Topic Discussion: Artificial Intelligence in Nephrology

Artificial intelligence(AI) is on a rise in science. Using it in medicine and specifically nephrology is sure to come.

According to the dictionary, AI is “the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.”

Dr Eric Topol has been a big proponent of this concept in medicine for years and recently has written a book called “Deep Medicine “ that details the potential uses of this in medicine.

Basically, AI can help in three main ways: 1) diagnosis that is often challenging in various challenging syndromes and even basic common ones. 2) make the physician’s life easier and decrease paper work and finally leading to the third -the most important 3) spending more time at the bedside.

AI is done via creating an  artificial neural network (ANN ) which is simply a collection of artificial neurons organized in layers. In a recent article in AJKD, authors discuss the potential use of this concept in Nephrology. They describe using it for IgA nephropathy(IgAN) as a recognizable cause for AKI. The ability to identify the patients that will progress to ESRD with IgAN would be useful for prognostic and therapeutic reasons. Geddes et al hypothesized that there exists a function that associates clinical and biological parameters measured at the time of IgAN diagnosis (namely age, sex, blood pressure, proteinuria, serum creatinine level, and antihypertensive treatments) to the probability of developing progressive IgAN. The authors designed and implemented an ANN to approximate this function. The results showed that their ANN could predict the occurrence of progressive IgAN more accurately than experienced nephrologists (correct predictions, 87% vs 69.4%; sensitivity, 86.4% vs 72%; and specificity, 87.5% vs 66%). Hmm, now this might be interesting to help guide a lot of therapies in Nephrology. This might be very useful in transplantation and prognosticating even need for dialysis for the elderly CKD patients.

Interestingly, many AI algorithms have been approved by FDA that are used in clinical practice:- some examples are of Atrial fibrillation detection, EF ECHO determination, Coronary calcium scoring, CT brain bleed diagnosis, device for paramedic stroke diagnosis, breast density via mammography to name a few.  No nephrology related such algorithms are approved to my knowledge.

There is an entire journal dedicated for this in medicine now

https://www.journals.elsevier.com/artificial-intelligence-in-medicine

Nephrologists, let’s get started and catch on!

Topic Discussion: Osmotic Nephrosis

Osmotic nephrosis describes a morphological pattern with vacuolization and swelling of the renal proximal tubular cells.

What does the pathology show:

Usually there is acute tubular necrosis–like changes. Histologically, osmotic nephrosis is characterized by a focal or, less often, diffuse “clear-cell” transformation of proximal tubular epithelial cells showing isometric fine vacuolization of the cytoplasm . The straight part of the proximal tubule primarily is involved and, in severe cases, also the convoluted part. Severely affected tubules are often seen side by side with normal-appearing tubules. Distal tubules and collecting ducts are more or less unchanged

Classic known causes of this entity are:

Intravenous immune globulin preparation(sucrose based)

Mannitol Dextrans
Contrast media
Hydroxyethyl starch
Glucose

How does one differentiate this from vacuolization seen with tacrolimus and cyclosporine? Is that a form of osmotic nephrosis?

Renal Pathologist Dr Lynn Cornell nicely describes this on twitter with these images. The image below shows isometric vacuolization in CNI toxicity. This leads to have focal tubules with this change( see arrow)

In osmotic nephrosis, tends to show vacuolated cytoplasm in tubules diffusely( see below)

Osmotic nephrosis describes a morphological pattern with vacuolization and swelling of the renal proximal tubular cells.

In addition, In paraffin sections, the isometric vacuolization seen in patients with calcineurin-inhibitor toxicity may be indistinguishable from osmotic nephrosis. However, electron microscopy shows dilated endoplasmatic reticulum as the cause of vacuolization in the former.  Osmotic nephrosis cannot be differentiated from lipid storage in tubular cells (foam cells), as seen in patients with nephrotic syndrome, liver failure, or intoxication. In such cases, foam cells also are often found in large amounts in the interstitial space. This does not occur in osmotic nephrosis.

The above image shows  osmotic nephrosis in a kidney biopsy specimen. (A, B) Tubular cross-section with seemingly no lumen. Epithelial cells are massively swollen, cytoplasm is completely filled by vacuoles of about the same size (isometric vacuoles), and nuclei are displaced to the base of the cells and distorted by adjacent vacuoles( source https://www.ajkd.org/article/S0272-6386(07)01592-2/pdf