Manual digitization can give a pathology laboratory a useful digital layer before it invests in whole-slide imaging (WSI). The goal is not to pretend that a few microscope photographs equal a complete digital slide. It is to create a controlled, traceable workflow for capturing representative fields, connecting them to the right case, enabling consultation, and building better data for future automation.
For Indian labs, this approach is especially useful when budgets, bandwidth, staffing, or scanner availability make a full digital pathology rollout impractical. It can support second opinions, multidisciplinary tumour boards, teaching, quality review, and selected research workflows. It should not replace the pathologist’s judgement or the laboratory’s applicable validation and regulatory requirements.
Start with a defined use case
Do not begin by buying a camera. First decide what the images must accomplish. Common starting points include:
- Sending difficult cases to an external consultant.
- Documenting selected regions for reports, tumour boards, or patient communication.
- Creating teaching sets for residents and laboratory staff.
- Reviewing IHC or special-stain findings across branches.
- Building a structured image library for later computer vision work.
Choose one or two specimen categories for a pilot—for example, skin biopsies, cervical smears, or common IHC panels. Define the expected turnaround time (TAT), number of images per case, responsible staff member, and reviewer. A narrow pilot exposes workflow failures faster than attempting to digitize every slide.
If repetitive coordination, renaming, or routing becomes the bottleneck, document it separately from diagnostic work. Principles from custom AI workflows for redundant administrative tasks can later help automate non-clinical steps without allowing an unvalidated system to make diagnostic decisions.
Capture images consistently
A manual setup can use a smartphone adapter, a trinocular microscope camera, or a USB CMOS camera. The equipment matters, but consistency matters more. Record the microscope model, objective, camera, illumination setting, and capture software used in the pilot.
Use a simple capture protocol:
1. Confirm the glass slide and case identity before opening the camera application.
2. Clean the slide, objective, and camera port.
3. Select the lowest magnification that shows the relevant architecture, then capture higher-power views for cellular detail.
4. Capture an overview or low-power image before focusing on hotspots.
5. Adjust focus at the objective being used; do not rely on autofocus alone.
6. Lock exposure and white balance where the software permits it.
7. Capture enough surrounding tissue to preserve context.
8. Review every image immediately at full size and retake blurred, clipped, or poorly exposed fields.
A representative set might include one overview, two to five intermediate-power fields, and targeted high-power images. The number should depend on the specimen and clinical question—not an arbitrary quota. For suspected focal disease, manual photography has a serious limitation: an uncaptured area is invisible to a remote reviewer. Record that limitation clearly when sharing images.
Smartphones are acceptable for documentation and many consultation scenarios, but a dedicated camera generally provides more stable exposure, better ergonomics, and easier capture of repeated cases. Avoid claiming diagnostic-grade performance until the laboratory has validated the complete combination of microscope, optics, camera, monitor, software, and intended use.
Create an identity and metadata standard
The most damaging manual error is attaching a correct-looking image to the wrong patient. Use two identifiers wherever practical: the laboratory accession number or UHID, plus a second check such as patient initials, specimen type, or barcode. Do not place unnecessary personally identifiable information in filenames.
A practical filename might be:
ACC-2026-004812_LiverBiopsy_HE_10x_F03_20260923.jpg
Keep the filename predictable and store richer information in a case record or spreadsheet. Minimum metadata should include:
- Accession number and specimen type.
- Stain, block, and slide number.
- Objective magnification and image sequence.
- Capture date, operator, microscope, and camera.
- Clinical question or region of interest.
- Reviewer, review date, and disposition.
Use controlled vocabulary: HE, not a mixture of H&E, HE stain, and haematoxylin eosin. A short data dictionary prevents inconsistent labels and makes later analysis possible. Never use a patient’s name as the primary filename, and avoid storing case images on personal phones after upload.
Store, back up, and control access
A shared folder is not a complete image-management system. For a small laboratory, a local network-attached storage (NAS) device can provide controlled access and predictable performance. Cloud storage may be more convenient for distributed teams, but assess vendor contracts, data residency, access logs, encryption, backup policy, and connectivity before uploading identifiable health information.
Use a case-level folder structure such as:
Accession_ID/01_RawAccession_ID/02_ReviewedAccession_ID/03_AnnotationsAccession_ID/04_Consultation
Keep raw files immutable. Save corrected or annotated copies separately, with the operator and date recorded. Apply role-based access: technicians may upload, pathologists may review, and administrators should not automatically see clinical images. Enable multi-factor authentication for cloud accounts and maintain an access log.
A sensible backup pattern is three copies, on two types of media, with one copy offline or geographically separate. Test restoration—not merely backup completion—at a defined interval. The lab should also have a retention and deletion policy aligned with its clinical, contractual, and legal obligations. For broader governance, review the principles behind secure autonomous AI workflows, particularly least privilege, auditability, and human approval gates.
Link images to the LIS and reporting process
The digital image should be reachable from the case record, not buried in an individual employee’s desktop. If the LIS supports attachments or hyperlinks, link to the controlled case folder rather than copying files repeatedly. If it does not, maintain a restricted accession-to-image register with a clearly assigned owner.
Barcodes or QR codes can reduce transcription errors, but they do not solve identity by themselves. Scan the accession barcode at the microscope station, verify the patient and specimen details on screen, and require a second check before final upload. A simple upload checklist is often safer than an elaborate integration that staff do not use.
For telepathology, share a time-limited, access-controlled link where possible. Include the clinical question, stain, magnification, number of fields captured, and a statement that the images are selected fields rather than a complete WSI. Retain the consultation response in the case record.
Quality assurance and validation
Set acceptance criteria before the pilot begins. At minimum, assess:
- Correct case-to-image matching.
- Adequate focus and exposure.
- Accurate colour reproduction for H&E, IHC, and special stains.
- Sufficient context around each region of interest.
- Upload completion and retrievability.
- Time from slide selection to consultant access.
- Rate of retakes, misfiles, and missing metadata.
Calibrate the display and camera setup, standardise illumination, and photograph a reference slide periodically. For IHC, do not rely on a smartphone screen or uncontrolled monitor to make subtle intensity judgments. If images will influence diagnosis, conduct a documented validation with representative cases, defined users, and an approved review process. Manual images are often excellent for consultation and education, but they may be insufficient for primary diagnosis or exhaustive tumour search.
Build toward AI without rushing into it
A manually captured image library becomes valuable only when labels are reliable. Store the accession, stain, region, diagnosis or finding, annotator, and uncertainty status separately from the image. Obtain appropriate permissions and de-identify data before research sharing. Do not train a model on inconsistent filenames, duplicated fields, or labels copied without pathologist review.
When the pilot is stable, automate low-risk actions first: barcode-based folder creation, metadata validation, duplicate detection, upload notifications, and consultation reminders. These are good candidates for cost-effective AI operational workflows for founders, while diagnostic classification should remain subject to clinical validation and human oversight. A later move to WSI or AI-assisted analysis will be easier because the lab already has defined identifiers, quality checks, and governance.
A practical 30-day pilot
Week 1: Select one specimen type, document the clinical use case, choose equipment, and approve the naming and access policy.
Week 2: Train two operators and capture a small set of cases. Measure retakes, upload time, and identity errors.
Week 3: Add pathologist review, remote consultation, and backup restoration testing. Fix the highest-frequency failure points.
Week 4: Compare TAT and consultation quality with the existing process. Decide whether to expand, change equipment, or stop.
Manual digitization is worthwhile when it improves access and traceability without creating false confidence. Start with a narrow, validated workflow; preserve the glass slide and diagnostic context; and treat every image as a clinical data asset that needs identity, quality control, security, and a clear owner.