LU103178B1 - Method and system fpr predicting deep brain stimulation parameters - Google Patents

Abstract

The present disclosure relates to a method for suggesting personalized and symptom-specific DBS stimulation parameters, comprising the steps of: creating a three-dimensional symptom network library comprising a data-driven three-dimensional model of symptom-specific tracts in stereotactic space created on a detailed and inclusive pathway atlas; registering patient data with the three-dimensional symptom network library for analyzing how a single patient's electrode maps to it; deriving and suggesting stimulation parameters as a function of a baseline symptom severity profile in each patient. Another object of the invention relates to a system for predicting DBS parameters, the system comprising an interface for providing DBS parameters to an electrode and receiving signals from the patient; and a computer programmable unit comprising a data storage for a three-dimensional symptom network library comprising a stored data-driven three-dimensional model of symptom-specific tracts in stereotactic space created on a detailed and inclusive pathway for predicting DBS parameters.

Description

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN Fort mann Tagethoi : 12077 204B1LU Fee se EE .

LU103178 %

METHOD AND SYSTEM FPR PREDICTING DEEP BRAIN STIMULATION .

PARAMETERS .

DESCRIPTION 4

Field of the invention : :

[0001] The present disclosure relates to a method and a system for predicting Deep Brain .

Stimulation (DRS) parameters. ©

Brief description of the related art 8 10002] Deep brain stimulation { DBS) is an established therapy for the treatment of Parkinson's 7

Disease. In the past decade, research in DBE is expen encing a paradigm shift towards studying p sa, Ying she effects of DBS on distributed brain networks (Horn et al, 2017; Hom and Fox, 2020).

Identifying and characterizing such networks in symptomespesific manner could pave the way to personalize DBS.

[0003] DRS to the subthalamic nucleus (STN) is an established treatment For Parkinson’s . disease (PD) and has revolutionized treatment of the disease as the most critical historical = development following the introduction of Levodopa, However, while the efficacy of DBS on © symptoms such as tremor and bradykinesia has been established in randomized clinical trials! à its effects on wait and other axial symptoris have heen variable, even including detrimental : effects of electrical stimulation under certain circumstance (Schrader, ©, et al. Neurology 77, : 4R3-488, 2011, Yin, Z. et al, J. Neurol. Sei, 393, 116-127, 2018), This is further emphasized . by the notion that, while 200,000 patients world-wide have undergone DRS (Veda-Mai, V. el . al. Front, Hum. Neurosci, 15, 2021, extrapolating from larger clinical trials, around 71,000 of 3 those (45%) did not achieve significant {improvements in quality of fife, despite motor- ) symptoms relief (Denschl, G. etal, N. Engl, J. Mad. 355, 896-908, 2006). Hence, while many 7 patients benefit from DBS, not all do (Aviles-Olmos, Let al, JL Neurol, Neurosurg. Psychiatry : 1 eta SOS SSI AE

CHARTE — LNIVERSITATSMEDIZIN BERLIN rormann Tegethoi | ©

RS, 1419-1425, 2014). One reason could be that, currently, a focal brain region was surgically . targeted to treat all symptoms of the disease. For instance, in STN-DBS, à coordinate is targeted | . within the posterolateral part of the nucleus defined by direct imaging and surgical landmarks ; such as the Bejjani line (Bejjani, BP. et al, J. Neurosurg. 92. 615-625, 2000). While stimulation is adjusted postoperatively during DBS programming and heuristics in symptoms | : specific parameter optimization exist in clinical practice (such as switching to dorsal contacts i far treatment of tremor), a more thorough and data driven model for patient-specific symptom : optimizations is jacking, Furthermore, segmented electrodes with up to sixteen contacts per . lead are currently employed, making the programming process thereasingly complex,

[0004] Thus, precisely defining symptom specific network targets appears 10 be s way to . imporve DBS. Evidence for symptom-specific networks spans back as far as the 1960s, when : ) meticulous analyses of lesion studies revealed networks causally involved with specific cardinal 7 ) symptoms of PD (Hassler et al., Brain 83, 337-350, 1960; MeGregor, M. M. & Nelson, A, B. .

Meuron 101, 1042-1056, 2019), For instance, in seminal work by the Freiburg school of stéreutaxy based on 560 ablation cases between 1950 and 1958, Hassler et al. concluded that . optimal control oftremor involved lesioning a loop between cerebellum ( and Mollaret triangle), . which sent axon collaterals to the red nucleus on its way to the posterior nucleus ventrooralis (V.o.p.) and primary motor cortex (Hassler et al, Brain 83, 337-330, 1960), In contrast, connections from pallidum to the anterior nucleus ventrooralis (V.0.4.) and a subregion of the supplementary motor area (defined by the Vogt/Hassler/Brodmanu school as area Hae) were ; associated with improvements in hypokinetic symptoms such as bradykinesia and rigidity.

Much later, modern neuroimaging work could confirm both concepts, for instance, Helmich et . al. associated Parkinsonian rest and, likely, action tremor with the cerebellothalamocortical , circuit (Ni et al, Ann. Neurol. 68. 816-824, 2010; Helmich et al, Curr, Neurol. Neurosci, Rep. ; 13, 378, 2013; Sturman et al, Brain 127, 2131-2143, 2004), Using DBS network mapping, this could be confirmed by Akram et al, among others, who ir addition confirmed improvements 7 in bradykinesia and rigidity © be related te connections from premotor areas and prefrontal . cortex {Akram et al. Neurolmage, 158,352-344, 2017; Strotzer, CL D et al, Ann. Meurel. 85, 7 852.864, 2019). 2 [00051 In the Hight of these gtanularities of involved segregated but related overlapping circuits, optimal outcomes of DBS for PD could profit from conceptualizing each symptom or symptom. 2 )

CHARITE — DNIVERSITATSMEDIZIN BERLIN rormaon FT Tagethotl I 12077,20483LU Le . cluster (not the entire disease) as an entity that may be treated by modulating a specific cironit. A

This leads to two conclusions: first, the need for an accurate and three-dimensional symptom- 7 circuit model in stereotactic standard space. Once established, patient-specific electrode 7 placement could be related to such a model to determine optimal stimulation settings for each . patient: If the patient presented with tremor-dominant symptoms, settings could be adjusted to 7 maximize modulation of the tremor circuit. Conversely, if axial symptoms primarily affected 7 quality of life, they would be adjusted to maximize impact on the axial cirouit. Second, it may . become possible to deliver treatment to several segregated circuits with a single DBS electrode by switching on different contacts {or combinations of contacts) ~ and potentially by using applying stimulation pulses at different frequencies. This would render parameter choices complex and underlines the necessity of automated methods to suggest symptom-spoctfic 7 stimulation settings. . 006] Published US, patent LS 10,905,882 B2 discloses à system and method for optimizing . parameters of à DBS pulse signal for treatment of a patient. In predicting optimal DBS 7 parameters, functional brain data is input into à predictor system, the functional brain data acquired responsive to à sweeping across a multi-dimensional parameter space of one or more

DBS parameters, Statistical metrics of brain response are extracted from the functional brain data for one or more ROIs or voxels of the brain via the predictor system, and a DBS functional , atlas is accessed, via the predictor system, that comprises disease-specific brain response maps â derived from DBS treatment at optimal DBS parameter settings for a plorality of diseases or . neurological conditions. One or more optimal DBS parameters are predicted for the patient based on the statistical metrics of brain response and the DES functions! atlas via the predictor . system.

[0007] Published US. patent US 11,395,020 B2 relates to a system and method for identifying ; à patient-specific Henrosurgerr target location, The system receives brain imaging data for a . patient that includes tracts and networks in the patient brain, accesses à quantitative connectome : atlas comprising population-based, disease-specific structural and functional connectivity maps . comprising a pattern of tracts and networks associated with an optimal target area (OTA) : identified from a population of patients, and defines the patient-specific neurosurgery target A location based on à comparison between a pattern of the tracts and netwerks from the brain imaging data for the patient and the pattern ol tracts and networks associated with the OTA ,

PEER coc COO SES SS DL...

CHARITE — UNIVERSITATSMELHZIN BERLIN roriman mn Tegeth off identified from the population of patients in the quantitative connectome atlas. The quantitative . connectoms atlas comprises a disease-specific, population-based quantitative connectome atlas à that identifies an optimal target location for treatment associated with à maximal clinical . improvement for each disease in the population of patients, 7

Summary of the disclosure .

[0008] À first aspect of the disclosare relates to a method for suggesting personalized and ; symptom-specific DBS stimulation parameters, comprising the steps oft creating a thee , dimensional symptom network library comprising a data-driven thrce-dimensional model of : sympiom-specifie tracts in stereotactic space created on à detailed and inclusive pathway atlas; 7 registering patient data with the fhrec-dimensional symptom network library for analysing how 7 a single patient’s électrode maps to it: deriving and suggesting stimulation parameters as a . function of a baseline symptom severity profile in each patient. 3

[0009] Another object of the invention relates to à system for predicting DBS parameters, the . system comprising an interface for providing DBS parameters 10 an electrode and receiving ; signals from the patient; and à computer programmable unit comprising a data storage for à . three-dimensional syraptom network library comprising à stored data-driven three-dimensional : model of symptom-specific tracts in stereotactic space created on a detailed and inclusive : pathway for predicting DBS parameters. 2 10010] Still other aspects, features, and advantages of ihe present invention are readily apparent 7 from the following detailed description, simply by illustratine preferable embodiments and , implementations. The present invention is also capable of other and different embodiments and its several details can be modified in various obvious respects, all without departing from the spirit and scope of the present invention. Accordingly, the drawings and descriphons are to be a regarded as illustrative in nature, and not as restrictive, Additional objects and advantages of A the invention will be set forth in part in the description which follows and in part will be obvious . from the description, or description or may be leamed by practice of the invention. . nn

Lu “A > € . iE eer TOA EFI BE Sear trays Tees ER .

CHARITÉ — UNIVERSITATSMEDIZIN BERUN rormann lagethon … 12077 20881 À eich Perv hares o

LU103178 À

Brief description of the figures = 0011) The invention will be described based on figures. It will be understood that the , embodiments and aspects of the invention described in the figures are only examples and do o not limit the protective scope of the claims in any way, The invention is defined hy the claims : and their equivalents. It will be understood that features of one aspect or embodiment of the . invention can be combined with a feature of a différent aspect or aspects of other embodiments = of the invention, in which .

[0012] FIG. | shows an exemplary electrode placement, wherein active contacts are visualized = separately for each of the three patient cohorts. = {0013 FIG. ZA shows symptom specific tracts in a sagittal view and magnified at the level of .

[0014] FIG. 2B shows the symptom-specifie tracts separately at the STN level with the other . tracts greyed out for spatial comparison, . 10015] FIG. 2C shows the segregation of symptoms within indirect pathway connections ES between STN and pallidum, following a similar rostrocaudal posteriofrontal gradient. T 10016] FIG. 20 shows the cortical origins of hyperdirect projections. ,

[0017] FIG. 3A shows tremor tracts include projections from the cerchellar nuclet (to thalamus) , as well as the cortical projections from primary motor cortex (do STNX . {0018} FIG. 3B shows tracts associated with axial symptoms include a brainstem connection to © the pedunculopontine nucleus region, 5 (00191 FIG. 3C shows segregating axial symptoms into gait items vs, all other lems reveals > that this connection is driven by gait (and not by other axial symptoms}. . (0020] FIG. 3D shows a comparison to the projection site with a matching slice from an = histological atlas (published by Conlombe and colleagues 2021 F rontiers in >

Neuroanatomy). ey co SS J Ss EY SA

Lu . ee = £ ke ; Tord ren oan ToacvaathosTr Es

CHARITÉ — UNIVERSITATSIMEDIZIN BERLIN ! OF HLS TIGE thon = 12077. 204814 teri funn Seager SH =

LU108178 À 10021} FIG, 4A shows a stimulation field of the same patient is shown together with symptom- = specific tracts. . 10022] FIG. 4B shows an example patient*s stimulation volume is shown alongside the optimal i tracts associated with global UPDRS improvements, .

[023] FIG. SA shows a quantitative representation of the method's performance as a function . of the contacts selected yields significant results (t= 2.62, p = 0.01). .

[0024] FIG. 5B shows clinical improvements of patients for which the method according to the : present disclosure suggested a matching or adjacent contact and patients for which the > method according to the present disclosure suggested a different contact :

[0025] FIG. SC shows contact choices suggested by the method according to the present .

[0026] FIG. 6 shows a comparison between clinical settings and settings suggested by the . method according to the present disclosure and in à single prospective patient. .

[0027] FIG. 7A shows a well-placed, standard omnidirectional (Medtronic 3389) electrode with 2 a single stimulation volume that equally covers all symptem-specifie tracts. .

[0028] FIG. 7B shows a hypothetical Fnture concept with a modern électrode. © 10029] FIG. RA shows an example fiber-tract from the pathway atlas (dashed line). © (60301 FIG. 8B depicts that the process is repeated across all fiber tracts in the pathway atlas 2 to create the symptom network Hbrary. .

[0031] FIG. 8C shows the single tract model (coding far global motor improvements) whoch 2 is gross-validated by estimating motor improvements in left-out patients based on their . activation of the tract. . (00321 FIG. 8D shows a more elaborate symptom-specific tract model repeats this procedure © four times (For each symptom tract) and weight estimates by baseline intensities of each © symptom. 7

RR RR ; SN AE

N , sta HET x , Enrrrnanennt TaratHntt Ss

CHARITE — UNIVERSITATSMEDIZIN BERLIN ro dran regen non 2

Eggs 8 Esegritecemomddie oe 12077. 2048ILU à directive! Fromers Simeon =

LU103178 {0633} FIG. 9 shows in an example a novel (hypothetical) patient who is treated based on the 7 method of the present disclosure. .

[0034] FIG. 10A shows a Subcortical view on tracts at the level of the subrortex. :

[0035] FIG, 10B shows à cortical view on fibers reaching the cortex. 7

[0036] FIG. 11A shows a S-fold cross validation. =

[0037] FIG, 11B shows a 7-fold cross validation. = {0038} FIG, 11C shows a 10-fold cross validation, = 10039] FIG. 12 shows statistics over 1000 iterations of k-10 cross validations. Left panel: The 7 greatest density of correlation values for 10-fold cross-validations. Right panel: 10-fald cross . validation estimates averaged across 1000 iterations both models, on

[0040] FIG. 13A shows bradykinesia and rigidity tracts visualized together. >

[0041] FIG. 138 shows Repeated analysis after regressing out bradykinesia improvements 2 from rigidity improvements and vice versa does not show qualitative changes on an anatomical “

Detailed description of the disclosure i

[0042] The technical problem is solved by the independent claims, The dependent claims cover further specific embodiments of the invention. [

[0043] The term stimulation parameters relate to the number of pulses per burst, pulse . frequency, pulse width, burst frequency, amplitude, applying combined pulses through a . plurality of electrodes and placement location for at least one electrode. . 10044] The term “symptom score” relaies io à subset of items from a validated, reproducible . scoring system for assessing disease severity and response to therapy. For instance, items that © measure severity, frequency ete. of tremor within the motor part of the unified Parkinson’s ES

A ss ER mi

CHARITE — UNIVERSITATSMEDIZIN BERLIN 5 0 ri mann Te a sthoff . 12077 2048110 Lira res Fo ray Arias : disease rating scale (UPDRES-I) would constitute a tremor “symptom score”, Changes in these itétns before and after surgery (or under DBS ON vs, OFF conditions) would relate te symptom score improvements, The term ‘symptom score” is within the meaning of the present disclosure © applicable to symptoms occurring in movement disorders. 10045} The present disclosure provides a method and à system and aims at making progress by . i) creating à three-dimensionsl circuit model for discase specific symploms and if) leveraging : the model in a method to suggest optimal stimulation parameters comprising a funetion of the : baseline symptom severity profile in each patient.

[0046] An extension of the model by further segregation into four symptom categories (tremor, 7 bradykinesia, rigidity, galt and remaining axial symptoms) has been exemplary developed . (comp, Fig. 3). Moreover, the concept can he readily applied to other diseases, such as obsessive-compulsive-disorder, where aymptoms have been segregated to obsessions, compulsions, anxiety, depression, cognitive flexibility, and global functioning. :

[0047] A model was developed according to the present disclosure and cross-validated on data â from three patient cohorts (N=43, Würzburg, N=35 Amsterdam and N=31, Berlin; comp. Table â 1). All patients snderwent bilateral STN-DBS to treat Parkinsons disease by using four contact . omnidirectional electrodes (Medtronic 3389) with stimulation activated in both hemispheres. â

Electrodes were localized and active contacts resided in the subthalamic region across all 129 A patients (FIG. 1). Clinical scores across all patients had an average Unified Parkinson Disease ,

Rating Scale (LUPDRS)-II baseline score of 44.59 £14.30 (SD) and mean improvement of $1.71 +24.26%. In the Würzburg cohort, the MDS-UPDRS was analysed, whereas the conventional â

UPDRS was investigated in the Amsterdam and Berlin cohoris, respectively (comp. Table 1). .

DBS centre | Age Disease UPDRS. Levodopa UPDRS LEDD Postop [M/female] | fsears] Duration HI response improvement réduction Imaging 7

Baseline [ei {Med OFF [%] .

SE AAA 0

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN Fartmann Tegethoff 7 "Wurgburg 1604 © 126445 SES LUE = 493%2ex 614 w= CT

Berlin | 600 + 104239 3R6 x 535 + 4534230 528 & MRIN=45) ir

Amsterdam [5053 x 127261 4703 Æ 688 & 469187 41824 ET .

Table 1. Demographic information 7

[0048] For creating a symptom network library according to the present disclosure, an extended 7 version of the DES tractography atlas (Middlebrooks, E. H. et al, Am. I Neuroradiol. 41, 1558-1568, 2020) was used to define anatomical connections from and to the STN (see l methods), Using the DBS fiber filtering method (Baldermann, LC, et al, Biol. Psychiatry 83, 735-743, 2019), it was investigated which tracts included in the pathway atlas correlated with 4 improvements in bradykinesia, rigidity, tremor, and axial symptoms (symptom network . library). The symptom network library may consider other tracts which are related to other { symptoms so that the symptom network library can be designed indication specific. . 10049] DRS fiber filtering is à mass-univariate approach which leads ta statistical coefficients 7 (such as, in the present case, Spearman’s rank correlation coefficients) for each tract that : connected to the group of stimulation volumes. Correlation coefficients were corrected for . multiple comparisons using the Benjamini-Hochberg technique (false discovery rate) at an a- . level of 0.03, and only significant tracts that survived were retained. This significant set of - fibers revealed a distinct rostrocaudal gradient of symptom improvements at the subthalamic ; level (FIG. 2), Importantly, the same gradient was reflected across the entire network, 7 suggesting a “sympiomatopic” arrangement both within hyperdirect pathways and pallidésubthalamie projections. The rostrocandal order of indirect projections (from pallidum . 16 STN) was consistent with the ordering of hyperdirect/corticosubthalamie findings mentioned E

[0050] Tracts associated with tremor improvements projected from primary motor cortex to the ; posterier-miost region of the motor STN. As expected, tremor tracts additionally included the . decussating cerebellothalamie pathway. Tracts assoctaied with rigidity improvements projected - eee oa NN SS SS SS Ss

EE EE

HW +é vs HR AE ; Sub San 16 yeti en FF ;

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN TONTITIONIT FEQOTON Es 13077 2648110 RE ee ;

LU103178 % from pre-Supplementary Motor Area (SMA) and similarly frontal premotor regions fo the . anterior part of the subthalamic premotor region. In between the tremor and ré gidity tracts, tracts ; that associated with improvements of bradykinesia and axial symptoms overlapped on the anteroposterior axis. However, while bradykinesia tracts entered from the medial surface of the :

STN, axial tracts terminated at its lateral aspect (sec insets in FIG, ZA) — both originating from .

SMA and laterally adjacent cortical regions, Axial tracts further included a connection to the , brainstem confined to the pedunculoponting nucleus (FPN) region. . [00511 FIG. 2 shows a symptom-network library according to the present disclosure. FIG. ZA 7 shows symptom specific tracts in à sagittal view and magnified at the level of the STN. The . sympiom specific tracts follow a rosttocaudal posteriofrontal gradient with tremor most . ocuipitalposterior, followed hy bradykinesia, axial symptoms, and rigidity. All shown tracts are . significant after correcting for multiple comparisons {p < 6,05), Iris to be, noted that the tracts , are in proximity to one another, making it possible to modulate ail — or most of them with a 7 single well-placed electrode (which agrees with clinical experience, where a single stimulation . field is often capable of alleviating many/all motor symptoms 10 some degree .

[0052] Referring to FIG. 2B, the symptom-specific tracts are visualized separately at the STN ; level with the other tracts greved out for spatial comparison, Insets represent permutation test . and 10-Ibl cross validation results and for each symptom specific tract. FIG. 2C shows the egregation of symptoms within indirect pathway connections between STN and pallidum, . following a similar rostrocaudal posteriofrontal gradient. FIG, 2D shows the cortical origins of : hyperdireet projections. Tracts associated with tremor improvements originate in the primary motor region of the cortex, whereas the tracts associated with improvements in hypokinetic . symptoms originate from premotor regions in à more interspersed fashion. -

[0053] As mentioned, all tracts shown in FIG. 2 are significant after correction for multiple 7 comparisons. Robustness of this model was tested further. To do so, first, syrptom-specific tracts were subjected tp a permutation analysis were subjected. Here, all but the tremor tract significantly explained more variance in outcomes than re-caleufaied tract models after permuting improvement values across patients 1,000 times {p < 0.001). Second, tract models 7 were subjected to cross-validations were subjected. Again, all but the tremor tract explained significant amounts of variance when subjected to 10-fold cross-validations (bradykinesia: R = 7

CHARITE — UNIVERSITÄTSMEDIZIN BERLIN Forim ann Tegethofi 12077 2048110 HCN NGA SE. . 0.23, p = 0.007: rigidity R = 0.18, p = 0.04: axial symptoms R = 0.23, p = 0.011, also sec FIG. . 2% Critically, while all symptiems except tremor could be calculated on the full cohort, the analysis of tremor included only 29 of the 129 patients, given 100 patients had a baseline tremor . score below 3 or improved by 100% (which explains why tramor models, albeit being : significant, were not as robust as the other tracts when subjected to permuplation or eross- 7 validation testing). The tremor analysis had to be restricted in a manner that since patients . without substantial tremor at haseline could of course not improve much by DBS (even if . stimulated optimally), and the ones that improved by 100% could have potentially improved further if more tremors would have been present, at baseline, The same issue does not apply to the other symptoms given the factorial structure of the UPDRS and the disease (also see â {0054} FIG. 3 shows the anatomical validation of s symptom network library. As shown in FIG, 2 3A, streamlines associated with improvements oftremorineluded cerebelle-thalamic pathways, ;

Together with cortical projections precisely originating from primary motor cortex (HG, 2D), , these exact connections have been widely implicated with tremor across a large body of the , literature (Hassler et al, Brain 83, 337-350, 1960; Akram, HL. et al, Neurolmage 158, 332-345, “ 2017; Coenen, V. À. et al. Acta Neurochir, (Wien) 18, 130-14, 2020; Helmich et al, Ann. :

Neurol. 69, 269-281, 201 1; Helmich et al, Brain J. Neurol. 135, 3206-3226; 2012), Similarly, . tracts associated with axial improvements included a brainstem connection confined to the PPIN : nuclens region {FIG, 3B). The PPN has served as the most promising stimulation target to treat 3 gait problems — which are part of the axial symptom group — although with variable success . {Mazzone et al, Neurosurgery 73, 894, 2013; Zrinzo, L. ot al, Brain 131, 1588-1598, 2008). .

Given this elinical relevance, it was tested whether these connections could be specific tu gait - improvement (or would be associated with all axial symptoms). To do so, gait-specific . symptoms were separated from all other axial symptoms. While gait-specific symplom s improvements alone isolated the same brainstem connection, repeating the analysis with all ; axial symptoms without the gait-items did not include this connection {Fit 3C & D). Since . axial symptoms and especially gait problems are by far the most problematic symptoms to treat . with DBS (Klawans, HL L., Mov. Disord. Of£ J. Mov, Disord. Soc. 1, 187—192, 1986; Bonnet . et al. Neurology 37, 1539-1542, 1987), specific value in this finding wag observed. ;

Furthermore, these results again match known associations between anatomy and motor

CHARITÉ — UNIVERSITÂTSMEDIZIN BERLIN Fortmann Tage! hoff functions (Le., à specific association between gait improvements and the PPN region), which : may qualitatively validate the present results further, .

[0055] Prior findings hinted at shared neural substrates for bradykinesia and rigidity (in contrast . to tremor) (Kühn, À, A. etal. Exp. Neurol, 215, 380-387, 200%), which does not directly match ; the degree of separation which was identified between rigidity and bradykinesia tracts. To explore this further, rigidity improvements were regressed out from bradykinesia improvements . and vice-versa and repeated the analysis, which led to the same se gregated results. 2

[0056] While all results shown in FIG. 2 were statistically significant after correction for multiple-comparisons, 11 was further explored whether the entire model (not sach individaai : symptora tract) would he able to robustly estimate global motor improvements when subjected . to eross-validations, FIG. 4A shows this analysis. To de so, patients were randomly assigned : to one of 10 folds. Then, the four-tracf model was iteratively recaloulated, sach time leaving : out one-fold of paticuts. Stimulation sites {expressed by electric field vector magnitudes, E- 7 fields) of left out patients sere overlaid with the tract model and multiplied with the weighted © optimal tracts (sach expressed by the Spéarman*s rank correlation assigned to each tract), - feading to fiber scores for each stimulation field and each symptom. Then, these estimates were . weighted based on the prenperative symptom score cach patient had (Le, for a patient with severe tremor at baseline, high fiber scores originating from the tremor tract were weighted . more strongly). From this, à single combined fiber score emerged, which was mapped to ; {predicted) UPDRS] improvements using a linear model calculated on the training set (Le. ÿ of 10 folds). Critically, the model did not see any data from the left-out fold which was 7 estimated in each iteration (see methods for details). These predicted improvements 7 significantly correlated with relative-UPDRS-IH improvements empirically observed across . patients (R = 0.34, p < 0.001, mean absolute error: 17.91% + 14.1%. FIG. 44), ;

[0057] The sume analysis was repeated after calculating a single tract that directly coded for . global %-UPRDIS-II improvements. As can be seen in FIG. 4B, this model was simpler. Instead ( of weighting fiber scores originating from each symplom-tract, a single fiber score was 7 calculated per patient, directly coding for $6-UPDRS-LI improvements. This single tract model . more readily mimics previous work, which aimed at determining the optimal structural . connectivity profile for global motor improvement (Hom, A. et al, Ann, Neurol. 82, 67-78, ;

CHARITÉ ~ UNIVERSITATSMERIZIN BERLIN Forfmann Tegethoi :

LU103178 § 2017; Trew, S. etal. Neurolmage 219, 117018, 2020). In direct comparison to the four-sympton 7 model, the single tract model performed worse (R = 8.27, p = 0.008, mean absolute error 7 18.57% « 14.3%. FIG. 48). :

[0038] During 10-Fold cross-validation, sets of patients were randomly assigned te sets which ; ; were then Îteratively left out. This process can be done In numerous ways tie. assigning 2 different combinations of patients ta folds). To rule owt that the current sélection was not : generalizable, the process was repeated 1,000 times, each time randomly assigning patients to ; : foids. This led to very comparable outcomes (multi-tract model, average R = 0.35, p < 0.008; | ı single tract model, average R = 0.25, p = 0.008), FiG.s describing statistics across the 1000 2 iterations, including error plots, and the distribution of R- and p- values across the folds are 7 provided as FIG. 11 and FIG 12. :

[0039] Finally, it was taken care to rule out that the choice of k {in E-fold cross validation) did ‘ ) not have a significant impact on the results, Therefore, the analysis was repeated with S-foid ı and 7-fold cross validation which again yielded similar results for both the multi-tract (R = : 6.33, p < 0.001 and R = 0.38, p < (L001, respectively) and single tract model (R= 0.26, p = ; . 0.008 and R = 0.27, p = 0.008} s

[0060] To take the very first steps toward an application in clinical practice, the method ) [ secording to the present disclosure was developed capable of maximizing stimulation of either 7 symptom tract in novel patients. This method according to the present disclosure is currently capable of suggesting monopolar stimulation settings and simulates stimulation fields at each ; ı contact and predicts their outcomes analogous to the multi-tract prediction analysis described ı above. Considering an individual patient’s baseline motor scores, the method seeks to optimize : contact position by maximizinge stimulation volume overlaps with the relevant sympiom- , specific tracts. In consequence, the exact same electrode placement would lead to different . settings in a patient with eg. high tremor severity vs. high axial impairment. The method according to the present disclosure could be used to update sngpestions for patients during the span of their disease, as their symptom profile develops over the years. ;

[0061] To validate the method according to the present disclosure, it was first applied to all 7 patients within the retrospective cohort to suggest optimal parameters, based on the baseline symptom profiles of each patient. Proposed settings were then compared to actual clinical A

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN Fortmann Tegethofi

LU103178 £ parameters by categorizing contact selection as matching, adjacent, or different with respect to ; the ground-trath parameters at 12 months follow-up. In relationship to chosen contacts, the s method according to the present disclosure suggested parameters that were significantly different from à random distribution {chi-square goodness of fit test, chilstat = 188.09, p < s 0.001; FIG. SA). Further, a three-way ANOVA revealed significant effects on clinical © improvement in relationship to the method according to the present disclosure suggesting the . same, an adjacent or à different contact than clinicians had selected (F(2,189)= 4.13, p = 0.02).

Fost-hoc tests revealed that clinical improvements In patients for which the method according to the present disclosure recommended the same contacts were significantly higher from clinical improvements for patients in which the method according to the present disclosure ; recommended different contacts (p= 0.01, FIG, 3B}. :

[0062] Second, it was tested whether the method according io the present disclosure could ~ in ; theory — have improved clinical improvements on a cohort level and, if so, quantify the extent, ;

Critically, actual clinical improvements for the parameter suggestions made by the method according to the present disclosure could not be obtained (given the retrospective nature of the ; cohort), hence this analysis hinges of the predictive model being valid. In other words, É predictions of outcomes for parameters suggested by the method according to the present . disclosure were compared te predictions of outcomes for clinical parameters, Based on this, ; 128 of the 129 patients (99.22 %) could have been optimized by the suggestions made by the . method according to the present disclosure (1 = 20.34, p < 0.001), and by an average of 21.12 . % on the UPRDS-EI scale 6 9.64%, min: 0.64%, max: 45.61%, FIG, § C & D). It has io be . reiterate, however, that this analysis was entively carried out in silice with the sole aim to ; quantify a potential room for tmprovement, ; 10063] Given the successful retrospective validation of the method, prospectively DBS stimulation parameters suggested by the method according to the present disclosure in a single Ë patient for 24 hours were applied. The 71-year-old male patient had undergone STN-DBS for :

PO with bilateral directional clectrodes (model: Boston Scientific Cartesia) at the University .

Hospital Würzburg six years after diagnosis with PD, Testing was carried out four months after } surgery, where DBS had led to an LEDD reduction of 63%. All measures were carried out after . withdrawal from dopaminergic medication for > 12 hours (Med OF F). Clinical programming . settings for this patient had been carried out at 1.3 mA on contact 2 and 4 (185 Hz, 60 us} in . the left electrode and 1.9 mA on contact 4 (80%) and contact 7 (20%) (185 Hz, 60 us) in the 7

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN Forman A lagethofi Ë 12077,20681LU Ae .

LU103178 £ right electrode (FIG, SAX This setting had led to an MDS-UPDRS-HI improvement of 33 puts ; (47.14%: to 37 from a DBS OFF baseline of 70 points). ;

[0064] Based on the baseline parameters and electrode localizations of this patient, the method . according to the present disclosure suggested to activale contact 7, with an amplitude of up to 5 4.5 mA in the le electrode and contact 8, with an amplitude of up to 5 mA in the right electrode. ;

These settings were switched on for 24 hours after adjusting amplitudes based on the observed 7 side-effect threshold (3.0 mA and 2.8 mA; FIG. 6B) and led to an improvement of 42 points Ë (60%: to 28 from the same DBS OFF baseline of 70 points). Of note, settings led to a 100% improvement in tremor, 60% improvement in axial symptoms, 50% improvement in rigidity : and 25% improvement in bradykinesia, which was more or equal for each symptom when 7 compared to the clinician programmed improvement (32.17%, 40%, 80% and 25%, 2 respeotively) i

[0065] The present disclosure presents for the first time the use of à three-dimensional model s of symptom-specifio tracts in stereotactic space created in a data-driven fashion on a detailed ; and inclusive pathway atlas. The disclosure shows that this symptom-network library is robust to eross-validations. As an example, a possible use has been shown on a single tract model Ë caléulated on global UPDRS-TH improvements, Based on the generated model, à method capable of suggesting personalized and symptoru-specific DBS stimulation parameters was . created and preliminarily validated. Within à retrospective database, patients, for which clinical :

DRS programming settings agreed with suggestions made by this method profited significantly more from DRS than the ones for which clinical choices differed. Subsequent modelling , suggested potential room for improvement of DBS settings in this three-center cohort. Finally, 7 the DRS settings suggested by the method according lo the present disclosure were ı prospectively applied in a single patient, This improved motor symptoms when compared to : the clinical setting. ;

[0066] The results provided by the present disctosure support the notion that different networks , correlate with imprévements of cardinal symptom categories in the example for Parkinson's -

Disease, Namely, it was possible to segregate the basal ganglia thalamo-cortical motor nop by 7 symploms arranged along a rostro-caudal gradient within the seasotimotor-premotor functional ; zone of the STN. While interspersed on a cortical level, each of the symptom-specific tracts ; predominantly originated from different cortical regions. Further, tremor connections included

CHARITÉ — UNIVERSITATSMEDIEN BERLIN Fortmonn Tegethoï s 120772048340 Leu re re 7 cormmections from cerebellar nuclei while axial symptoms included connections io PPN region 2 10067] H is important to clarify at this point that the results do not suggest that ons symptom demain can be modulated independently by a specific tract. There were considerable overlaps : hefween connections, most especially on à cortical level and along the indirect (pallidosubthalamic) projections. On the other hand, projection zones of hyper direct (cortical) . input to the STN seemed quite segregated — albeit very close to one another. At first glance, . this could be seen ax iff it bpposed clinical experience: Indeed, the same DBS setting can s madulate many symptomes at once, seemingly with similar intensity. However, this notion does . not conflict with the presented results: it has to be emphasized that identified tracts are very s close to each other, spanning across a region of millimetres within the sensorimotor functional sane of the STN level. As FIG. 7A shows, a single well-placed electrode may produce a stimulation volume that modulates all identified tracts (and hence symptoms), simultancousty.

However, FIG. 7B shows potential use of the tract model with a modern l6-contact segmented electrode {such as the Boston Scientific model Cartesia X) Using Multiple Independent Current .

Control (MICC) technology, distinet stimulation volumes can be activaled along the same ; électrode, wach with different amplitudes and frequencies (Timmermann, L. et al, Lancet .

Neurol, 14, 693-701, 2015), In the hypothetical example shown in FIG. 7, one could steer one volume of high frequency (180 Hz) to the tremor tract and a second of low frequency (25 Hz) tu the axial & gait tract to treat the two symptoms as optimally, as possible. .

[0068] Hence, the results of the present disclosure could potentially become clinically relevant: 7

First, segmented electrodes allow for an increasingly refined precision on where exactly to . focus stimulation volumes, This feads to an explosion of parameter space where imaging guided 7 methads will become indispensable (Roediger, J. et aL, Mov. Disord. 37, 574-584, 2022; .

Roediger, J. etal. Lancet Digit. Health 5, e59-e70, 2023). With imaging methods and electrode : focalizations becoming ever more precise, the present disclosure provides a method for using . symiptom-network libraries such as the present one 10 finé-tune stimulation settings depending . on the symptom demand of each patient. This will facilitate prospective validations of present . results (Roediger. JL et al. Lancet Digit. Health 5, €59-e70, 2023; Gadot, R. et al, Biol. :

Peychiatry, 2023), Second, while à single stimulation site of a well-placed electrode can cover ; 16

CHARITE — UNIVERSITATSMEDIFIN BERLIN Forimann Teget hoff ; 120772048110 Le . the majority of identified tracts , it might «till mater where the peak of the stimulation resides 2 {and not all leads are placed optimally}. 7

[0069] To develop this further, a proof-of-concept process was introduced, to obtain . suggestions of symplom-specific DBS parameters. The process allows fo sel weights of . symptoms that are both prevalent and burdensome for à patient and which adjust parameters to maximize stimulation of symptom-speciiic networks, For instance, one could ran the process 3 with a symptom profile of a typical tremor-dominant patient with high intensity of tremor and , lower intensity of bradykinetic-rigid symptoms, which would favour settings that maximally 7 target the tremor connections from primary motor cortex and cerebelhum. ;

[0070] The present disclosure proposes to first use normative connectivity data to define ; symptom-specific tracts où a group level, Next, patient data is spatially registered with the ; resulting sympiom-tract-library to analyze how a single patients electrode maps to it. Then, , the actual personalization of the approach takes place on the level of svmptoms, using à concept . which can be termed network blending, in the past (Hollander, Bo et al, Pro-gress in .

Neurobiology 102211, 2021; Hollunder et al, Biologieat Psychiatry: Cognitive Neuroscience and Neuroimaging 6, 939-941, 2021). The concept is to blend — or weight — the identified : symptom networks to derive at an optimal stimulation target for the symptom profile prevalent in an individual patient. Brain sensing combined with machine leaming might provide mediate feedback to the DBS system that could automaticaly inform the process to switch 2 network targets based on the individual need of the patient at the time a symplons — such us ; iremor under stress — emerges (Merk, T. et al. Exp. Neurol, 351, 113595, 2022). One day, this could open new horizons to an integration of adaptive DBS technology and symptom specific ; connectomics, towards an individealized precision medicine approach to DBS in real-time. 2 {0071} [In conclusion, the present disclosure provides a meihod which is based on a three- 7 dimensional circuit model for disease specific symptoms and its use to suggest stimulation parameters comprising a function of the baseline symptom severity profile in a patient. :

Moreover, the present disclosure introduces a method capable of leveraging this model to . suggest symptom-specific stinmlation parameters for DBS programming, which may ultimately . improve clinical outcomes and patient satisfaction. '

LU : = —_ en dem a Sense Veen FE ES

CHARITÉ — UNIVERSITÄTSMEDIEIN BERLIN cor 3 ann agen haf 2 12077. 2048114 HR SR SL . tose 8

[0072] For showing the effects of a method according to the present disclosure, 129 patients nl who underwent STN-DRS for Parkinson's Disease (FD) were exemplary retrospectively = studied. OF these, 51 were treated in Berlin, 43 in Würzburg and 34 in Amsterdam (patient = characteristics and demographic data are provided in Table 3). .

Symptom UPDRS- HMDS tems Mean and standard deviation of . a | | basetine clinical scores on

Bradykinesia : 23 (Finger tap Rand L) Baseline: = 24 (Foe tap Rand D) 17.24 7.00 on (Hand move R and L} Improvement: TE 26 (Leg agility Rand Lj 44,99 32.98% i and MDS — Tas tapping {K and D). =

Rigidity 1 22 (rigidity neck, rigidity arm R and L, Baseline: ) rigidity leg R and LA. $3 +510 5

Taprovement: i 70.40 à 28,33% .

Tremor | 20 {tremor rast arm R and L, tremor rest leg Baseline: .

Rand L} 458 + 5,24 .. 21 faction tremor hand R and LYMDS Improvement (twesholded): . {kinetic tremor hand R and L} 57.24 + 42.38% .

MIS tremor (lip/jaw} .

MDS constancy of rest remor ..

Axial 27 {Arise chair) Baseline: i 28 (posture) 00x 471 i 29 {gait} improvement:30 (postural stability) 46.93 = 30.24% . : 31 (Body bradykinesia) i

MDS-Freszing of Gait (FC) :

Table 3: Siatistios of clinical scores and sub scores across cohorts. a 10073] All patients were bilaterally implanied with two quadripolar DBS electrodes (model € 3389: Medtronic, Minneapolis, MN). The study was carried out in accordance with the 7

Declanation of Helsinki and was approved by the institutional review board of Charité- 1

Universititsmedizin, Furthermore, the method prospectively apply DBS settings based on the 1 retrospective database 10 a 7i-vear-old malo with PD who underwent DBS surgery at the .

University Hospital Würzburg, This application was approved by the institutional review board . of University Würzburg, i 6 ;

CHARITE UNIERSITATSVIEDIZIN BERLIN rorimann Tegethoff ; 12077.20481LU Le a 7

LU103178 § 10074] Percentage improvements measured by the motor part of the Unified Parkinsons ;

Disease Rating Scale (LPDRS-IN) were caloulated based on the difference between prevperative and postoperative scores divided by preoperative scores as a measure of global , treatment outcome. For the Würzburg cohort, the revised MDS version of the UPDRS-IT was ; used. Improvements of UPDRS items were similarly calonlated that represented Four major s motor symptoms in PD: bradykinesia (items 23.24.2526 which measure finger tapping, hand . movement, rapid afternating and leg agility [& MDS items Toe tapping where available), . rigidity {items 22 which measure rigidity of the neck, arm & leg), tremor {items 20 and 21 of the UPDRS [& MDS tremor Gems: MDS postural tremor, MDS kinetic tremor, MDS tremor . test lip/jaw and constancy of rest tremor where avalable]}, and axial symptoms (items 27, 28, = 29, 30 which measure posture, postural stability, and gait [& MDS îtems Freezing Of Gait . where available]). Gait scores (tems 29 and 30) were further singled-ouf in a sub-analysis, as 2 were a combination of all axial symptoms without these gait items. In the analysis, tremor , subscore at the baseline had a high standard deviation {4.27 +5,62 points), when compared to 2 ihe other sub score values (Table 3). Patients with tremor scores below two points at baseline, ; and those patients who improved to 100% relative improvement, were excluded from tremor . analyses, Multispectral preoperative MRI scans were acquired during clinical routine to define . patient-specific anatomical targets. Post-operatively, patients either underwent CT (N= 84) or -

MRI scanning (N = 45) to localize electrodes.

[0075] DBS électrodes were localized using Lead-DBS software (Horn et al, Nearolmage 134, . 293-316, 2019: Horn, A. & Kohn, À A. Neurolmage 107, 127—135, 2015) following the 2 revised protocol of version 3 (Neudorter et al, Neuralmage 268, 119862, 2023). In brief, this . included linear co-registration of post- and preoperative images using Advanced Normalization .

Tools (ANTS) {Avants et al, Med. Image Anal. 12, 26-41, 2008). Co-registered images were ; then normalired into the ICBM 20006 Nonlinear Asymmetric MNT”) template space using . the ANTs SyN approach with the Effective: Low Variance + subcortical refinement protocol 2 as implemented in Lead-DBS (Bwert, S. et al, Neurolmage 184, 586-598, 2019). The results of each pre-processing step were visually inspected and refined if necessary. Normalization . errors In particular were revised using the WarpDrive module available in Lead-DBS . (Neudorfer, ©. et al. Neurolmage 268, 119862, 2023). Following pre-processing, DBS . clectrodes were localized using the phantom-validated PaCER approach (Husch et al, ;

Neurolmage Clin, 17, 80-89, 2018) for postoperative CT or the TRAC/CORE method or ;

CHARITE — UNIVERSITÄTSMEDIZIN BERLIN Fort TIN Tegeathoft , 12077.20481LU Re STR non . anual localization method for postoperative MRI data (Horn, A. & Kühn, A, À, Neurolmage ,

[0076] To estimate the stimulation volume. electric field vector magnitudes around the 7 electrode (E-Fielde) were caleulated. This was done based on a four-compariment mesh distinguishing gray and white matter, slectrode contacts, and insulated parts. Gray matier regions were defined by the DISTAL atlas (Fwert, S, et al. Neuroïmage 170, 271-283, 2018). .

An adapted version of the FieldTrip-SimBio pipeline (Vorwerk et al. Biomed. Eng. OnLine s 17, 37, 2015) was then used to solve the static Formulation of the Laplace equation on 8 discretized domain represented by the tetrahedral four-compartment mesh.

[0077] Creating an anatomically inclusive pathway atlas for the subcortical region: Multiple 7 network mapping approaches have used data-driven connectomes derived from diffision- , weighted imaging based tractography {Baldermiann et al, Biol. Psychiatry 85, 735-743, 2019: .

Hom et al, Ann, Neurol. 82, 67-78, 2017; Al-Fatiy et al, Brain 142, 3086-3098, 2019; Li et . al, Nat. Commun, 11, 3364, 2020). However, especially in the subthalamic region, these datasets often lack accuracy especially regarding thin projection bundles, such as the . pallidothalamic projections (ansa and fascicalus lenticulares), Edinger's comb fibers or the . cerebellothalamic pathways (Middlebrooks et al, Am. J. Neuroradiol. 41, 1358-1368, 2020; .

Petersen et al, Neuron 104, 1056-1064,e3, 3619; Noecker et al, Neuromadulation J. Int. ;

Neyrontodulation Soc. 24, 248-258, 2021; Alba et al. Mov. Disord. 35, 75-80, 2020: Horn et ; al., Neurology 92, e1663-+1664, 2015), To overcome this limitation, a key development was . the concept of manually curated pathway atlases defined using prior anatomical knowledge (Hamani et al, Brain 127, 4-20, 2004; Marani et al, Adv. Anat. Embryol. Cell Biol, 198, 1-- : 113, vii, 2008). These datasets comprise meticulously curated pathways and are hence Free from ) false-positive connections (tracts that do not exist in the brain). However, one downside of these ; atlases is that tracts that have not been defined by the anatomical team can also not be identified : as critical for DBRS success by post-hoc use af the atlas. This could lead ta faise-negative ; conclusions (not seeing tracts that exist in the brain). For instance, certainly the most accurate ; atlas defined by Petersen et al, lacks bundles such as the anse subthalamica (Petersen et al, ;

Neuron 104, 10356-106483, 2019), connections between STN and PPN or strigtomgral J connections, among others, simply because an anatomical team must focus on a number of : tracts and cannot exhaustively include all tracts that exist in the brain. Furthermore, the STN

LU nul - | .

Le ee Reben, Tavera bbe $F =

CHARITÉ ~ UNIVERSITATSMEDIZIN BERLIN i orfrmann FEET S ; $ Fépgct § BER sois ES 12077 2048104 {iol Smt AHR =

LU103178 2 receives cortical (hyperdivect) input from the entire frontal cortex (Yeh, Nat. Commun. 13, . 4933, 2022; Isaacs et al, Front. Neurosnat. 12, 2018), which is not represented in typical © pathway atlases. To overcome this limitation, here, existing atlases (Middiebrooks et al, Am. J.

Neuroradiol 41, 1558-1568, 2020; Petersen et al, Neuron 104, 1056-1064.63, 2019) were . compiled comprising cortical and subcortical pathways traversing the subthalamus and added 7 missing connections based on anatomical data (Table 2). .

Tract name Was i Adopted from | Calculated. from HCP . included in | Petersen et al, © | 1000 diffusion template . version 1 5 | {seeds defined by) .

Anterior Thalamie Radiation = ; | .

Cerebelfothalamic Tract x : =

Dentatoruhtothalamie Tract = 2

Ansa Lenticularis | # ; .

Fascivalus Lenticnlaris x | .

Fasciculus Subthalamicus x . {fibers of Edinger's comb) © projections fromli BA2 | x {9 =

BA3 | x (©) .

BAGS | (©) :

BAS | Ex) .

BA 1D | (© 2

LBA 13 x (*) .

BA ' (7) .

BA2S | x {9 2

BAR | “(0 7

BA 45 x

BA 47 | =) ©

Substantia nigra, reticular part | Px) i = STN 1

Substantia nigra, compact part | L(y .

Pedunculopontine nucleus +. [ x 23 =

STN | ; | .

Table ? Tracts defined fn the revised version 2 of the DBS tractography atlas (comp. FIG. 10). .

[0078] Missing connections not represented in any atlas were calculated following the same . methodology used to create the first version of the atlas, described in detail elsewhere .

nr — UNIVERSITATSMEDIZIN BERLIN rormann Tagethoi - 12077 2048110 lus ARTE 2

LU108178 À {Middiebrooks, E. H. etal., AINR Am J Neuroradiol 107, 64, 2018). Brietly, using a diffusion template (Yeh, FC. ot al, Neuroimage 178, 37-68, 2018) compiled fram 1065 participants of the Homan Connectome Project (Van Essen, D. C, et al, Neurolmase 62, 2222-2231 2012), seeds defined from various atlases (Table 2) were used to map their connections to and from 7 the STN {as defined by the DISTAL atlas, Ewert, 8. of al, Neuroimage 170, 271-282, 2018). -

As à result, this revised version 2 of the DBS tractography atlas (Middlebrooks, E, H, et al, .

Am. J Neuroradiol. 41, ISSS-1568, 2020) consists of a more exhaustive set of connections . from and to the STM. The entire atlas is shown in FIG, T0, : 10079] Multi-Tract implementation of DBS fiber filtering: The present method is built upon the .

DRS fiber filtering concept introduced in (Baldermann, LC, et al, Biol. Psychiatry 85, 735- . 743, 2019) and extended in (Li, N. et al, Nat Commun. 11, 3364, 2020) to isolate tracts , associated with changes across multiple motor symptom domains (FIG, 8). In the first step, this 4 method was used to build exemplary a “symptom network Hbrary” associating tracts with improvements of clinical snbscores for tremor, bradykinesia, rigidity, and axial symptoms, .

Activation of these four tract sets correlated with improvements in respective symptoms. To be a included into the brary, each tract had to pass through low number of E-fields &2 % of total number of E-Helds) at a rather high peak intensity of »1.5 Vim This constrained was set up . since it was intended to exclude tracts that were not strongly modulated by any stimulation field 7 at all (which in theory could still obtain high correlation values if sub-threshold intensities . cenrelated with clinical improvements). Changing the arbitrary chosen values (72 % E-flelds , and >1.5 V/mm) eg, to >3 and > 4 V/mm did not qualitatively alter results, For each tract, .

Spearman's rank correlations were then calculated for each symplom group separately, by ; correlating the respective sub-score with the peak amplitude of each patient's E-field a given : streamiine passed through, This mass-univariate approach leads to a high number of rank ; correlation coetficients, which were thresholded at a p-value < 0.03 after correction for multiple . comparisons using the false-discovery rate (FDR). . {0080] Fstimating clinical improvements based on the symptom network library: Top 1000 7 fibers from the resulting tract sets were used to estimate clinical improvements (in à k-fold . cross-validation design) by overlaying E-Tields of left-out patients with respective symptom . tracts, Hare, the k (number of folds) was set lo 10 since this is a standard choice in the machine 7

CHARITÉ — UNIVERSITATSMEDIZIN BERLIN Fortmann Tegethoff © 130772048104 enr es .

LU103178 À learning field. Different numbers (eg, k = 7 and 5) were probed and it yielded similar, 7 significant results, ë [00811 Again, each pair of tracts and E-fields was quantified by the peak intensity through . which the tract had passed, multiplied by the R-value assigned to the tract (upvoting tracts . strongly associated with improvement of a specific symptom). For each E-field, these values ; were averaged resulting in à mean fiber score per sympiom and E-Field. Scores were then : weighted by the haseline scores of symploms prevalent in sach patient, As a result, moduiation . of the tremor tract in each patient would vield a greater impact in à patient with predominant . {0082} This led to = combined fiber score coding for a patient-specific blend of symptom improvements. Fiber scores were further mapped to the % improvements in each symptom s usine à linear model applied to the respective training cohort. The predicted improvement for : gach symptom was then transformed to the predicted % UPDRS improvement by a weighted average method, where each weight is the normalized baseline score for thal symptom. To . exclude the possibility of à random assignment of patients into the train- and test-set resulting in a significant result by chance, the process was iterated of k-10 cross validation 1000 times, wherein each iteration resulted in a random assignment of patients into the folds. 7

[0083] À method to suggest DBS programming parameters based on the symptom network 2 library: The present disclosure proposes a method capable of suggesting DBS stimulation . parameters based on the symptom network Hbrary (FIG. 9). The method considers all possible 7 monopolar DBS stimulation fields {within a range of 1 to 5 V in 0.5 V steps) for a given . electrode and estimates improvements for each hypothetical stimulation field the same way as described above. Tt then ranks predictions to identify the most efficacious stimulation setting.

As described earlier, tracts may be weighted, either by symptom severity, patient burden or J patient preference. For instance, à strong weight may be put on tremor improvement if the patient is burdened maximally by tremor, or à setting is sought for specific situations in which 7 tremor should be suppressed.

[0084] The method builds on and extends upon the ‘prediction’ as described above.

Hypothetical stimulstion volumes are generated (for all possible DBS settings), and their resulting symptom improvements are ‘predicted’, making the method versatile and extendable.

CHARITÉ — UNIVERSITATSMIEDIZIN BERLIN Fortmann Tegethoif ,

Hence, the cross-validation of the symptom network library (FIG, 4) equally supports validity of the method. However, the method was further examined by applying # to the same , retrospective cohort, Tt has been examined, for how many patients the method proposed the . same active contact, an adjacent, or a different one and compared clinical improvements for these subochorts. It has further been examined whether the method could have potentially led to enhanced clinical improvements in this cohort by comparing predictions for clinical stimulation settings with the ones suggested by the method (FIG, 3) . 10085] Another abject of the present disclosure is a system for predicting DBS parameters and providing the predicted stimulation parameters to electrodes. The system comprises a computer programmable unit (CPU) with a data storage comprising a fhree-dimensional symptom : network library comprising à data-driven three-dimensional model of symptom-specifie tracts , in stereotactic space created on à detailed and inclusive pathway, The CPU is further conti gured ; to register patient data with the created three-dimensional symptom network library for analyzing how a single patients electrode maps to it for deriving an optimal stimulation volume - for the patient’s symptom profile. 2

[0086] When DBS parameters have been predicted to meet the patient’s requirement, they can ; be applied via an interface of the system to electrodes of a patient which have been previously ; surgically implanted, Ît is obvious for a skilled person that at least one electrode will have to be : implanted in at least one region of the brain as well as there may have been a plurality of ; electrodes been implanted. The electrodes may be segmented electrodes which are able © , stimulate more than one tract ,

[0087] The system may further comprise a pulse generator which is required for generating pulses for DRS. It is also envisaged that the pulse generator may have been surgically implanted ) previously so that the predicted parameters of the system will be transferred through the . interface te the pulse generator which transmits the pulses to the at least one électrode. .

[0088] À sensor system may also be part of the system according to the present disclosure . which is able to sense and track DRS signals which are transmitted by the system. The sensor ) system may also be used to detect incoming signals from the electrodes and provide it to the 7

CPU so that patient data is at hand which can be used for predicting optimized DBS parameters, )

CHARITÉ — LNIVERSITATSMEDIZIN BERLIN Fonmionn Tegethof . 13077,20681LU tt

LU103178

[0089] The foregoing description of the preferred embodiment of the invention has been 7 presented for purpases of illustration and description. It is not intended to be exhaustive or to . jfrmit the invention to the exact form disclosed, and modifications and variations are possible in light of the above teachings or may be acquired from practice of the invention. The embodiment . has been chosen and described to explain the principles of the invention and its practical ) application, sa that those skilled in the art may use the invention in various snibodiments ; suitable for the particular application. It is intended that the scope of the invention he defined , by the appended claims and their equivalents. ,

Claims (20)

1. LU Se grès ‘ CT - seCHARITÉ — UNIVERSITATSMEDIZIN BERLIN For ON rege thon 12077.20481LU J Sion Habart Air { LU103178 CLAIMS

   I. A method for suggesting personalized and symptom-specific DBS stimulation parameters, comprising the steps of: - Creating a three-dimensional symptom network library comprising a data-driven three-dimensional model of symptom-specific tracts in stereotactic space created on a detailed and inclusive pathway atlas; : - Registering patient data with the three-dimensional symptom network library ; for analyzing how a single patient’s electrode maps to it; - deriving and suggesting stimulation parameters as a function of a baseline symptom severity profile in each patient. ;
2. 2. The method of claim 1, wherein creating the three-dimensional symptom network library ) comprises the steps of - defining symptom-specific tracts; . - determining correlation coefficients for each symptom-specific tract which ; connects stimulation volumes with any changes in a symptom score; À ~ determining significant symptom-specific tracts by correcting the correlation coefficients. ;
3. 3. The method of claim 2, wherein defining symptom-specific tracts relates to pathways ; from and to the subthalamic nucleus. ;
4. 4. The method of claim 2 or 3, wherein the stimulation volumes correlate with : improvements in bradykinesia, rigidity, tremor, and axial symptoms.
5. 5. The method of any one of claims 2 to 4, wherein the correlation coefficients are corrected with multiple comparisons using the Benjamini-Hochberg technique. 26 / ss NSS SS Ss SE Ss NN SAN

   CHARITE — UNIVERSITÄTSMEDIZIN BERLIN ronmann Tegetho LU103178
6. 6. The method of any one of claims 1 to 5, wherein the three-dimensional symptom network library comprises a four-symptom network library relating to the four cardinal motor symptom regions comprising tremor, bradykinesia, rigidity and axial symptoms.
7. 7. The method of any one of claims 1 to 6, wherein the three-dimensional symptom network library is validated prior to registering the patient data with it.
8. 8. The method of claim 6, wherein the stimulation parameters are suggested for each of the symptom-specific tracts.
9. 9. The method of any one of claims 1 to 8 wherein individual patient’s baseline symptom scores are considered as part of the patient data.
10. 10. The method of any one of claims 1 to 9, wherein the suggested stimulation parameters comprise maximizing stimulation volume overlaps with the respective symptom specific tracts.
11. 11. The method of any one of claims 1 to 10, wherein the suggestion of stimulation parameters considers segmented stimulation electrodes.
12. 12, The method of any one of claims 1 to 11, wherein weights of symptoms are considered in symptom profiles in the patient data for suggesting stimulation parameters.
13. 13. The method of any one of claims | to 12, wherein normative connectivity data is used to define the symptom-specific tracts on a group level and spatially registering patient data with the resulting three-dimensional symptom-tract-library to analyse how a single patient’s electrode maps to it.
14. 14. The method of any one of claims | to 13, wherein the stimulation parameters relate to suggest a location for a placement of electrodes prior to surgery.
15. 15. A system for predicting DBS parameters, the system comprising - an interface for providing DBS parameters to an electrode and receiving signals from the patient; and 27 Ms

    NN ss LU ; N hae wn gd "YY "4 fx Tans ony dh SN ££ CHARITÉ — UNIVERSITATSMEDIZIN BERLIN OPT EOS ION 12077.20481LU ENERO Atos LU103178 - a computer programmable unit comprising a data storage for a three- dimensional symptom network library comprising a stored data-driven three- dimensional model of symptom-specific tracts in stereotactic space created on a detailed and inclusive pathway for predicting DBS parameters.
16. 16. The system of claim 15, wherein the computer programable unit is configured to register patient data with the three-dimensional symptom network library for analyzing how a single patient’s electrode maps to it.
17. 17. The system of claim 16, wherein the computer programmable unit is configured to derive and suggest stimulation parameters as a function of a baseline symptom severity profile in each patient.
18. 18. The system of any one of claims 15 to 18, comprising a pulse generator for generating pulses which are provided through a connection to a patient’s at least one electrode.
19. 19. The system of any one of claims 15 to 19, comprising a sensor system for sensing and tracking DBS signals and/or for receiving signals from a patient.
20. 20. A method for the treatment of movement disorders, comprising the steps of - Connecting a system for predicting DBS parameters to at least one surgically implanted electrode in at least one region of a patients brain; - Applying DBS parameters to the at least one electrode of a patient suffering from a movement disorder, wherein the DBS parameters were previously determined by creating a three-dimensional symptom network library comprising a data-driven three-dimensional model of symptom-specific tracts in stereotactic space created on a detailed and inclusive pathway atlas; registering patient data with the three-dimensional symptom network library for analyzing how the at least one electrode maps to it; and deriving and suggesting the DBS parameters to be applied as a function of a baseline symptom severity profile in each patient. 28