Diagnostics & medtech
Thirty-six subjects on how a measurement earns the right to be believed — from transduced biosensors to organs on chips — and the recurring discovery that the binding constraint is never the sensor alone but the biology that stands between it and the answer.
Explanatory reference: why a technology works and where the science is unsettled. Drafted with AI assistance; pages carry no per-claim references and have not yet been reviewed by a subject expert — use them for orientation, not as a citable source. The evidence for each technology, with graded sources, is in its dossier. How this reference is made →
Read the thirty-odd subjects of this cluster together — biosensors, sequencing, imaging, organ-on-chip, the consumer formats — and three arguments keep reappearing. Hold them before starting anywhere.
Every method lives or dies by one bounding quantity. Not by its feature list: by the affinity window of its binding chemistry, the sequencing-error rate standing against the variant frequency it claims to call, the Poisson floor under single-molecule counting, the stability window between collection and analysis, the prevalence term inside every screening calculation. The pages here name that quantity for each subject, because it is where the method’s honest limits and its engineering frontier both live — diagnostics biosensors sets out the transduction vocabulary the others reuse.
Measurement is a negotiation with the body’s own noise. The interesting designs do not silence the noise; they route around it. Synthetic reporters turn pathology itself into the signal. Liquid biopsy reads tumour fragments the immune system has already filtered. Wearables trade absolute accuracy for continuous context and recover decisions from the trend. Organoids and chips replace the noise of a whole animal with controlled, legible physics. In each case the design question is not “how sensitive” but “what does the body do to the measurement” — and the answer is where device coatings, biobanks and specimen handling earn their keep.
The field’s direction is temporal. Diagnosis began as an answer after the fact — pathology confirmed, resistance measured, disease staged. Its newest chapters move the same measurements earlier: antibiotic susceptibility in hours instead of days, minimal residual disease instead of relapse, a genomic surveillance signal instead of an outbreak. What has changed is not the chemistry but the temporal structure of the decision, from answer-after-the-fact to answer-in-time-to-use.
Start with the rank-one page if you want the measurement spine, or anywhere else — each page links to whichever of these three arguments it demonstrates.
Articles 36
| How biosensors measure Every diagnostic is a wager between two molecules — one that recognises, one that counts. What limits the result is rarely the sensor's cleverness; it is molecular affinity, background noise, and how few target molecules the sample happens to contain. |
| Molecular diagnostics at the point of care A PCR machine is a heater pretending to be a laboratory; moving molecular testing to the bedside meant solving two problems — doing amplification without a thermocycler, and smuggling years of technician skill into a disposable plastic cartridge. |
| CRISPR diagnostics The same molecular scissors that edit genomes can be borrowed as a switch: point them at a pathogen's DNA and their appetite turns indiscriminate, shredding whatever synthetic reporters float nearby. Presence becomes programme. |
| Liquid biopsy and circulating tumour DNA A tumour leaks fragments of its genome into the bloodstream as it dies cell by cell. Reading that exhaust from a blood draw sounds impossible until you see the numbers: the task is finding roughly one forged sentence in a book of ten thousand genuine ones. |
| Multi-cancer early detection Screening healthy people means asking a question whose answer is almost always no — which makes threshold mathematics, not chemistry, the binding constraint on every blood test promising dozens of cancers from one draw. |
| Direct-to-consumer genetic testing A mailed tube of saliva becomes either a fixed six-hundred-thousand-question quiz or a thirty-times-proofread copy of your whole genome. What separates them matters more than the price tag, because nearly all common-disease genetics lives in fractions. |
| Companion diagnostics: tissue assays for drug eligibility A targeted therapy without its companion test can harm patients it cannot help — so regulators bolt the two together at approval. The interesting science lives in how continuous tumour biology gets forced through yes-or-no gates. |
| Microbiome diagnostic panels A stool sample reports on an ecosystem, not an organ — and ecosystems answer questions nobody asked. The measurement science is mature; the hard part is that its raw outputs are proportions, and proportions systematically mislead. |
| Epigenetic clocks Certain chemical marks on DNA drift with age along paths reproducible enough that regression turns them into a year count. The fascinating subtlety: what such a number means depends entirely on what its algorithm was trained to imitate. |
| Multi-omics profiling in precision oncology One tumour, three witnesses: the genome testifies to possibilities, the transcriptome to current activity, the proteome to what physically stands on deck. Integration sounds like arithmetic and is actually a predictive model wearing lab credentials. |
| Genomic pathogen surveillance Case counts say how many; genome sequences say which lineage, arriving from where, and carrying what. Turning that into public-health action is an exercise in statistics, logistics and humility about missing denominators. |
| AI interpretation of tissue slides A digitised biopsy is a hundred-thousand-pixel square that no network can read whole. Making machines useful in pathology required admitting the label problem: nobody has ever annotated every tile — only verdicts about entire slides exist. |
| Cell-free DNA: from blood draw to error-corrected call Cell-free DNA exists because cells die; sequencing it is a chain of noise suppression, from stabilizing tubes to molecular barcodes, and each clinical market simply sets a different depth requirement. |
| AI-driven biomarker discovery: when the pattern finds the marker Machine learning over population-scale molecular data finds biomarkers no hypothesis would have named — methylation patterns that report a cancer signal and its tissue of origin, multimodal profiles that predict therapy response — and the field's real work is converting those patterns into locked, regulated, covered tests. |
| CRISPR lateral flow: the genome as a paper strip CRISPR diagnostics repurpose bacterial immune enzymes into molecular recognition and lateral flow strips into the readout — an architecture that needs no thermocycler and no instrument, which is why it reached emergency-use products faster than any platform before it. |
| AI reading of OCT eye scans Optical coherence tomography fires harmless infrared into the eye and times its echoes — a microscope that never touches tissue. Screening engines built on it taught medicine what an autonomous diagnostic must contain: a graceful way to say 'cannot answer.' |
| Retinal biomarkers for neurodegeneration The retina is brain tissue that migrated to the window of the eye. Reading neurodegeneration from it trades fidelity for reach: an imperfect proxy, but one a village clinic can afford and repeat yearly. |
| Wearable biodiagnostics A wearable is a bet on access routes: needle-thin filaments sample the skin's interstitial water, sweat glands surrender their salts, skin electrodes eavesdrop on cardiac voltage. Each route carries different chemistry at different honesty. |
| Implantable biosensors Moving a diagnostic under the skin removes every surface problem — adhesives, sweat, lost patches — and replaces them with one deeper adversary: the body itself, which walls off intruders and rethinks the wall for months. |
| Ingestible biosensors and smart pills A swallowed diagnostic borrows the body's own freight system: peristalsis drives it, stomach acid powers its wake-up call, and nothing needs retrieving. The costs return as physics — no steering, no biopsy, and an occasional pill that never leaves. |
| Breathomics Every exhale is the headspace above your bloodstream: metabolism's volatile exhaust diffusing out of lung capillaries at parts per trillion. Reading it requires instruments worth millions — and discipline about what those instruments might be recognising besides disease. |
| Breath-based metabolic monitoring Glucose itself never reaches the breath — it isn't volatile. What exhales instead is acetone, reporting fat metabolism; in ketoacidosis it shouts loud enough to smell. That gulf between proxy and target defines this whole category. |
| Tear-fluid diagnostics Tears are the only human fluid you can sample without breaking anything — three microlitres of chemically informative bath sitting on an immunologically busy surface. The catch is that everything interesting hides in tiny volumes that evaporate while you look. |
| Synthetic biomarkers Classic diagnostics eavesdrop on whispers the body releases; this strategy hands the body something loud to shout through its own pathology — an injected question no healthy tissue answers but diseased tissue cannot keep quiet about. |
| Home blood-analysis kits Home testing moves phlebotomy, not the laboratory. That sounds modest and is exactly why it works — and why the ghost of Theranos is best understood as physics: a test menu cannot be larger than the sample volume it is asked to serve. |
| Wastewater-based epidemiology A city reports on its own infections through the plumbing: every resident sheds daily, one sewer sample summarises a catchment, and the signal rises a week or two before hospitals feel anything. Public health's cheapest census runs on sewage. |
| One Health monitoring: zoonoses and resistance Antimicrobial resistance does not respect species boundaries, so monitoring any single species reads one act of a longer play. One Health surveillance samples animals, humans and environment together because resistance genes commute between them constantly. |
| Intraoperative tissue identification Surgery ends before pathology's verdict arrives, and the margin between them is where tumours return. A family of optical and molecular tools now answers 'what is this tissue' inside the operation — in minutes, over whole surfaces, without leaving the room. |
| Rapid antibiotic-susceptibility testing Every hour on the wrong antibiotic worsens outcomes and breeds resistance, yet the laboratory answer takes days because bacteria must be grown. New platforms attack the growth requirement itself — by watching single cells instead of populations, or by asking for the gene. |
| Organoids & organ-on-chip Why stem cells given the right niche signals build tissue autonomously, what flowing fluid adds that a dish cannot, and where microphysiological systems still fall short of the organs they stand in for. |
| Biocompatible coatings for medical devices Why a swollen hydrogel is the slipperiest thing you can legally put in a blood vessel, how diffusion through one polymer layer sets a drug's release from minutes to years, and what a delaminating coating costs. |
| Personalized diagnostics Why the same measurement means different things in different genomes — metabolism, immune recognition and tumour heterogeneity are the mechanisms that turn a population test into an individual answer. |
| Next-gen bio-forensics How forensic DNA moved from confirming an already-suspected identity to predicting one: population structure makes ancestry readable, heritability bounds what faces and colours can be guessed, and shared segments turn databases into family trees. |
| Biobanks & human population samples A biomarker measurement is only as good as the sample's biography: why decay begins at the needle, how freezing, fixing and drying each preserve some molecules at the expense of others, and why a bank's representativeness is a scientific instrument in itself. |
| Sleep & circadian bioeconomy What a ring or a mattress mat actually measures: light through perfused skin, stillness mistaken for sleep, and the two clocks — the suprachiasmatic nucleus and adenosine — whose phase a wearable can only estimate. |
| Sports bioeconomy What an athlete's biomarkers mechanistically report — a lactate crossover, a leaky muscle membrane, an acute-phase protein masquerading as iron stores — and which of the panels sold on recovery are reading physiology versus reading trend lines. |