Frequently asked questions

It’s an educational tool that uses a general-purpose AI model (Claude) to translate the medical writing around a brain-tumor diagnosis into plain language: radiologists’ written reports (single or compared across time), pathology reports, operative (surgery) reports, NGS / molecular profiling reports, and arbitrary medical text like doctor emails or research abstracts. Alongside the translation tools, the site also has practical organizers that don’t involve AI: a medication tracker, a symptom journal, a unified case timeline, a pre-appointment prep packet, a second-opinion packet generator, a curated treatment-options landscape, and patient-facing primers on driving status and work / disability. The goal is to help you arrive at your doctor’s appointment understanding the terms and ready with better questions, not to diagnose you.

No. The scan-report flows translate the written report your radiologist produced, not the underlying MRI/CT images themselves. Earlier versions of this site did interpret images directly; we retired that because LLM-based image interpretation can sometimes produce confident false positives. The pathology decoder does accept screenshots or photos of paper pathology reports (as a convenience for getting the text in), but the model only reads the text on those reports.

No. This is an orientation and education tool. The AI can miss things in a report, invent details that aren’t there, and misjudge what matters. Every result includes a reminder to share with your clinician. Nothing on the site replaces a qualified neuro-radiologist, neuro-pathologist, neurologist, neurosurgeon, neuro-oncologist, or pharmacist.

Patients, family members, and caregivers who have just received a scan or a pathology report and feel lost in the terminology. It also works well between appointments: for tracking medications, comparing follow-up scans, and building a case timeline so you walk into each visit oriented.

Yes. There is no account, no sign-up, no paywall, and no ads. If that changes we’ll say so clearly on this page before turning anything on.

When you submit a report (as a PDF or pasted text), the text is transmitted to Anthropic (the model provider) to be translated. The response comes back and renders in your browser. The original report content is not retained on our servers. There is no database of uploads. Only a short (~240-character) preview is saved in your browser’s local storage so the history list can show context. See the Privacy Policy for the full flow and the legal basis.

Yes, by default. Before any report text leaves your device, your browser automatically removes patient name, MRN, date of birth, address, ZIP, phone, email, SSN, and accession/account/specimen numbers. Each detected item is replaced with a categorical placeholder like [PATIENT NAME] or [MRN]. Age, sex, scan/exam dates, and provider and facility names are kept; removing them would degrade the AI’s translation without privacy benefit. You can see the list of categories that were removed in a chip-list below the textarea, and you can manually edit the cleaned text before clicking Analyze. This is a best-effort precaution rather than HIPAA-grade de-identification: rare names or unusual report formats may not be caught by the patterns, so reviewing the cleaned text is encouraged.

Past analyses (scan-report explanations, scan-report comparisons, pathology decodes, surgery-report translations, NGS / molecular decodes), your medication tracker, your symptom journal, your appointment-prep details (next visit, custom and dismissed questions), your beyond-treatment view (care-phase choice, surveillance rhythm, and late-effects log), your second-opinion-packet preferences (cover sheet, which sections are included), your driving and work preferences (state, employer size, tenure, latest seizure date), and the case timeline all live in your browser’s localStorage, never on our servers. Clearing your browser storage, using the “Clear all” buttons on individual pages, or the Clear all data control in Cookie Settings wipes everything.

No. Anthropic’s standard API is not used for training by default, and we don’t opt into any training programs on your behalf.

No, we don’t run any analytics. No page-view tracking, no usage stats, no behavioral data, no advertising pixels. Essential functional storage (your past analyses, medications, journal entries, preferences) lives in your browser’s local storage only, because the app can’t work without it. Error tracking is on the roadmap as an opt-in feature for catching crashes, but isn’t running today. Manage your choices on the Cookie settings page.

Your data lives in your browser’s local storage, so you control it directly. Delete individual entries, use the Clear all buttons on the History, Medications, or Case timeline pages, or use Clear all data in Cookie Settings to wipe everything in one click. We don’t store your data on our servers. There’s nothing for us to export or delete on the backend. For privacy questions, email info@gliowise.ai and we’ll respond within 30 days.

GlioWise is a personal educational tool for individual use. It is not a covered entity under HIPAA (Health Insurance Portability and Accountability Act) and does not function as an electronic health record system.

Anything you enter in GlioWise is for your personal use only. It lives in your browser’s local storage and is not part of any official medical record system. We don’t store your data on our servers.

For full details on how data flows, see our Privacy Policy.

For scan-report explanations and compare-reports: the radiologist’s written report, as a PDF or as pasted text. The scan-report flows do not accept images of MRI/CT slices, only the written report. For pathology decodes, surgery (operative) report translations, and NGS / molecular profile decodes: PDF, an image/screenshot of a paper report, or pasted text. For the plain-language translator at /explain: pasted text only. The organizer pages (second-opinion packet, treatment options, driving, work) don’t ask you to submit anything. They work off what’s already saved in your browser from the other tools, plus a few small preferences.

Explain a scan report takes one radiology report and translates it into plain English: a lifted impression, a measurements table, definitions of the medical terms, an incidental-findings callout for items unrelated to your tumor, and report-specific questions to bring to your oncology team. Compare reports over time takes two reports (typically your last one and the most recent), aligns them, and surfaces what changed: the radiologist’s framing shift, before/after measurements, new findings, and stable items collapsed out of the way. You can also pivot into compare mode from a single-report result via the “Compare with a prior report” button.

Explain a scan report translates a radiologist’s MRI/CT report, what the imaging showed. Understand pathology translates a pathologist’s biopsy/resection report: tumor type, grade, histology, and every molecular marker (IDH, MGMT, 1p/19q, ATRX, Ki-67, p53, etc.). Different document, different decoder. Most patients end up using both at different points.

Earlier versions did. We retired image interpretation after a patient advocate flagged that LLM-based MRI reading can sometimes produce confident false positives. The radiologist’s written report is what your oncology team will act on anyway, so we focus on translating that into plain English.

Submit the full report: the FINDINGS section plus the IMPRESSION section, not just one or the other. The impression tells the AI what the radiologist concluded; the findings let it lift specific terms and measurements. If you’re using compare mode, the two reports should be for the same patient and ideally from the same imaging facility so the framing is comparable. If a report is a poor-quality scan or a photo of a printed page, type out the impression rather than relying on OCR.

Yes. Every scan-report explanation, comparison, and pathology decode is saved to your Past analyses page (they all live in your browser’s localStorage, not on our servers). You can also print or save a summary as a PDF from any analysis page (email is on the roadmap, not in this release).

The model is a general-purpose AI language model, not a trained neuro-radiologist, neuro-pathologist, neuro-oncologist, or pharmacist. It will be confidently wrong sometimes: misreading a term, paraphrasing softer or sharper than the radiologist intended, missing a measurement, or quietly dropping a caveat that was in the source. Treat each result as a starting point for a conversation with your care team, not a verdict.

The scan-report flows are intentionally translation-only. They preserve the radiologist’s exact clinical framing in quotes (so “concerning for tumor progression” stays exactly as written, not softened to “a possible change” or sharpened to “tumor is growing”). They show measurements verbatim from the report with a direction tag, but never compute growth percentages or label severity. Findings the radiologist explicitly called incidental (like mild sinus disease) are surfaced separately so you don’t read them as tumor-related. And the model is instructed never to estimate prognosis, recommend treatments, or interpret beyond what’s in the text.

Yes, it can, and it will, sometimes. That’s a known failure mode of AI. The plain-English impression is generated text, and the model can invent plausible-sounding measurements or definitions that aren’t in the source report. Always cross-check with the original radiologist’s report in your patient portal, and with your oncology team.

On a pathology decode, a chat panel lets you ask follow-up questions in context. It already knows what the decode said, so you can ask “what does my MGMT methylation mean alongside my IDH status?” without re-explaining. Questions and answers are stored with that specific analysis in your browser. You can star any question; starred questions automatically flow into your next pre-appointment prep packet so you can carry them into the visit. A scoped Q&A panel for the scan-report explainer and comparator is on the roadmap (the existing panel was built around the retired image-analysis kinds and needs its own boundary prompts (severity, prognosis, and treatment recommendations are intentionally off-limits on that surface) before it can responsibly handle scan-report content).

A general-purpose explainer at /explain for medical text outside of radiology and pathology reports. Paste any medical text: an email from your care team, a forum post, a research-paper abstract, a chunk of a discharge summary. The tool returns a plain-language explanation plus a decoded glossary of the key terms it found. The pasted text is sent to the AI model the same way an analysis is (transmitted to Anthropic and the response renders in your browser), but the translator output is not saved to your past-analyses history; each translation is a fresh one-off. Treat the result as orientation, not as advice or as a substitute for asking your care team directly. If your text is a full radiology report, use /explain-scan instead. You’ll get a richer structured output (impression, measurements, key terms, incidental findings, questions) and the entry will save to your history.

The surgery (operative) report decoder at /explain-surgery translates an operative note into plain English. It surfaces the extent of resection in the surgeon’s own words (gross total vs subtotal, with the verbatim quote), the techniques used (awake mapping, 5-ALA fluorescence, intraoperative MRI, neuronavigation, ultrasonic aspirator), the specimens sent to pathology, the complications or surprises noted, and a set of questions for your neurosurgical team. Accepts PDF, screenshots/photos, or pasted text. The result saves to your Past analyses for re-reading later.

The NGS (next-generation sequencing) decoder at /understand-ngs translates a molecular profiling report (Foundation One CDx, Tempus xT, Caris, MSK-IMPACT, and others) into plain language. It buckets variants by clinical relevance (driver vs passenger vs VUS), spells out what each driver mutation means for treatment, flags therapy implications with FDA / NCCN / trial framing, and optionally fetches live-matched clinical trials from the NCI’s open trial database based on the variants in your report. Accepts PDF, screenshots/photos, or pasted text. The decoded report saves to your Past analyses. The trial-match fetch is opt-in. Nothing about your variants leaves your browser until you tap the button.

The page at /second-opinion assembles a clean, printable summary of your case from what GlioWise already has on file (scans, pathology, NGS, surgery, current treatment, recent symptoms) for a second-opinion visit. You fill in a small cover sheet (name, DOB, the question you want the second opinion on), tick which sections to include, and a live preview renders on the right. Print as PDF uses your browser’s native print dialog; nothing is uploaded. Starred questions from your past analyses can pre-seed the cover-sheet questions field. Everything runs in your browser; no AI is called for the packet itself.

The page at /treatment-options is a patient-facing map of how diffuse gliomas are typically treated (standard of care, NCCN-listed alternatives, and the kinds of investigational directions active right now), plus a list of NCI-designated comprehensive cancer centers with neuro-oncology programs. It covers six diffuse-glioma profiles in v1 (GBM IDH-wildtype, IDH-mutant astrocytoma grades 2/3/4, oligodendroglioma grades 2/3). The page auto-detects the matching profile from your latest pathology + NGS records on file; if you don’t have records yet, a manual picker is available. The content is curated by hand from current NCCN guidelines and FDA approvals (not generated by AI), so every claim has a traceable source and the same profile produces the same view every time. It is deliberately educational: no doses, no rankings, no personalization to your specific case. Your team does the personalizing.

The page at /driving is a patient-facing primer on the driving question that comes up for almost everyone with a brain tumor. It pulls your latest seizure date from the symptom journal (if you’ve been logging) so you can see your days-since-last-seizure count, surfaces your state’s rules on physician reporting of seizures (a small number of US states require doctors to report patients to the DMV), and gives a set of talking points to bring to your neurologist about driving. It never tells you you can drive. That call is your medical team’s and your state DMV’s. The page exists to organize the question, not answer it.

The page at /work organizes the work side of a brain-tumor diagnosis: federal FMLA eligibility (based on employer size, tenure, and hours), state short-term-disability (highlighting the few states that mandate STD), the ADA accommodations patients most often request, and questions to bring to your care team and to HR. It also points to askjan.org, a free federally-funded service for accommodation guidance. All preferences (state, employer size, tenure) live in your browser only. This is general information, not legal advice . For a contested case, see an employment lawyer.

The page at /trials has two parts: a short primer explaining how brain-tumor trials work (phases, what randomization means, what eligibility criteria typically include) and a guided form where you fill in what you know about your case (tumor type, IDH/MGMT status, prior treatments, geographic radius) and the form builds the right clinicaltrials.gov URL for you to open. It is not a trial-matching service, doesn’t check eligibility for you, and doesn’t apply on your behalf. The form runs entirely in your browser; nothing about your filter selections is sent to our servers. Clicking the search button opens clinicaltrials.gov in a new tab where the usual filtered listing appears.

The day after an appointment you’ve set on /prep, the countdown widget on the home page transforms into a “How did the visit go?” prompt for up to a week. Tapping it opens a short reflection form: what the doctor said, what changed, what’s next, and anything you wish you’d asked. Saving captures the debrief to your browser, closes out the past appointment, and any “wish I’d asked” lines automatically appear as pre-starred questions on the next prep packet so you don’t have to remember to add them. A recap of your most recent debrief shows at the top of /prep when you set the next appointment, so context carries forward. The entire flow is browser-only. Nothing about the debrief is sent to our servers or to Anthropic.

A curated set of common brain-tumor medications (temozolomide, bevacizumab, vorasidenib, steroids, seizure meds, etc.) with plain-English side-effect and call-the-doctor cards. You can also add custom entries for meds outside the curated list.

When you add two or more medications from our curated list, the Medications page shows a small panel that flags well-documented pairings: for example, phenytoin with the newer IDH inhibitors, or valproate’s effect on lamotrigine levels. Each flag comes with a plain-English explanation, what to watch for, and what a care team typically does about it. This is not a clinical interaction checker. The list is intentionally narrow, it only covers drugs in our curated database, and it cannot see over-the-counter medicines, supplements, herbal products, foods, or your full medication history. The absence of a warning is not evidence that a combination is safe. Your pharmacist has the authoritative interaction tools. Please check any new combination with them and with your neuro-oncology team.

The check runs entirely inside your browser against a small, hard-coded list that ships with the site. Nothing about your medication list, and nothing about the interactions it flags, is sent to our servers, to Anthropic, or to any third party.

The symptom journal at /symptoms is a simple diary for the days between appointments. Each day you can tap an overall mood (Good / Okay / Rough / Hard), select which symptoms you’re feeling (headache, fatigue, nausea, word-finding, sleep, mood, and so on) with a 0–10 severity, and log seizures with as much or as little detail as you want (type, duration, time of day, aura, after-effects, possible trigger). You get a 14-day at-a-glance strip, an expanded list of recent entries, a 30-day trends chart, and a small Patterns we noticed panel above the chart with rule-based observations. Things like “5 seizures on 4 days in the last 30, worth flagging at your next visit,” “headaches have been worse on Mondays for the last three weeks,” or “headache severity has eased this week.” Each pattern has an Add to next visit button that pre-stars the related question on your next pre-appointment prep packet, and an × to dismiss patterns you’ve already acted on (you can restore them later). An Export last 30 days button produces a one-pager you can print or save as a PDF to bring to a clinic visit. Nothing is analyzed by the AI here and nothing is sent to our servers. All entries, trends, patterns, and the export are computed inside your browser.

No. It’s a self-reported diary to help you notice patterns and share them with your care team. Not a medical device and not a substitute for clinical follow-up. The journal doesn’t detect seizures automatically, doesn’t call for help, and doesn’t replace an epilepsy diary your neurologist may have given you. The red-flag reminders on the page (call your care team if a seizure lasts more than 5 minutes, or if new weakness or sudden severe headache appear) are general educational prompts; always follow your own care team’s instructions over anything the app says.

One horizontal strip with six lanes (scans, procedures, pathology, radiation/chemo treatments, medications, and symptoms) automatically assembled from everything you’ve already entered across the site. It gives you an at-a-glance view of your treatment story.

Not yet. Sending a summary by email is on the roadmap for an upcoming release. When it ships, you’ll be able to review the content first, type the recipient yourself, and we won’t retain the email body after sending. For now, the most reliable way to get a summary out of GlioWise is your browser’s “Print” or “Save as PDF” from any analysis page or the pre-appointment prep page.

The prep page at /prep assembles a one-page packet from data you’ve already entered elsewhere on the site: past analyses, the medication tracker, the symptom journal, and the drug-interaction check. It produces a list of suggested questions to bring to your visit, a snapshot of recent activity (latest imaging trend, active medications, top symptoms, recent seizures), and a small form for your next visit’s date, provider, and visit kind. A Save as PDF button uses your browser’s native print feature to render a single, printable page. Nothing is sent anywhere.

No AI is called. Question generation is rule-based and runs entirely in your browser. A small set of deterministic rules looks at what you’ve already entered (for example, a recent imaging trend, an active medication that’s commonly paired with a side-effect concern, or a high-frequency symptom on the journal) and attaches a relevant generic prompt. Treat the suggestions as starting points to discuss, not as a checklist to enforce. Your care team will know which ones actually matter for you.

Yes. The prep page has an input to add custom questions, which sit at the top of the list with the rest, and an × on each generated question to dismiss it. Dismissed questions are remembered per-question. If the same rule fires again on a future visit you won’t see it, but you can use Restore N dismissed to bring them all back. Custom and dismissed questions live in your browser only, under the bsc:appointment:v1 key.

The page at /beyond-treatment is a long-term organizer for the chapter after active treatment ends, when the rhythm of care shifts to periodic surveillance scans, watching for late effects, and managing energy. It has three pieces: a surveillance-rhythm card with your next-MRI countdown and a scanxiety check-in inside the two weeks before a scan; a late-effects monitor where you log observations (cognition, fatigue, hormones, hearing, etc.) with a 0–10 severity and optional notes; and a printable care-plan summary that includes the items you’ve flagged for your next visit. Most useful in the surveillance phase. The view is not a clinical surveillance schedule. Always follow what your neuro-oncology team has planned.

The page is built for the chapter after active treatment, so we keep the setup small: one button to Set up my surveillance view, and a smaller Not yet, tuck this away for now link if you’d rather come back later. We deliberately don’t guess your care phase from your timeline data. Patients move through these phases in their own way, and the rule “you finished radiation 9 months ago therefore you’re in surveillance” doesn’t always hold (some are on long-term maintenance, some are in re-treatment). Whatever you choose is stored in your browser only, under the bsc:carePhase:v1 key, and you can change it any time using the Currently: … · Change link in the page header.

It’s a small running log of long-term issues you’re keeping an eye on. Categories include cognition, fatigue, mood, sleep, hormones (especially after radiation), hearing/vision, weakness/balance, headache, and seizure activity. For each entry you pick a category, slide a severity 0–10, jot an optional note, and optionally Flag for next visit. Flagged items show up front-and-centre on the printable care-plan summary. None of this is medical advice or a clinical assessment. It’s a patient-prepared organizer to bring to your team. Long-term effects can show up months or years after treatment, and even “small” observations are worth flagging.

The same way the symptom-journal export and pre-appointment prep packets are generated: with your browser’s built-in Print dialog. The page hides its on-screen chrome under a print stylesheet and renders a tightened, doctor-ready summary instead. Choose “Save as PDF” as the destination from your print dialog; nothing is uploaded to us. The PDF includes your surveillance plan, your flagged late-effects, and a few standing reminders.

Current versions of Chrome, Safari, Firefox, and Edge. The interface is designed primarily for desktop and tablet . Mobile browsers (Mobile Safari and Chrome on Android) will work, but several views like the case timeline, 30-day symptom trends, and the printable care plan are easier to read on a bigger screen. If you’re on a very old browser, localStorage or PDF preview may misbehave.

Retry once. Transient model errors happen. If a report is unusually long (close to the 20,000-character cap), try submitting just the FINDINGS and IMPRESSION sections rather than the whole document. For a pathology PDF that won’t parse, try pasting the text directly. If the error persists or mentions “invalid JSON,” email info@gliowise.ai with a screenshot. We read every report.

History lives in your browser’s localStorage. Clearing browser data, switching to a different browser or device, or using private/incognito mode will make past analyses look gone. They’re not recoverable because we never had them on our end.

Email info@gliowise.ai. Include a screenshot if you can, and the browser + OS you’re on.