It has been a year since we were exposed to a totally new disease. Now we are at the
start of the second wave that historically has been more severe than the first wave.
Are there any lessons to learn from the first wave that may help us reduce the impact
of the second wave?
As the first wave progressed, it was discovered that severe acute respiratory syndrome
coronavirus 2 (SARS-CoV-2) that causes coronavirus disease 2019 (COVID-19) pneumonia
was different from its predecessor SARS-CoV-1. SARS-CoV-1 mainly occurred in symptomatic
individuals. SARS-CoV-2 occurs in a large number of asymptomatic.[1] These asymptomatic go undetected and act as super spreaders. A study done very early
in the first wave on the Diamond Princess cruise ship revealed that 54% of asymptomatic
individuals with SARS-CoV-2 had positive computed tomography (CT) findings.[2] This finding has subsequently been supported by numerous international and more
importantly national researchers.[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10] Kashyape and Jain[8] did an interesting study of 1,499 patients and divided them into two groups, symptomatic
and asymptomatic. Interestingly the percentage of positive high-resolution computed
tomography (HRCT) was similar irrespective of their symptoms, debunking the theory
that asymptomatics have a very low percentage of positive imaging studies.
Another negative aspect discovered during the first wave was the high false negative
rate of reverse transcription polymerase chain reaction (RT-PCR) (30–40%).[11] This means that individuals who are false negative can be super spreaders. Again,
numerous international and national researchers have demonstrated HRCT positivity
in individuals with negative RT-PCR.[12]
[13]
[14]
[15]
[16]
[17]
[18] HRCT can help reduce false negative rate as well as curb super spreading of SARS-CoV-2.
It is important to note that CT may be negative in a positive RT-PCR and vice versa.
These tests are complementary and not competitive.
Artificial intelligence (AI) in medical imaging is a major buzzword as we enter the
third decade of the twenty-first century. COVID-19 was a perfect testing ground for
AI especially as it needed to concentrate on only two modalities, chest X-ray and
CT scan. The findings were also fairly typical and specific of ground glass densities/consolidations
in a subpleural/peribronchovascular location. Numerous articles have evaluated the
performance of AI systems especially in India where the cohort of patients has been
very high, all concluding the experienced radiologists performed better than the AI
system in the detection of COVID-19 pneumonia.[19]
[20]
[21]
Where AI is making a significant impact is on the quantification of extent of lung
involvement. In a pandemic, the main concern is as large volumes of patients present
at the same time, medical facilities will get overwhelmed triggering panic and societal
problems. Triage is needed to help determine rapidly where these patients will be
managed so as to optimally utilize medical facilities as well as not overwhelm the
medical facilities. It is also useful to determine when the patient needs to be admitted,
whether in a ward or directly to intensive care unit (ICU), what oxygen requirements
are envisaged, O2 by nasal canula, noninvasive ventilation, or invasive ventilation. Patients may also
rapidly progress and this needs to be predicted. The extent of lung involvement is
a good surrogate to detect disease burden; a low percentage of aerated lung correlates
with a poor prognosis.
Several CT scoring systems have been proposed, 20, 25, 40, 72 point scale as well
as percentage of lung involvement. All these evaluate the extent of lung involvement
depending upon the percentage of involvement of each lobe, based on the extent of
involvement these are converted to points that are summed up to provide a final score.
These scores are based on visual assumption; thus, this is a subjective method with
significant inter- and intraobserver variation, resulting in significant under- and
overestimation; there is no standardization. Additionally, this is a time-consuming
process that does not help in a pandemic when there are numerous patients; quick accurate
results are required to help triage these patients.
A recent study utilizing AI system (pulmo density package)[22] that provided percentage of lung involvement was utilized in a large cohort of patients
that demonstrated a significant correlation between extent of lung involvement and
hospitalization, oxygen requirement, ICU, or ward admission.
Very early in the first wave various radiological societies and health organizations
advised strongly against the utilization of CT in the detection of COVID-19 pneumonia
caused by SARS-CoV-2. As the first wave progressed, the realization of the great value
of CT in this pandemic unfolded, especially in countries like India where there is
democratization in the utilization of imaging equipment, not controlled by national
policies or insurance companies. In states like Maharashtra, the government realized
the value of CT scan and controlled the price of scans between Rs 2000 and 3000 for
HRCT. This made it easily accessible and affordable. The large cohort of patients
in India resulted in numerous excellent studies on the role of HRCT in asymptomatic
and negative RT-PCR. In fact, there has been no disease that has been imaged so much!
Understandably, there has been a call for a review of recommendations and guidelines
issued by national and international societies on the utilization of CT scan in COVID-19
pneumonia.
In conclusion, two important lessons learned from the first wave are:
-
CT is very useful in the detection of COVID-19 in asymptomatic and/or RT-PCR-negative
individuals. It has the added advantage of a very quick turnaround time as compared
with RT-PCR. CT can play a previously undiscovered role as a public health tool to
detect super spreaders early. The important point to remember CT may be negative in
a positive RT-PCR individual and vice versa, so they are complementary investigations
and not competitive.
-
CT is very useful in detecting the extent of lung involvement that helps triage/further
management prognostication of individuals. AI is playing a strong role to accurately
quantify the extent of disease removing subjectivity. Another important conclusion
of the first wave is that AI at present will not be able to replace the experienced
radiologists but will function as a useful ally.
