DSM
The Diagnostic and Statistical Manual of Mental Disorders (DSM) is a handbook authored by the American Psychiatric Association that classifies mental disorders through operationalized symptom criteria, aiming to standardize diagnoses for clinical, research, and administrative purposes without relying on unproven etiological theories.[1] First published in 1952 as a concise guide modeled on U.S. military classifications from World War II, it sought to address inconsistent diagnostic practices by emphasizing observable behaviors and excluding speculative causes like Freudian dynamics.[2] Subsequent editions expanded the manual's scope, incorporating feedback from field trials and epidemiological data; DSM-III (1980) marked a paradigm shift toward atheoretical, reliability-focused criteria, while DSM-5 (2013) and its 2022 text revision (DSM-5-TR) introduced dimensional assessments for some conditions and harmonized elements with the World Health Organization's ICD system to facilitate global use.[1] This evolution has enabled consistent coding for insurance reimbursements, structured research protocols, and cross-study comparisons, contributing to advances in psychopharmacology and therapy outcomes tracking.[3] Nevertheless, the DSM's categorical approach has drawn empirical scrutiny for limited inter-rater reliability—often yielding kappa values below 0.5 for many disorders in field trials—and validity concerns, as diagnoses frequently fail to align with distinct neurobiological profiles or longitudinal outcomes, potentially inflating prevalence through lowered thresholds and symptom overlap.[4][5] Critics highlight risks of false positives, where normal variations or transient distress are pathologized, exacerbating over-medicalization without corresponding causal insights; DSM-5's revisions, in particular, amplified these debates by broadening criteria for conditions like autism and ADHD, amid questions of undue influence from stakeholder interests over rigorous falsifiability.[6][7] Despite such challenges, the manual remains a pragmatic tool, though ongoing calls urge integration of genetic, neuroimaging, and causal modeling data to enhance scientific grounding.[8]Psychiatry and mental health
Diagnostic and Statistical Manual of Mental Disorders
The Diagnostic and Statistical Manual of Mental Disorders (DSM) is a classification system for mental disorders published by the American Psychiatric Association (APA), containing diagnostic criteria, descriptive text, and codes to standardize clinical assessments in the United States and internationally.[9] First issued in 1952, it evolved from earlier statistical manuals focused on census data for hospitalized patients to a comprehensive tool emphasizing observable symptoms over inferred causes, aiming to enhance diagnostic reliability across practitioners.[10] [11] The current edition, DSM-5-TR (text revision), released on March 18, 2022, includes updates to criteria, new disorders like prolonged grief disorder, and revised text reflecting post-2013 research, while retaining the core structure of over 200 diagnostic categories.[9] [12] Development of DSM editions involves APA-appointed workgroups of experts reviewing empirical literature, conducting field trials for inter-rater reliability, and incorporating public feedback, as seen in the decade-long process for DSM-5 starting in 2000 with NIMH collaboration.[11] Early editions like DSM-I and DSM-II (1968) drew from psychoanalytic theory and World Health Organization influences, listing 106 and 182 disorders respectively with brief descriptions lacking explicit criteria.[11] The paradigm shift occurred with DSM-III in 1980, led by Robert Spitzer, adopting a descriptive, atheoretical approach with polythetic criteria sets (requiring a subset of symptoms from a list) to prioritize reliability over etiology, resulting in 265 disorders and field trial kappas averaging 0.60-0.80 for many categories.[11] Subsequent revisions—DSM-III-R (1987), DSM-IV (1994), and DSM-IV-TR (2000)—refined criteria via literature reviews and trials but avoided major restructuring until DSM-5, which eliminated the multiaxial system, introduced dimensional assessments for some disorders, and added conditions like binge eating disorder based on prevalence data exceeding 3% in community samples.[11] [12] The manual's structure comprises a diagnostic classification (alphabetical and numerical codes aligned with ICD), criteria sets specifying symptom thresholds (e.g., five or more symptoms for major depressive episode persisting two weeks), and textual elaborations on prevalence, course, and differentials, without prescribing treatments.[13] It facilitates billing, research, and epidemiology, with U.S. lifetime prevalence estimates derived from it indicating 20-25% for anxiety disorders and 6-7% for major depression in national surveys.[14] However, empirical critiques highlight persistent limitations: inter-rater reliability in DSM-5 field trials ranged from kappa 0.20 for complex diagnoses like major depressive disorder to 0.80 for schizophrenia, often falling below the 0.70 threshold for clinical utility, undermining consistency.[4] Validity concerns persist, as categories lack robust biological validators like biomarkers or family study boundaries, with critics arguing the system conflates heterogeneous syndromes and prioritizes consensus over causal mechanisms, potentially inflating prevalence by lowering thresholds (e.g., DSM-5 autism spectrum disorder encompassing prior subtypes).[15] [16] Pharmaceutical industry funding of trials and task force members has raised questions of bias toward expanding diagnoses amenable to medication, though APA maintains transparency via conflict disclosures.[17] Cultural applicability is limited, with Western-centric criteria overlooking somatic presentations common in non-Western populations, prompting calls for alternatives like the Research Domain Criteria (RDoC) framework emphasizing neural circuits over symptom clusters.[8] Despite these issues, DSM remains the de facto standard for insurance reimbursement and legal determinations, influencing global practice via harmonization with ICD-11.[9]Business and organizations
DSM-Firmenich
DSM-Firmenich AG is a Swiss-Dutch multinational corporation specializing in science-based innovation for nutrition, health, and beauty markets. Formed through a merger of equals between Royal DSM N.V., a Dutch company founded in 1902 originally focused on coal and fertilizers before pivoting to chemicals, nutrition, and materials, and Firmenich SA, a Swiss firm established in 1895 as a leader in flavors and fragrances, the combined entity leverages complementary expertise in sustainable ingredients, perfumery, taste enhancement, and nutritional solutions.[18][19] The merger was announced on May 31, 2022, and completed on May 9, 2023, with Firmenich becoming a wholly owned subsidiary of the new holding company, DSM-Firmenich AG, listed on the Euronext Amsterdam and SIX Swiss Exchange. At inception, DSM shareholders held approximately 65.5% of the combined entity, reflecting the transaction's structure to integrate operations while preserving innovation-driven cultures from both predecessors. The company maintains dual headquarters in Kaiseraugst, Switzerland (including a new R&D facility opened in 2023), and Maastricht, Netherlands (with a new site in 2024), alongside operations in nearly 60 countries.[20][21][22] DSM-Firmenich's business segments encompass perfumery and beauty, taste, texture, and health, health, nutrition, and care, and animal nutrition and health, emphasizing purpose-led science to address challenges like sustainable protein production and climate impacts on food systems. In 2024, the company reported annual sales of €12.8 billion, organic sales growth of 6%, and employed around 30,000 people globally, supported by over 2,000 R&D professionals. Sales synergies contributed approximately €50 million to 2024 revenues, with expectations for an additional €100 million in 2025, underscoring post-merger integration focused on accelerated innovation and sustainability.[23][24][25]Other organizations
The Leibniz Institute DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH) is a German non-profit research institute and culture collection specializing in microorganisms, cell lines, and bioresources. Established in 1969, it maintains over 92,500 strains, including bacteria, archaea, fungi, protists, bacteriophages, and human/animal cell lines, serving as a key global repository for biodiversity research and industrial applications.[26] The institute, part of the Leibniz Association, employs around 230 staff, with research focused on prokaryotic and eukaryotic biodiversity, authentication services, and bioinformatic tools; it operates under rigorous quality standards certified by ISO and other accreditations.[27] The Department of Standards Malaysia (DSM) functions as Malaysia's National Standards Body (NSB) and National Accreditation Body (NAB), established under the Standards of Malaysia Act 1996 to promote standardization, conformity assessment, and accreditation for enhancing global competitiveness.[28] It develops and adopts over 7,000 Malaysian Standards (MS) across sectors like manufacturing, services, and sustainability, while accrediting laboratories, certification bodies, and inspection entities through programs such as SAMM (for testing labs) and ACB (for certification).[29] DSM collaborates internationally with bodies like ISO and IEC, supporting Malaysia's trade and regulatory frameworks as of 2023.[28] DSM Capital Partners is a U.S.-based, employee-owned investment management firm founded in 1999, specializing in growth equity strategies with a valuation discipline to achieve long-term capital appreciation.[30] Headquartered in Palm Beach Gardens, Florida, it manages portfolios for institutional clients including pension plans, foundations, and endowments, employing concentrated high-conviction approaches across U.S. large-cap growth, global growth, and small-cap strategies; as of recent filings, it oversees assets under management exceeding $10 billion.[31] The firm emphasizes fundamental analysis of quality companies with sustainable revenue growth, registered as an SEC investment adviser since 2001.[32]Science and technology
Digital surface model
A digital surface model (DSM) is a raster-based dataset representing the elevation of the uppermost surface of the Earth, encompassing both terrain and overlying features such as vegetation, buildings, and infrastructure.[33] [34] Unlike bare-earth models, DSMs capture the "first-return" elevations from remote sensing data, providing a continuous 3D surface suitable for visualizing surface obstructions.[35] DSMs differ from digital terrain models (DTMs), which depict only the ground surface after filtering out non-terrain elements like trees and structures to represent bare earth.[33] [36] The term "digital elevation model" (DEM) is sometimes used broadly but often aligns with DTM characteristics, emphasizing topographic relief without artificial or vegetative cover.[33] [34] This distinction is critical in geospatial analysis, as DSMs include canopy heights and urban features, while DTMs/DEMs facilitate hydrological modeling and slope calculations.[36] DSMs are generated primarily through LiDAR surveys, where airborne or terrestrial lasers emit pulses and record the initial reflections from the highest surfaces, yielding point clouds interpolated into grids with resolutions as fine as 1 meter.[35] [37] Photogrammetry offers an alternative, deriving elevations from overlapping stereo aerial or satellite images via structure-from-motion algorithms, though it may introduce errors in low-texture areas compared to LiDAR's direct ranging.[37] [38] Data processing involves georeferencing, classification, and interpolation using software like ArcGIS or LAStools to produce georeferenced TIFF or ASCII grid formats.[35] Applications of DSMs span urban planning for volume calculations of built environments, telecommunications for line-of-sight analysis in antenna placement, and aviation for obstacle detection in approach paths.[33] [36] In environmental monitoring, they support canopy height modeling by subtracting DTM elevations and aid in flood risk assessment by simulating water surface interactions with structures.[34] [39] The integration of DSMs with GIS enables 3D visualizations and simulations, such as urban growth projections, though accuracy depends on source data density—LiDAR-derived models often achieve vertical errors below 15 cm in open terrain.[33] The conceptual foundation for DSMs traces to early digital terrain modeling in the 1950s, pioneered by Charles L. Miller at MIT around 1955, initially focusing on bare-earth representations but evolving with remote sensing advancements.[40] Widespread DSM production accelerated in the early 2000s with commercial LiDAR adoption, enabling inclusion of surface objects for comprehensive topographic datasets.[41]Direct Stream Digital
Direct Stream Digital (DSD) is a proprietary digital audio encoding format developed by Sony and Philips, utilizing 1-bit delta-sigma modulation to represent audio signals at a sampling rate of 2.8224 MHz, or 64 times the standard compact disc rate of 44.1 kHz.[42] This approach enables a frequency response extending up to 100 kHz and dynamic range exceeding 120 dB within the audible band through noise shaping, which shifts quantization noise to ultrasonic frequencies beyond human hearing.[42] Unlike pulse-code modulation (PCM), which quantizes amplitude into multi-bit samples, DSD maintains a single-bit stream of pulses representing the audio waveform's density, aiming to preserve analog-like fidelity by minimizing multi-bit processing artifacts.[43] Sony and Philips initiated DSD development in the mid-1990s as the core technology for Super Audio CD (SACD), a high-resolution optical disc format introduced commercially in 1999 to supersede the compact disc.[44] The format draws from earlier delta-sigma principles patented in 1954 but was refined for consumer audio, with SACD production leveraging DSD to encode stereo and multichannel content on a single layer.[45] Initial mastering involved analog-to-digital conversion directly to DSD streams using specialized converters, though subsequent editing often required conversion to higher-rate PCM due to DSD's computational demands for manipulation.[46] Technically, DSD employs fifth- or higher-order delta-sigma modulators to achieve its performance, resulting in a bit rate of approximately 5.6446 Mbps per channel—over 10 times that of uncompressed CD audio.[42] Noise shaping ensures low in-band noise (below 20-50 kHz) but introduces rising noise above 100 kHz, necessitating careful analog filtering in playback to avoid intermodulation distortion.[47] Extensions include DSD128 (double rate, 5.6448 MHz) and higher multiples like DSD256, adopted in native digital downloads and professional workflows since the early 2010s, though these increase storage and processing burdens.[48] Comparisons to PCM highlight trade-offs: DSD proponents argue it reduces quantization errors and non-linearities inherent in multi-bit PCM, potentially yielding a more "direct" analog emulation, but empirical blind listening tests indicate high-resolution PCM (e.g., 24-bit/192 kHz) and DSD are often indistinguishable to listeners.[46][49] DSD's disadvantages include inefficient editing (requiring delta-sigma remodulation), higher bandwidth demands, and vulnerability to noise floor elevation during processing, leading many studios to use PCM intermediaries despite native DSD capture advantages in converters.[46] Despite limited mainstream adoption—SACD sales peaked modestly in the 2000s—DSD persists in audiophile circles for downloads and hybrid discs, with ongoing debate over audible benefits unsubstantiated by large-scale perceptual studies.[50][51]Dynamic shared memory
Dynamic shared memory in CUDA refers to a mechanism for allocating on-chip shared memory to thread blocks at kernel launch time, rather than fixing the size at compile time.[52] This approach uses the declarationextern __shared__ followed by the data type, such as extern __shared__ int s_data[];, which defers sizing until runtime.[53] During kernel invocation, the third parameter specifies the bytes per block, for example, myKernel<<<blocks, threads, sharedMemSize>>>(args);, where sharedMemSize is calculated based on input parameters or data needs.[52]
This dynamic allocation contrasts with static shared memory, declared as __shared__ int s_data[fixed_size];, which requires a compile-time constant and limits flexibility for varying workloads.[53] Dynamic allocation enables kernels to adapt memory usage to runtime conditions, such as partitioning data irregularly or handling inputs of unknown size beforehand, thereby optimizing performance in parallel algorithms like matrix multiplication or reduction operations.[54]
Performance implications include potential overhead from variable sizing but gains in occupancy and efficiency for memory-bound tasks; however, the total shared memory per multiprocessor remains hardware-limited, typically 48 KB for static plus dynamic on older architectures, extensible to nearly the full multiprocessor limit (e.g., 64 KB or more) via dynamic methods on newer GPUs like those in the Volta or Ampere families.[55] Access within the kernel treats the pointer as a base address, with threads computing offsets manually to avoid bank conflicts and ensure coalesced loads from global memory.[54] Synchronization via __syncthreads() is required for safe inter-thread communication, as with static variants.[52]
Introduced in early CUDA versions to enhance programmability, dynamic shared memory supports advanced features like cooperative groups in CUDA 9.0 (2017) and beyond, allowing finer control in multi-block collaborations.[52] Limitations include a single dynamic allocation pointer per kernel—subsequent arrays must offset from it—and no support for recursive or templated sizing without host-side computation.[54] In practice, developers profile with tools like Nsight Compute to balance dynamic usage against launch configuration trade-offs, as excessive per-block allocation reduces achievable occupancy.[53]