1. AI and diagnostics

1. Voice biomarker "health check": A phone app that listens to 30 seconds of speech each week. It flags early changes linked to Parkinson's, depression, heart failure (breathlessness patterns) or cognitive decline. Novelty: it is passive and has no hardware. Difficulty: medium, because you need a large labelled voice dataset.
2. Cough-sound triage for TB and pneumonia: Smartphone cough analysis used by community health workers where X-ray and lab access is limited. Difficulty: medium.
3. "Second-opinion" AI that argues against the diagnosis: Instead of suggesting a diagnosis, it lists the most dangerous things the clinician might be missing, such as anchoring bias on a common diagnosis. Novelty: it targets cognitive error, not image reading.
4. Retinal photo as a whole-body scan: Use a cheap fundus camera to screen for diabetes, hypertension, kidney disease, and cardiovascular risk in one visit.
5. Digital twin for chronic disease: A personalised simulation of a patient's physiology (for example, a diabetic's glucose response) that tests insulin or diet changes virtually before they are tried in real life.

2. Wearables and devices

1. Sweat-sensing patch for medication levels: Continuous drug monitoring (antiepileptics, lithium, antibiotics) without blood draws.
2. Smart bandage: It senses wound pH, temperature, and moisture. It alerts the patient early to infection and releases antibiotic on demand. This would be especially useful in diabetic foot ulcers.
3. Smart inhaler with trigger mapping: It logs each use with location, pollen and pollution data. It then predicts asthma attacks and shows which triggers matter for that person.
4. Fall-prevention insole for older adults: It detects gait changes weeks before a fall and sends the data to a physiotherapist.
5. Ingestible capsule sensor: A swallowable capsule that samples gut microbiome and gas, and gives a non-endoscopic read of gut health.

3. Therapeutics and drug approaches

1. Personalised "n-of-1" drug combinations: Test a patient's own tumour or infection cells against drug panels on a chip, then choose the therapy.
2. Phage therapy banks: Libraries of bacteriophages matched to a patient's resistant infection, as a response to antibiotic resistance.
3. Digital therapeutics prescribed like drugs: Evidence-tested apps for insomnia (CBT-I), chronic pain, or substance use, with dosing and side-effect monitoring.
4. Microneedle vaccine patches: These need no cold chain and no trained injector, so they are suited to remote areas.
5. Chronotherapy scheduling tool: Software that times chemotherapy, antihypertensives, or steroids to the patient's own body clock, based on wearable data.

4. Surgery and procedures

1. Augmented-reality surgical overlay: It projects CT or MRI anatomy (vessels, tumour margins) onto the patient during surgery.
2. Haptic simulators with real patient anatomy: Surgeons rehearse a specific patient's operation in 3D-printed or virtual form beforehand.
3. Tele-mentored surgery in low-resource hospitals: A remote expert guides a local surgeon through a headset and shared view. It uses lower-cost, non-robotic gear.
4. Biodegradable implants that carry drug or growth factor: These dissolve after healing and avoid a second operation for removal.

5. Public health and low-resource settings

1. Solar-powered, offline AI ultrasound: A handheld ultrasound with built-in guidance that lets a midwife screen for high-risk pregnancy.
2. Mobile "clinic in a box" with a pharmacist-led model: It handles chronic disease (hypertension, diabetes) with refills and monitoring via SMS.
3. Wastewater surveillance dashboard: Early outbreak detection at neighbourhood level, covering viruses, antimicrobial resistance genes, and drug use.
4. Community "health credit" system: Local health workers earn credits for follow-ups completed, which fund local services.
5. Pill-pack reminders with low-tech tracking: Blister packs with simple printed circuits that log when a pill is taken and send it to a clinician. This targets TB and HIV adherence.

6. Medical education and the patient experience

1. Virtual patients that "get sicker" in real time: Students' decisions change the case, with AI-generated variation each time.
2. Procedure-practice with video AI feedback: Students record suturing or intubation and get scoring and corrections.
3. Plain-language discharge "translator": It rewrites discharge summaries at a reading level and language chosen by the patient, with a verification step by the nurse.
4. Patient-held "health passport" on a phone: A single record the patient owns and shares selectively with any clinic, useful for refugees and migrants.
5. Clinician burnout early-warning: It uses schedule, EHR time, and sleep data (with consent and privacy safeguards) to flag overload before it harms care.

7. Cross-disciplinary ideas

1. Smart pill organiser plus pharmacist AI: It checks drug interactions and flags duplicate prescriptions across prescribers for people on many medicines.
2. Gamified rehab with brain-computer interfaces: Stroke patients move a virtual limb by thought, which supports neuroplasticity. A 2026 review on AI and wearables in upper-limb neurorehabilitation covers this area.
3. Menstrual and hormone-cycle health tracker with clinical-grade sensors: Early flags for PCOS, endometriosis, and perimenopause.

How to pick one

• Real need: is there a clear clinical problem, and who would pay or benefit?
• Evidence: is there a data source to train or validate it?
• Regulation: does it need FDA or CE approval, and is the risk class low?
• Feasibility: can a small team build a first version in under a year?