How to Use Anki for Anatomy and Image Occlusion

Disclosure: I build AnkiQuiz, the card-based quiz tool discussed later in this guide. AnkiQuiz is not affiliated with Anki, AnKing, AnkiHub, UWorld, AMBOSS, or the NBME. It does not replace an anatomy practical, cadaver lab, or validated question bank, and AI-generated questions can contain errors.

Anatomy seems perfectly matched to Anki Image Occlusion: cover a label, name the structure, repeat later. The method is useful, but a high review accuracy can hide a narrow memory. You may know that the hidden rectangle in the upper-left corner is the radial nerve while failing to find the nerve on a cadaver, an axial CT, or the same illustration rotated to a new view.

The solution is not to stop using images. It is to give image identification a defined role inside a broader anatomy workflow. Start from course objectives, use images you are allowed to use, create a limited number of focused occlusions, preserve orientation and relational context, and then test the same anatomy from unfamiliar views and different prompts.

This guide is written for U.S. medical students, but it does not prescribe one universal deck, Image Occlusion mode, or daily card count. Your course objectives, practical format, school policies, image licenses, prior knowledge, and available study time should determine the details.

The broader medical-school Anki workflow shows how this fits around lectures, due reviews, and question blocks.

Why anatomy cards can feel easier than anatomy

An image card does not test “anatomy” as a single ability. It tests whether you can retrieve one answer from the exact visual information left on the front. That can be valuable, but it can also become much narrower than the task your faculty will ask you to perform.

You learn the rectangle's location

After enough repetitions, the shape and position of an occlusion become a cue. You may answer because “the long label near the top is always the median nerve,” even when you cannot trace the nerve or name its neighboring artery. This is a version of the recognition gap : a familiar card presentation supports the answer, but the knowledge is less accessible when the presentation changes.

Do not conclude that all position memory is useless. Anatomy is spatial, and location should matter. The problem is learning a label's location within one card layout instead of learning a structure's location relative to stable landmarks.

A clean atlas does not look like lab

A color-coded atlas removes fat, fascia, tissue distortion, imaging artifacts, pathology, and normal variation. A prosection or cadaver may expose a different layer. A CT reverses some intuitive expectations about viewpoint and replaces colored structures with density. A new photograph may rotate the field or crop away the landmark you depended on.

If every review uses one polished superior view, success mainly establishes retrieval from that view. It does not establish transfer to an inferior view, a specimen, a cross-section, or radiology.

You remember the name but not the anatomy around it

Naming the femoral nerve is not the same as knowing its position relative to the femoral artery, the compartment it enters, its major motor and sensory functions, or the deficit expected after injury. Identification is one useful layer. Relationships, function, and clinical meaning require their own retrieval opportunities.

One image silently creates a review burden

It is easy to draw twenty masks on one diagram and create twenty cards in minutes. Those cards are not free. Each one can return repeatedly for months or years. A few heavily labeled images can produce enough new cards to crowd out lab preparation and practice questions.

Judge Image Occlusion by the future workload it creates, not by how quickly masks can be drawn. If the queue is already growing, use the cards-per-day framework before adding another anatomy batch. If reviews are overdue, stabilize them with the medical-school backlog plan .

Step 1: define the course target before making cards

Open the syllabus, laboratory objectives, faculty image list, and practical instructions. For one region, classify the required outcomes:

• identification: name a tagged structure on a specimen, model, image, or scan;
• orientation: identify side, plane, view, level, or compartment;
• relationship: state what lies anterior, posterior, medial, lateral, superficial, deep, or within a space;
• pathway: trace a nerve, vessel, duct, or fascial route;
• function: retrieve action, innervation, blood supply, drainage, or physiologic role;
• clinical application: predict a deficit, localize a lesion, or connect anatomy to a procedure;
• radiology: identify structures across slices, modalities, and contrast phases.

Then ask which outcomes genuinely need spaced flashcards. A practical dominated by tagged specimens justifies identification cards, but still requires time with specimens. A course emphasizing cross-sectional anatomy needs serial CT or MRI practice, not only atlas plates. A broad “know the brachial plexus” objective should be broken into the specific relationships and operations your assessment expects.

This boundary protects you from turning every visible label into a scheduled obligation. Make cards for durable, testable targets. Keep reference-level detail in an atlas rather than forcing all of it into active recall.

Step 2: choose images you may use and can actually read

Use your own drawings or photographs, openly licensed material, school-provided images when the terms permit personal study use, or commercial resources whose license allows your intended use. A visible image online is not automatically licensed for copying, modification, uploading, or AI processing.

For clinical images, remove patient identifiers and follow your institution's privacy, security, and educational-use policies. De-identification is more than cropping out a name if dates, accession numbers, facial features, metadata, or rare clinical context could identify a patient. When in doubt, use an approved teaching image or ask the responsible faculty member.

Do not upload protected question-bank figures, proprietary atlas plates, patient information, or other material you lack the right to process. “Educational use” is not a universal permission that overrides a license, privacy rule, or school policy.

Quality also matters. Prefer an image with:

• enough resolution to distinguish the target;
• a known side, plane, and viewpoint;
• landmarks that remain visible when the label is hidden;
• contrast appropriate to the structure;
• a source and caption you can record;
• no unnecessary labels that reveal sibling answers.

One clear image can support learning. One ambiguous image can train confident guessing.

Step 3: use Anki's native Image Occlusion deliberately

Anki 23.10 and later includes native Image Occlusion. The official Image Occlusion documentation explains how to add an image, draw shapes, select the occlusion behavior, and use fields including Header and Back Extra.

Draw each mask tightly enough to hide the answer without covering the anatomical evidence needed to locate it. If the arrowhead itself points to the structure, preserve the arrow while masking the text. If a leader line crosses several structures ambiguously, use a better image or edit the annotation before creating cards.

Preview the generated cards. Ask whether a student who understood the anatomy—but had never seen that exact plate—could determine what is being requested. A tiny blank rectangle floating among labels may test label geometry more than anatomy.

Hide All, Guess One

In Hide All, Guess One, the target and the other masks are hidden while you answer one target. This can reduce answer leakage from neighboring labels and is often useful when the printed labels would directly reveal the region or sequence.

The tradeoff is visual clutter and lost context. A field of identical boxes can make the target depend on mask position. If hiding every label removes landmarks needed to interpret the view, the card may become artificial or ambiguous.

Hide One, Guess One

In Hide One, Guess One, only the target is hidden and the other labels remain visible. Those labels can provide useful orientation, especially early in learning a dense region. They can also become answer cues: seeing “ulnar nerve” and “brachial artery” may narrow the missing label without requiring you to identify the underlying structure.

Use this mode when the remaining labels supply legitimate context rather than giving away the answer. Test it by mentally covering the other label text. If your answer disappears, the card depends too heavily on exposed labels.

No mode is universally optimal

Choose mode per image and learning stage. Hide All may be better for reducing direct label hints; Hide One may be better when orientation would otherwise vanish. You can also use an unlabeled duplicate, crop the image around meaningful landmarks, or create a separate orientation card.

Do not create both modes for every mask by default. Duplicate cards increase workload without necessarily creating a useful new retrieval route.

Step 4: write context into Header and Back Extra

An occluded image needs enough text to identify the task and preserve meaning. Use Header, Back Extra, or other fields supported by your note type. Field choices should fit your deck's update and syncing rules, especially if you use a maintained premade deck.

For each card, consider recording:

• named region and laterality;
• view or imaging plane;
• superior/inferior or anterior/posterior orientation;
• one stable landmark;
• the answer at the specificity your course expects;
• one high-value relationship;
• one function, pathway, or clinical implication;
• source and permission or license information when appropriate.

Do not overload every back with a textbook chapter. The tested answer should remain clear. Supporting text should make the card interpretable, verifiable, and reusable from another prompt.

Step 5: build a mixed anatomy retrieval set

Image Occlusion is strongest when it is one layer rather than the whole deck. For a topic such as the inguinal canal, combine several operations.

Image identification

Show an appropriate image and identify the masked structure. This is the direct use case for Image Occlusion. Vary the source and view when the curriculum requires it.

Relationship retrieval

Ask a focused question such as, “Which structure forms the anterior wall of the inguinal canal?” or “What lies immediately lateral to this vessel at the marked level?” Relationship prompts should use explicit orientation and enough context to allow one answer.

Function and pathway retrieval

Ask what the identified nerve supplies, what movement a muscle produces, where a vessel originates, or which nodes receive drainage. When a pathway matters, retrieve a short ordered route from a blank prompt or redraw it rather than making a card for every line segment.

Clinical and radiology retrieval

Use a permitted clinical image or text prompt to identify a lesion level, predict a deficit, or find a structure in a cross-section. Keep the task within taught anatomy. Do not invent patient context merely to make a card appear “board style.”

A balanced set does not require four cards for every structure. Choose the smallest prompts that collectively match your objectives. Identification-heavy practicals may need more image cards; clinically integrated blocks may need more relationship and consequence retrieval.

Step 6: control the number of cards created from each image

Before clicking save, rank the masks:

1. Must identify independently for the practical or durable clinical foundation.
2. Must understand relationally, but may be better tested by a text or comparison prompt.
3. Useful landmark, which can remain visible as context.
4. Reference detail, which does not deserve recurring reviews.

Create cards for the first category and selected items from the second. Do not automatically occlude every annotation. If twenty labels belong to one plate, begin with the five to eight highest-value structures and add more only after the existing workload remains sustainable.

Track the full daily workload: due reviews, new cards, lab preparation, image practice, lectures, sleep, and legitimate questions. There is no evidence-based universal number of anatomy cards per day. The sustainable number is the one whose later reviews still leave time for the learning tasks Anki cannot replace.

If one image produces several siblings, avoid reviewing them back-to-back when one answer reveals another. Check your burying settings and audit whether exposed labels or recently seen sibling answers are acting as prompts.

Step 7: review and grade strictly

Before revealing the answer, say the full structure name at the required specificity. If the course distinguishes “left recurrent laryngeal nerve” from “vagus nerve,” a broader family name is not correct. If the card asks for a relationship or function, naming the structure alone is not enough.

Use Again when you did not retrieve the required answer. Do not choose Hard simply to avoid acknowledging a miss. Conversely, do not fail yourself for synonyms your course accepts; edit the back to list approved alternatives.

Keep a short error code:

• ID: could not identify the structure;
• ORIENT: misread side, plane, level, or view;
• REL: knew the name but not the relationship;
• FUNC: knew the structure but not its role;
• CUE: answered from mask or label position;
• TRANSFER: knew the familiar image but missed a new specimen or scan.

The code tells you what to repair. More repetitions of the same atlas image will not reliably fix an orientation or transfer failure.

Step 8: repair anatomy leeches

Anki calls repeatedly forgotten cards leeches. A leech is not proof that you need more willpower. It may signal poor image quality, ambiguous masking, missing orientation, interference, weak understanding, or low-value detail.

For an anatomy leech:

1. Return to the specimen, atlas, or course source and locate the structure from stable landmarks.
2. Check whether the image and arrow identify one unambiguous target.
3. Add region, side, plane, or orientation to the Header.
4. Replace an exposed-label cue with an unlabeled or differently annotated image.
5. Compare the target directly with the structure you confuse it with.
6. Add one relationship or function that makes the identity meaningful.
7. Suspend the card if the detail is outside the objective or not worth future reviews.

Changing the learning problem is more useful than repeatedly pressing Again on an unclear rectangle.

Step 9: test transfer outside the original image

At least once per topic, close the familiar deck and retrieve the anatomy in another format:

• rotate or mirror a legally usable image when doing so remains anatomically valid;
• identify the structure from another atlas or an unlabeled diagram;
• locate it on a cadaver or prosection during authorized lab access;
• use a model with a different color scheme;
• move through serial cross-sections rather than one memorized slice;
• identify it on CT, MRI, ultrasound, or radiograph when required;
• explain its relationship to two neighboring structures;
• predict the deficit or procedural risk associated with damage;
• answer legitimate course or validated question-bank items.

Transfer failures are useful data. If you know an atlas plate but not a cadaver, spend more time finding landmarks in real tissue. If you miss axial CT orientation, practice level and orientation before adding isolated labels. If you know names but miss deficits, retrieve function and pathway.

Question banks and practical materials have different jobs from Anki. During clerkships, anatomy review should increasingly support imaging, procedures, examination, and current patient-care learning rather than preserve every preclinical label.

Where the evidence fits—and where it does not

Spaced retrieval is a reasonable component of anatomy learning, but avoid converting a broad learning principle into a claim that one Image Occlusion setting is proven best. A 2025 BMC Medical Education qualitative study described a small, single-school elective that combined longitudinal spaced retrieval with opportunities for testing and application.

That study can support the practical idea of pairing retrieval over time with opportunities to apply knowledge. It does not prove that Image Occlusion itself causes superior outcomes, establish one optimal card design, or quantify generalizable effects for all medical students. Your course assessment, performance on unfamiliar material, and workload remain important checks.

Where AnkiQuiz fits—and its important image limitation

AnkiQuiz is not an image-recognition system. The server converts images and other media in selected Anki fields into placeholders before sending card content to the question-generation model. The model does not see the anatomy image.

That means a pure image card—or an Image Occlusion card whose usable content is only masks, label coordinates, or media placeholders—is not suitable for generating a meaningful anatomy question. AnkiQuiz cannot infer which vessel is under a rectangle, inspect a CT slice, or judge whether you selected the correct structure.

It can provide a changed-prompt textual retrieval check only when the fields you select contain enough meaningful text that you are permitted to process. For example, a Header and Back Extra may state the region, view, structure name, key relationship, function, and clinical relevance. From that text, AnkiQuiz may ask a reworded relationship or function question.

A careful workflow is:

1. Select a small anatomy topic rather than an entire deck.
2. Inspect Header, Back Extra, and other selected fields.
3. Exclude image-only cards and fields containing only coordinates, tags, IDs, or source metadata.
4. Confirm that the remaining permitted text contains enough context for one unambiguous question.
5. Request a short quiz focused on relationships, functions, and distinctions present in that text.
6. Verify every generated answer against the card and an approved anatomy source.
7. Use misses to revise understanding or card text, not as a new scheduler.

The AnkiQuiz card-to-quiz tutorial explains field selection and quiz generation. AnkiQuiz does not process image identification, change Anki or FSRS scheduling, or replace an anatomy practical, cadaver lab, radiology practice, or validated question bank.

Do not submit patient information, protected question-bank material, proprietary images or text, or anything you do not have the right to process. Even when an image will become a placeholder, related text fields may still contain sensitive or protected information.

Questions medical students commonly ask

Is Anki good for anatomy?

Yes, Anki is useful for retaining selected anatomical identities, relationships, pathways, and functions over time. It works best after initial learning and alongside cadaver or prosection work, multiple viewpoints, cross-sectional imaging, practical practice, and legitimate questions. It is less effective when success depends on memorizing one mask position.

Should I use Hide All, Guess One?

Use it when hiding other labels reduces direct answer cues without removing the landmarks needed to interpret the image. Use Hide One, Guess One when visible context is educational and does not give away the target. Preview both on the actual image; neither mode is universally best.

Why can I identify a structure in Anki but not in lab?

Your card may repeat a clean image, fixed view, color, crop, label position, and neighboring cues. Lab requires orientation through variable tissue and unfamiliar views. Relearn the structure from stable landmarks, then practice on permitted cadaver or prosection material and different images rather than only repeating the original card.

Can AnkiQuiz quiz image-only anatomy cards?

Not meaningfully. The service turns images and media into placeholders, so the question-generation model cannot see the structure or masks. It can make a changed-prompt textual retrieval check only if selected Header, Back Extra, or other fields contain sufficient permitted text. It does not perform image recognition.

How many anatomy cards per day?

There is no universal target. Add only the number whose future reviews fit alongside lab, lectures, sleep, transfer practice, and legitimate questions. If due cards are rising or reviews displace those activities, pause or reduce new cards. Count the workload created over time, not just today's new cards.

Build anatomy that survives a new view

Good anatomy cards do more than hide labels. They preserve orientation, ask one gradable question, connect identity to relationships and function, and occupy a review load you can maintain. Their value is confirmed when the structure remains findable after the atlas, crop, plane, or prompt changes.

Start with one current region. Define five essential structures, create only the necessary occlusions, add view and relational context, and test those structures later on an unfamiliar permitted image or specimen. If the names survive but the relationships do not, repair that layer instead of generating another wall of masks.