ± 0.0333

0.6841 ± 0.0523

EHR & Notes

0.8078 ± 0.0232

0.7978 ± 0.0453

0.7268 ± 0.0435

0.7643 ± 0.0255

0.8125 ± 0.0353

MS-BERT( [11])

0.6010 ± 0.0222

0.2064 ± 0.0356

0.3090   ± 0.0265

0.7936 ± 0.0512

0.7788 ± 0.0398

MRI & Notes & EHR

0.8380 ± 0.0438

0.7963 ± 0.0520

0.7489   ± 0.0502

0.7936 ± 0.0488

0.7960 ± 0.0312