Academic Journal
Limitations of simple machine-learning surrogates for cross-tissue prediction of radial177Ludose point kernels.
| Τίτλος: | Limitations of simple machine-learning surrogates for cross-tissue prediction of radial177Ludose point kernels. |
|---|---|
| Συγγραφείς: | Ennassiri H; Medical Physics and Radiation Protection Department, Centre Hospitalier Intercommunal Toulon-La Seyne-sur-Mer, Toulon, France. |
| Πηγή: | Biomedical physics & engineering express [Biomed Phys Eng Express] 2026 Sep 21; Vol. 12 (5). Date of Electronic Publication: 2026 Sep 21. |
| Τύπος έκδοσης: | Journal Article |
| Γλώσσα: | English |
| Στοιχεία περιοδικού: | Publisher: IOP Publishing Ltd Country of Publication: England NLM ID: 101675002 Publication Model: Electronic Cited Medium: Internet ISSN: 2057-1976 (Electronic) Linking ISSN: 20571976 NLM ISO Abbreviation: Biomed Phys Eng Express Subsets: MEDLINE |
| Imprint Name(s): | Original Publication: Bristol : IOP Publishing Ltd., [2015]- |
| Ιατρικοί όροι (MeSH): | Radiotherapy Planning, Computer-Assisted*/methods , Machine Learning*, Monte Carlo Method ; Humans ; Algorithms ; Linear Models ; Computer Simulation ; Phantoms, Imaging |
| Περίληψη: | Objective.To determine whether simple regularized linear regression models, relying solely on radial distance, mass density and elemental tissue composition as input features, can outperform the conventional water-kernel interpolation baseline for predicting radial dose point kernels (DPKs) ofacross a diverse set of homogeneous tissues and one additional low-density homogeneous medium.Approach.Radial DPKs forwere generated using OpenGATE 10 Monte Carlo simulations in 12 homogeneous media covering soft tissues, bone-like structures, and low-density lung-like tissues, together with one additional low-density homogeneous medium (g cm) evaluated separately as an out-of-distribution extrapolation test. Three scikit-learn models (RidgeCV, LassoCV, and ElasticNetCV) were trained with robust scaling and evaluated through leave-one-tissue-out (LOTO) and leave-one-family-out cross-validation, as well as on the independent low-density case. Performance was assessed using mean absolute percentage error on physical dose (), logarithmic root mean square error (), relative integral energy error, and relative errors onand, systematically compared against a water-kernel baseline obtained by linear resampling of a single simulated water DPK.Main results.ElasticNetCV was the best-performing linear surrogate overall, achieving a meanof 224% in LOTO validation (median 117%), followed by LassoCV and RidgeCV. However, none of the linear surrogates outperformed the water-kernel baseline on the main physical performance metrics. Mean gains relative to the baseline remained negative across tissue families, ranging from aboutpercentage points in bone-like tissues to aboutpercentage points in low-density lung-like tissues. Relative integral energy errors reached several thousand percent in lung-like media, and the separate low-density homogeneous medium also showed severe degradation (best surrogateversusfor the water-kernel baseline). Although some surrogate predictions showed lower log-domain errors than the baseline in selected cases, these improvements did not translate into better dosimetric accuracy in physical dose units.Significance.Simple linear machine-learning surrogates based on radial distance, mass density and macroscopic elemental composition did not improve, and often substantially worsened, the accuracy of radialDPK prediction compared with standard water-kernel interpolation. This negative benchmark delineates the limitations of basic tabular linear regression for capturing density- and composition-dependent kernel changes across a range of homogeneous media relevant to nuclear medicine dosimetry, and provides a reproducible reference framework for future development of more advanced surrogate models. (© 2026 IOP Publishing Ltd. All rights, including for text and data mining, AI training, and similar technologies, are reserved.) |
| Contributed Indexing: | Keywords: 177Lu; Monte Carlo; OpenGATE; dose point kernel; machine-learning surrogate; nuclear medicine dosimetry; tissue composition |
| Entry Date(s): | Date Created: 20260921 Date Completed: 20260921 Latest Revision: 20260924 |
| Update Code: | 20260925 |
| DOI: | 10.1088/2057-1976/aea427 |
| PMID: | 42764806 |
| Βάση Δεδομένων: | MEDLINE |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: cmedm DbLabel: MEDLINE An: 42764806 AccessLevel: 3 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Limitations of simple machine-learning surrogates for cross-tissue prediction of radial177Ludose point kernels. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AU" term="%22Ennassiri+H%22">Ennassiri H</searchLink>; Medical Physics and Radiation Protection Department, Centre Hospitalier Intercommunal Toulon-La Seyne-sur-Mer, Toulon, France. – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22101675002%22">Biomedical physics & engineering express</searchLink> [Biomed Phys Eng Express] 2026 Sep 21; Vol. 12 (5). <i>Date of Electronic Publication: </i>2026 Sep 21. – Name: TypePub Label: Publication Type Group: TypPub Data: Journal Article – Name: Language Label: Language Group: Lang Data: English – Name: TitleSource Label: Journal Info Group: Src Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22IOP+Publishing+Ltd%22">IOP Publishing Ltd </searchLink><i>Country of Publication: </i>England <i>NLM ID: </i>101675002 <i>Publication Model: </i>Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>2057-1976 (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2220571976%22">20571976 </searchLink><i>NLM ISO Abbreviation: </i>Biomed Phys Eng Express <i>Subsets: </i>MEDLINE – Name: PublisherInfo Label: Imprint Name(s) Group: PubInfo Data: <i>Original Publication</i>: Bristol : IOP Publishing Ltd., [2015]- – Name: SubjectMESH Label: MeSH Terms Group: Su Data: <searchLink fieldCode="MM" term="%22Radiotherapy+Planning%2C+Computer-Assisted%22">Radiotherapy Planning, Computer-Assisted*</searchLink>/<searchLink fieldCode="MM" term="%22Radiotherapy+Planning%2C+Computer-Assisted+methods%22">methods</searchLink> <br /><searchLink fieldCode="MM" term="%22Machine+Learning%22">Machine Learning*</searchLink><br /><searchLink fieldCode="MH" term="%22Monte+Carlo+Method%22">Monte Carlo Method</searchLink> ; <searchLink fieldCode="MH" term="%22Humans%22">Humans</searchLink> ; <searchLink fieldCode="MH" term="%22Algorithms%22">Algorithms</searchLink> ; <searchLink fieldCode="MH" term="%22Linear+Models%22">Linear Models</searchLink> ; <searchLink fieldCode="MH" term="%22Computer+Simulation%22">Computer Simulation</searchLink> ; <searchLink fieldCode="MH" term="%22Phantoms%2C+Imaging%22">Phantoms, Imaging</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Objective.To determine whether simple regularized linear regression models, relying solely on radial distance, mass density and elemental tissue composition as input features, can outperform the conventional water-kernel interpolation baseline for predicting radial dose point kernels (DPKs) ofacross a diverse set of homogeneous tissues and one additional low-density homogeneous medium.Approach.Radial DPKs forwere generated using OpenGATE 10 Monte Carlo simulations in 12 homogeneous media covering soft tissues, bone-like structures, and low-density lung-like tissues, together with one additional low-density homogeneous medium (g cm) evaluated separately as an out-of-distribution extrapolation test. Three scikit-learn models (RidgeCV, LassoCV, and ElasticNetCV) were trained with robust scaling and evaluated through leave-one-tissue-out (LOTO) and leave-one-family-out cross-validation, as well as on the independent low-density case. Performance was assessed using mean absolute percentage error on physical dose (), logarithmic root mean square error (), relative integral energy error, and relative errors onand, systematically compared against a water-kernel baseline obtained by linear resampling of a single simulated water DPK.Main results.ElasticNetCV was the best-performing linear surrogate overall, achieving a meanof 224% in LOTO validation (median 117%), followed by LassoCV and RidgeCV. However, none of the linear surrogates outperformed the water-kernel baseline on the main physical performance metrics. Mean gains relative to the baseline remained negative across tissue families, ranging from aboutpercentage points in bone-like tissues to aboutpercentage points in low-density lung-like tissues. Relative integral energy errors reached several thousand percent in lung-like media, and the separate low-density homogeneous medium also showed severe degradation (best surrogateversusfor the water-kernel baseline). Although some surrogate predictions showed lower log-domain errors than the baseline in selected cases, these improvements did not translate into better dosimetric accuracy in physical dose units.Significance.Simple linear machine-learning surrogates based on radial distance, mass density and macroscopic elemental composition did not improve, and often substantially worsened, the accuracy of radialDPK prediction compared with standard water-kernel interpolation. This negative benchmark delineates the limitations of basic tabular linear regression for capturing density- and composition-dependent kernel changes across a range of homogeneous media relevant to nuclear medicine dosimetry, and provides a reproducible reference framework for future development of more advanced surrogate models.<br /> (© 2026 IOP Publishing Ltd. All rights, including for text and data mining, AI training, and similar technologies, are reserved.) – Name: SubjectMinor Label: Contributed Indexing Group: Data: <i>Keywords: </i>177Lu; Monte Carlo; OpenGATE; dose point kernel; machine-learning surrogate; nuclear medicine dosimetry; tissue composition – Name: DateEntry Label: Entry Date(s) Group: Date Data: <i>Date Created: </i>20260921 <i>Date Completed: </i>20260921 <i>Latest Revision: </i>20260924 – Name: DateUpdate Label: Update Code Group: Date Data: 20260925 – Name: DOI Label: DOI Group: ID Data: 10.1088/2057-1976/aea427 – Name: AN Label: PMID Group: ID Data: 42764806 |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=42764806 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1088/2057-1976/aea427 Languages: – Code: eng Text: English Subjects: – SubjectFull: Monte Carlo Method Type: general – SubjectFull: Humans Type: general – SubjectFull: Algorithms Type: general – SubjectFull: Linear Models Type: general – SubjectFull: Computer Simulation Type: general – SubjectFull: Phantoms, Imaging Type: general – SubjectFull: Radiotherapy Planning, Computer-Assisted methods Type: general – SubjectFull: Machine Learning Type: general Titles: – TitleFull: Limitations of simple machine-learning surrogates for cross-tissue prediction of radial177Ludose point kernels. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Ennassiri H IsPartOfRelationships: – BibEntity: Dates: – D: 21 M: 09 Text: 2026 Sep 21 Type: published Y: 2026 Identifiers: – Type: issn-electronic Value: 2057-1976 Numbering: – Type: volume Value: 12 – Type: issue Value: 5 Titles: – TitleFull: Biomedical physics & engineering express Type: main |
| ResultId | 1 |