2 results
An empirical staging model for schizophrenia using machine learning
- M.-C. Clara, F. Sánchez-Lasheras, A. García-Fernández, L. González-Blanco, P. A. Sáiz, J. Bobes, M. P. García-Portilla
-
- Journal:
- European Psychiatry / Volume 66 / Issue S1 / March 2023
- Published online by Cambridge University Press:
- 19 July 2023, pp. S626-S627
-
- Article
-
- You have access Access
- Open access
- Export citation
-
Introduction
One of the great challenges still to be achieved in schizophrenia is the development of a staging model that reflects the progression of the disorder. The previous models suggested have been developed from a theoretical point of view and do not include objective variables such as biomarkers, physical comorbidities, or self-reported subjective variables (Martinez-Cao et al. Transl Psychiatry 2022; 12(1) 1-11).
ObjectivesDevelop a multidimensional staging model for schizophrenia based on empirical data.
MethodsNaturalistic, cross-sectional study. Sample: 212 stable patients with Schizophrenia (F20). Assessments: ad hoc questionnaire (demographic and clinical information); psychopathology: PANSS, CDS, OSQ, CGI-S; functioning: PSP; cognition: MATRICS; laboratory tests: C-Reactive Protein (CRP), IL-1RA, IL-6, Platelets/Lymphocytes (PLR), Neutrophils/Lymphocytes (NLR), and Monocytes/Lymphocytes (MLR) ratios. Statistical analysis: Variables selection was performed with an ad hoc algorithm developed for this research. The referred algorithm makes use of genetic algorithms (GA) to select those variables that show the best performance for the patients classification according to their global CGI-S. The objective function of the GA maximizes the individuals correct classification of a support vector machines (SVM) model that employs as input variables those given by the GA (Díez-Díaz et al. Mathematics 2021; 9(6) 654). Models performance was assessed with the help of 3-fold cross-validation and these process was repeated 10,000 times for each one of the models assessed.
ResultsMean age(SD): 39.5(13.54); men: 63.5%; secondary education: 59.50%. Most patients in our sample had never been married (74.10%), and more than a third received disability benefits due to schizophrenia (37.70%). The mean length of the disease was 11.98(12.02) years. The best SVM model included the following variables: 1)Clinical: number of hospitalizations, positive, negative, depressive symptoms and general psychopathology; 2)Cognition: speed of processing, visual learning and social cognition; 3)Functioning: PSP total score; 4)Biomarkers: PLR, NLR and MLR. This model was executed again 100,000 times applying again 3-fold cross-validation. In 95% of the algorithm executions more than a 53.52% of the patients were classfied in the right CGI-S category. On average the right classification was of 61.93%. About specificity and sensitivity the average values obtained were of 0.85 and 0.64 respectively.
ConclusionsOur staging model is a robust method that appropriately distributes patients according to the severity of the disorder. Highlights the importance of clinical, functional and cognitive factors to classify patients. Finally, the inflammatory parameters PLR, NLR and MLR have also emerged as potential biomarkers for staging schizophrenia.
Disclosure of InterestNone Declared
The Mechanical Properties of Common Interlevel Dielectric Films and Their Influences on Aluminum Interconnect Extrusions
- Fen Chen, Baozhen Li, Timothy D. Sullivan, Clara L. Gonzalez, Christopher D. Muzzy, H. K. Lee, Mark D. Levy, Michael W. Dashicll, James Kolodzcy
-
- Journal:
- MRS Online Proceedings Library Archive / Volume 594 / 1999
- Published online by Cambridge University Press:
- 10 February 2011, 421
- Print publication:
- 1999
-
- Article
- Export citation
-
Knowledge of the mechanical properties of interlevel dielectric films and their impact on sub-micron interconnect reliability is becoming more and more important as critical dimensions in ULSI circuits are scaled down. For example, lateral aluminum (Al) extrusions into spaces between metal lines, which become a more of a concern as the pitches shrink, appear to depend partially on properties of SiO2 underlayers. In this paper, the mechanical properties of several common interlevel dielectric SiO2 films such as undoped silica glass using a silane (SiH4) precursor, undoped silica glass using a tetraethylorthosilicate (TEOS) precursor, phosphosilicate glass (PSG) deposited by plasma-enhanced chemical vapor deposition (PECVD) and borophosphosilicate glass (BPSG) deposited by sub-atmosphere chemical vapor deposition (SACVD) were studied. Among the four common interlevel layers, BPSG exhibits the smallest modulus (E), hardness (H) and the highest the coefficients of thermal expansion (CTE). BPSG again has the lowest as-deposited compressive stress and the lowest local Si-O-Si strain before annealing. Stress interactions between the various SiO2 underlayers and the Al metal film are further investigated. The impact of dielectric elastic properties on interconnect reliability during thermal cycles is proposed.