Multi-auxiliary domain transfer learning for diagnosis of MCI conversion

Bo Cheng,Bingli Zhu,Shuchang Pu
DOI: https://doi.org/10.1007/s10072-021-05568-6
2021-09-12
Neurological Sciences
Abstract:<p class="a-plus-plus">In the early stage of Alzheimer's disease (AD), mild cognitive impairment (MCI) has a higher risk of progression to AD, so the prediction of whether an MCI subject will progress to AD (known as progressive MCI, PMCI) or not (known as stable MCI, SMCI) within a certain period is particularly important in practice. It is known that such a task could benefit from jointly learning-related auxiliary tasks such as differentiating AD from PMCI or PMCI from normal control (NC) in order to take full advantage of their shared commonality. However, few existing methods along this line fully consider the correlations between the target and auxiliary tasks according to the clinical practice of AD pathology for diagnosis. To deal with this problem, in this paper, treating each task domain as a different one, we borrow the idea from transfer learning and propose a novel multi-auxiliary domain transfer learning (MaDTL) method, which explicitly utilizes the correlations between the target domain (task) and multi-auxiliary domains (tasks) according to the clinical practice. Specifically, the proposed MaDTL method incorporates two key modules. The first one is a multi-auxiliary domain transfer-based feature selection (MaDTFS) model, which can select a discriminative feature subset shared by the target domain and the multi-auxiliary domains. In the MaDTFS model, to combine more training data from multi-auxiliary domains and simultaneously suppress the negative effects resulting from the irrelevant parts of multi-auxiliary domains, we proposed a sparse group correlation Lasso that includes a proposed group correlation Lasso penalty (i.e., <span class="a-plus-plus inline-equation id-i-eq1"><span class="a-plus-plus equation-source format-t-e-x"><span class="mjpage"><svg xmlns:xlink="http://www.w3.org/1999/xlink" width="9.513ex" height="3.176ex" style="vertical-align: -1.171ex;" viewBox="0 -863.1 4095.7 1367.4" role="img" focusable="false" xmlns="http://www.w3.org/2000/svg"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"> <use xlink:href="#MJMAIN-2225" x="0" y="0"></use> <use xlink:href="#MJMAINB-57" x="500" y="0"></use> <use xlink:href="#MJMAINB-48" x="1690" y="0"></use> <use xlink:href="#MJMAIN-2225" x="2590" y="0"></use><g transform="translate(3091,-287)"> <use transform="scale(0.707)" xlink:href="#MJMAIN-32" x="0" y="0"></use> <use transform="scale(0.707)" xlink:href="#MJMAIN-2C" x="500" y="0"></use> <use transform="scale(0.707)" xlink:href="#MJMAIN-31" x="779" y="0"></use></g></g></svg></span></span></span>) and a proposed correlation Lasso penalty (i.e., <span class="a-plus-plus inline-equation id-i-eq2"><span class="a-plus-plus equation-source format-t-e-x"><span class="mjpage"><svg xmlns:xlink="http://www.w3.org/1999/xlink" width="9.513ex" height="3.176ex" style="vertical-align: -1.171ex;" viewBox="0 -863.1 4095.7 1367.4" role="img" focusable="false" xmlns="http://www.w3.org/2000/svg"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"> <use xlink:href="#MJMAIN-2225" x="0" y="0"></use> <use xlink:href="#MJMAINB-57" x="500" y="0"></use> <use xlink:href="#MJMAINB-48" x="1690" y="0"></use> <use xlink:href="#MJMAIN-2225" x="2590" y="0"></use><g transform="translate(3091,-287)"> <use transform="scale(0.707)" xlink:href="#MJMAIN-31" x="0" y="0"></use> <use transform="scale(0.707)" xlink:href="#MJMAIN-2C" x="500" y="0"></use> <use transform="scale(0.707)" xlink:href="#MJMAIN-31" x="779" y="0"></use></g></g></svg></span></span></span>). The second module in MaDTL is a multi-auxiliary domain transfer-based classification (MaDTC) model that improves the voting with linear weighting-based ensemble learning. This model extends the constraints of the linear weighting method so that it can simultaneously combine training data from multi-auxiliary domains and achieve a robust classifier by minimizing negative effects from the irrelevant part of multi-auxiliary domains. Experimental results on 409 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database with the baseline magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) data validate the effectiveness of the proposed method by significantly improving the classification accuracy to 80.37% for the identification of MCI-to-AD conversion, outperforming the state-of-the-art methods.</p><svg xmlns="http://www.w3.org/2000/svg" style="display: none;"><defs id="MathJax_SVG_glyphs"><path stroke-width="1" id="MJMAIN-2225" d="M133 736Q138 750 153 750Q164 750 170 739Q172 735 172 250T170 -239Q164 -250 152 -250Q144 -250 138 -244L137 -243Q133 -241 133 -179T132 250Q132 731 133 736ZM329 739Q334 750 346 750Q353 750 361 744L362 743Q366 741 366 679T367 250T367 -178T362 -243L361 -244Q355 -250 347 -250Q335 -250 329 -239Q327 -235 327 250T329 739Z"></path><path stroke-width="1" id="MJMAINB-57" d="M915 686L1052 683Q1142 683 1157 686H1164V624H1073L957 320Q930 249 900 170T855 52T839 10Q834 0 826 -5Q821 -7 799 -7H792Q777 -7 772 -5T759 10Q759 11 748 39T716 122T676 228L594 442L512 228Q486 159 455 78Q433 19 428 9T416 -5Q411 -7 389 -7H379Q356 -7 349 10Q349 12 334 51T288 170T231 320L116 624H24V686H35Q44 683 183 683Q331 683 355 686H368V624H323Q278 624 278 623L437 207L499 369L561 531L526 624H434V686H445Q454 683 593 683Q741 683 765 686H778V624H733Q688 624 688 623L847 207Q848 207 927 415T1006 624H905V686H915Z"></path><path stroke-width="1" id="MJMAINB-48" d="M400 0Q376 3 226 3Q75 3 51 0H39V62H147V624H39V686H51Q75 683 226 683Q376 683 400 686H412V624H304V388H595V624H487V686H499Q523 683 673 683Q824 683 848 686H860V624H752V62H860V0H848Q824 3 674 3Q523 3 499 0H487V62H595V326H304V62H412V0H400Z"></path><path stroke-width="1" id="MJMAIN-32" d="M109 429Q82 429 66 447T50 491Q50 562 103 614T235 666Q326 666 387 610T449 465Q449 422 429 383T381 315T301 241Q265 210 201 149L142 93L218 92Q375 92 385 97Q392 99 409 186V189H449V186Q448 183 436 95T421 3V0H50V19V31Q50 38 56 46T86 81Q115 113 136 137Q145 147 170 174T204 211T233 244T261 278T284 308T305 340T320 369T333 401T340 431T343 464Q343 527 309 573T212 619Q179 619 154 602T119 569T109 550Q109 549 114 549Q132 549 151 535T170 489Q170 464 154 447T109 429Z"></path><path stroke-width="1" id="MJMAIN-2C" d="M78 35T78 60T94 103T137 121Q165 121 187 96T210 8Q210 -27 201 -60T180 -117T154 -158T130 -185T117 -194Q113 -194 104 -185T95 -172Q95 -168 106 -156T131 -126T157 -76T173 -3V9L172 8Q170 7 167 6T161 3T152 1T140 0Q113 0 96 17Z"></path><path stroke-width="1" id="MJMAIN-31" d="M213 578L200 573Q186 568 160 563T102 556H83V602H102Q149 604 189 617T245 641T273 663Q275 666 285 666Q294 666 302 660V361L303 61Q310 54 315 52T339 48T401 46H427V0H416Q395 3 257 3Q121 3 100 0H88V46H114Q136 46 152 46T177 47T193 50T201 52T207 57T213 61V578Z"></path></defs></svg>
neurosciences,clinical neurology
What problem does this paper attempt to address?