We investigate parasite diversity and host associations through comparative morphology, preserved specimens, and reproducible visual evidence.
Detailed microscopy makes parasite structures visible and gives comparative research something concrete to examine.
Explore how we document morphology →
To understand parasite diversity and its place in changing host communities.
SEPS is a nonprofit research organization focused on parasite systematics and disease ecology. Our work connects field observations with comparative morphology, taxonomy, and preserved specimens to investigate neglected and understudied parasites. Computer vision supports this research by organizing images and measurements for expert review.
Building on this foundation, we aim to advance parasite systematics and translate that knowledge into affordable, accessible imaging tools that strengthen diagnostic capacity. Our long-term goal is to develop and validate approaches suited to settings with limited laboratory infrastructure, including those equipped with a smartphone and a microscope, to support reliable parasite identification and efforts to combat parasitic neglected tropical diseases worldwide.
Florida nonprofit organization · U.S. 501(c)(3)
Which parasites occur in sampled hosts, and which anatomical characters support their identification?
How do parasite records vary among hosts and collection sites? Observed associations provide a starting point for testing transmission and host-switching hypotheses.
Do conventional measurements, geometric morphometrics, and computer vision add reliable identification value when compared with expert morphological review?
Parasite Recognition through Integrated Specimen Morphology
FieldView, TEGUMENT, HookFuse, and FUZZARIA connect specimen images, diagnostic measurements, evidence quality, and expert review.
Explore the SEPS:PRISM workflowThe images, diagnostic structures, and documentation behind our research.
Three RO views connect the original capture, proposed anatomical labels, and expert judgment.
Field-to-voucher research infrastructure
Comparative morphology and curated visual evidence
Reproducible visual evidence through the SEPS:PRISM methodology
Measurable outputs
Preserved vouchers Traceable datasets Diagnostic image atlases Peer-reviewed records GIS/data products Agency-ready implementation tools
Field → image → evidence.
Field validation, voucher generation, reproducible data products and agency-ready outputs.
Parasite Recognition through Integrated Specimen Morphology
SEPS:PRISM is our integrated methodology for specimen imaging, morphological measurement, and evidence assessment. FieldView supports acquisition, TEGUMENT prepares image evidence, HookFuse examines diagnostic morphology, and FUZZARIA assesses uncertainty and reporting limits. Taxonomic interpretation remains grounded in diagnostic anatomy, reference literature, and expert review.
Each record connects the available images to the host, collection event, preservation history, and voucher. Missing information remains visible. Molecular evidence can be linked when it is available and appropriate to the question.
This page describes the SEPS:PRISM methodology and its development requirements. Prototype image-processing changes, proposed robustness pilots, and biological validation are distinct. The workflow diagram does not establish validated automated identification or measured end-to-end performance.
Explore one specimen photograph, its provisional mask, and a mathematical description of the extracted boundary.
A close view documents the structures in an image. Outlines and landmark placements show different ways to record shape for comparison.
Illustrative examples. Anatomical labels and landmark correspondence require expert review. The additional scale bars displayed in Figure 3 have not been independently verified. View the original annotated SEM plate ↗
FieldView and laboratory uploads enter the same specimen-linked workflow. TEGUMENT is the first processing phase within SEPS:PRISM.
Original field photographs, host context, specimen identity, and capture quality.
Existing microscopy images, specimen metadata, and calibration where available.
Egg microscopy and adult-worm or hook analysis use separate profiles. Unsupported inputs may remain unresolved.
Segment the specimen and prepare linked image evidence. Review boundaries and retain the original image.
Review diagnostic structures, outlines, landmarks, and measurements. Track uncertainty in the contour and its measurements.
Use a supported task model or reference comparison when available. The egg profile bypasses adult-worm and hook extraction.
Assess evidence quality, diagnostic coverage, calibration, provenance, and disagreement. A high model score cannot override missing diagnostic evidence.
Inspect the evidence and record a supported conclusion, a correction request, or an unresolved result.
Keep the source image, derived views, measurements, quality flags, and review history connected in the specimen record.
Correction path: a reviewer can request a new capture, a corrected mask, or a revised outline. Automated feedback and reprocessing remain proposed work.
These requirements guide development and validation. They describe what must be demonstrated before a feature or identification claim is released.
Retain originals and failed acquisitions, separate capture failure from no detection, and keep outputs linked to the current image. Offline recovery and delayed or stale outputs require device testing.
Reported whole-worm prototype changes separate specimen outlines and confine internal contrast patches to their selected outlines. Boundary preservation and thin-structure continuity need anatomical review; contrast patches are not confirmed anatomy.
Preserve fine attachment structures, review contour uncertainty, and carry that uncertainty into measurements. Shape comparison requires comparable views and reviewed traces. Retrieved look-alike images support expert inspection rather than establish identity.
Keep quality dimensions and contradictory evidence visible. Assess boundary and anatomical errors separately from mask overlap. Unsupported inputs can be rejected; an unresolved outcome is acceptable. Restored or generated anatomical detail is inadmissible identification evidence.
Spatial coverage for selecting review examples, constrained thin-structure tracing, measurement-uncertainty propagation, and FieldView response-time checks remain proposed pilots until tested on relevant SEPS specimens or devices. The methodology does not treat those pilots as completed validation.
Diagnostic anatomy is compared with taxonomic descriptions, keys, and appropriate reference specimens. Calibrated lengths, widths, angles, and ratios are recorded where the structures and views support them. Hook geometry and other attachment structures are a focused application within HookFuse.
Measurements retain their units, calibration source, view, and reviewer. Missing calibration prevents claims about absolute physical size. Incomplete anatomy or conflicting metadata limits the taxonomic conclusion.
The proposed extension uses anatomically corresponding landmarks and reviewed semilandmarks to compare shape. Generalized Procrustes alignment removes differences in position, rotation, and overall scale; size is retained separately so allometry can be investigated.
Principal component analysis can describe shape variation without establishing species identity. Elliptic Fourier analysis is an optional branch for suitable closed outlines. View, orientation, preservation, and tracing error must be controlled before comparisons are interpreted.
TEGUMENT organizes conservative image preparation, segmentation, and diagnostic-region review. Masks must preserve the anatomy needed for identification and measurement. Original photographs remain available alongside every processed derivative.
Vision Transformers are a proposed computer-vision extension for learning image features and testing candidate classification or reference-image comparison. Attention maps alone do not establish an anatomical explanation. Model version, reference data, preprocessing, and review status accompany any experimental output.
FUZZARIA organizes evidence adequacy and uncertainty to prioritize review. The proposed integration retains separate assessments for image quality, diagnostic coverage, calibration, and provenance, with explicit notes when analyses disagree.
Insufficient evidence can lead to recapture, further examination, or an unresolved record. No-detection images and inadequate acquisitions are retained with distinct outcomes. A model detecting no target is not, by itself, evidence that a host is uninfected.
The pilot compares conventional morphology and measurements with the added value of geometric morphometrics, Vision Transformers, and their combination. Expert-reviewed references are required before performance is reported.
Repeated images of a specimen remain in one data partition. Host and collection grouping are respected where required by the study design.
Detection, measurement error, shape repeatability, and taxonomic classification are evaluated separately. Unresolved cases remain visible.
SEPS:PRISM names the complete methodology. FieldView documents acquisition, TEGUMENT prepares visual evidence, HookFuse organizes diagnostic characters, and FUZZARIA supports review. Lens is the separate mobile image-readiness interface, not an extra processing phase.
Image preview only. No record is saved and no file is uploaded. Use the documentation template to prepare a linked record.
Original views, diagnostic details, and linked documentation.
These are image entries. Physical specimen identity, host association, and voucher linkage require verification. R-001, R-006, and R-017 now open structured draft records; their missing fields are shown explicitly.
A formal SEPS publication index will appear here as papers, natural history notes, technical reports and conference materials are released.
About SEPS
Society for Evolutionary Parasite Systematics
SEPS is a U.S. 501(c)(3) nonprofit research organization focused on parasite systematics and disease ecology. Comparative morphology, taxonomy, ecology, and preserved specimens guide our work on neglected and understudied parasites. Computer vision and quantitative shape analysis support these scientific questions.
Building on this foundation, we aim to advance parasite systematics and translate that knowledge into affordable, accessible imaging tools that strengthen diagnostic capacity. Our long-term goal is to develop and validate approaches suited to settings with limited laboratory infrastructure, including those equipped with a smartphone and a microscope, to support reliable parasite identification and efforts to combat parasitic neglected tropical diseases worldwide.
For research partnerships, specialist review, and questions about specimens or images, contact M.R. Dwyer, Director of Research & Development.
Email SEPSSociety for Evolutionary Parasite Systematics
