MOBILENET ARCHITECTURE

Let F be the input feature map, G be the output feature map, and К be the convolutional kernel. G is given by

G is the output feature map for standard convolution while assuming stride one and padding.

Depthwise separable convolutions comprise two layers. Depthwise convolutions are used to apply a single filter per each input channel.

We can define depthwise convolution with one filter per input channel as

where F is the input feature map, G is the output feature map, and К is the convolutional kernel of size Dk>Kx M where the wth filter in convolutional kernel К is applied to the with channel in input feature map F to produce the with channel of the filtered output feature map G.

The computational cost of a depthwise separable convolution is

where a is the width multiplier and ae [0,1]. The width multiplier reduces computational cost and the number of parameters quadratically by around a2 DK x DK is the kernel size, and Df x DF is the feature map size.

PROBLEM IDENTIFIED IN SKIN DISEASE DETECTION

This chapter understands problem definition, observing the possible solutions to the problem and formulating the solution to it. A research problem is a statement that specifies the area of concents, a difficulty to be eliminated, or trouble with existing theory, or concept, a condition to be improved upon to a need for necessary investigation.

5.3.1 UNDERSTANDING THE PROBLEM

Infants are prone to skin diseases more often than adults because their skin is delicate, and their tissues can be easily ruptured. If one encounters with a skin disorder, immediate treatment is essential with proper diagnosis and medication, without which infant can die at early stages of skin disease due to a lower immune system. Remote areas where the treatment centers are not nearby to household regions hence become challenging to diagnose cases of skin disease and may lead to death due to lack of adequate treatment and faster transport. Even in the rural areas, although health centers are nearby, due to unawareness of the skin diseases, it is quite challenging to provide medication for it.

  • 5.3.2 OBJECTIVES OF THE PROPOSED SYSTEM
  • • To develop a system that provides immediate or faster than existing medication.
  • • To build a platform that serves the patient and the doctor in then terms.
  • • To deploy a system with a user-friendly interface for ease of usage.
  • • To analyze the effect of a disease over a region of the map and visualize it for better actions to be taken.
  • • To provide the location of nearby hospitals, pharmacies for medicines.
  • • To establish a connection between a doctor and a patient for useful, healthy conversation without any leak of doctor’s contact information.
  • • To make aware of the common solutions or home-made remedies for some trivial diseases at initial stages.
  • • To improve the experience level of beginner-level medical graduates.
  • 5.3.3 FORMULATING THE SOLUTION

The solution for the problem mentioned is to build a platform for both doctors and patients, where patients can register themselves for getting a diagnosis from a doctor. It solves the problem of transportation and makes the immediate connection between the doctor and the patient. With new patients, the app is tracking diseases for further analysis by location. Later on, it can be visualized to cure the diseases of patients within the same local geographical regions. Vaccines can be provided based on geographic areas and can prevent a significant loss of humanity.

With the access of patient’s location, we can provide them with their nearby location of hospitals and pharmaceutical stores, according to disease identified from image. In case the patient has additional queries, the system should provide the chat facility but ensuring the privacy of both a patient and a doctor. The system has an encrypted way to store details of patients and doctors. The patients do not need to discuss a small health issue with a doctor every time, and hence, the system itself can provide the solution in such cases. The solution is provided based on the classification of disease by the machine learning model.

This system benefits the patients to cure their diseases and also provide platform to medical graduates to provide medication to patients. Not only freshers but also higher level medical students can do practice in the system. It provides a summarized description of the spreading of certain diseases and their root cause of it from analysis of patient’s locations. In addition, the system can recommend effective drugs immediately, which cures most of the patients. ApnaDermato consists of a centralized database for doctors and patients, which stores the entire information about patients and doctors, along with their queries. Few intelligent questions are asked to the patients to know how the condition of the patient is and accordingly gets treated by doctors. If there is no critical case of skin disorder, then home remedies are suggested to heal naturally. Doctors can register this system manually, which prevents the legitimate issues of medical science. With this feature, a patient can save a lot of time waiting to get the doctor’s appointment and hence can save millions of lives.

COMPARATIVE STUDY WITH EXISTING SYSTEMS

This chapter gives a brief comparison between the proposed method and the existing systems. Existing systems along with their pros and cons are listed in the following.

5.4.1 FIRST DERM: ONLINE DERMATOLOGY

This system bridges the gap between an Internet search and an in-person dermatologist consultant.

Advantages:

  • • User-friendly graphical user interface.
  • • Able to get nearby clinics and pharmacies.

Disadvantages:

  • • Paid and nonrefundable.
  • • Provides only suggestions about the disease.
  • • Does not cine the diseases at first place.

5.4.2 PRACTO: DOCTORS, CONSULT ONLINE, ORDER MEDICINES

Practo is a folly featured app from doctors and consultants to deliver medicines; all functions are integrated into this system.

Advantages:

  • • Provides plenty of sendees.
  • • Provides chat with the doctor.

Disadvantages:

  • • Response time is huge.
  • • Rude doctors on chat.
  • • The support team is not professional.
  • 5.4.3 mfine: CONSULT TOP DOCTORS FROM BEST HOSPITALS

The system provides seamless interaction between patients and doctors on chat and video call

Advantages:

• First-time free consultancy.

Disadvantages:

  • • Provides unmanaged reports.
  • • Multiple reschedule the appointments.
  • • Poor customer sendee.
  • 5.4.4 DOCTOR ON DEMAND

The system provides sendees in 50 states and the District of Columbia, and it ensures 24 h of on-time sendee.

Advantages:

• User-friendly interface.

Disadvantages:

  • • Cannot detect a disease from an image.
  • • Nonprofessional advice from doctors on this platform.
  • • Poor response time (Table 5.1).

Proposed System (ApnaDermato)

Existing System(s)

Emergency Cases

The app takes as input various medical conditions which patient is suffering through and based on those conditions as well as based on prediction our application had made it decides the case as an emergency case or a normal case

Existing applications on the play store does not support the consideration of emergency cases [1.4.1, 1.4.2,1.4.3]J

Morals And Ethics

On the stall of the app. it shows that these are the morals and ethics which should be kept in mind before doing any steps which are supposed to be forbidden

Instructions for using the app are not mentioned in the existing systems [1.4.1, 1.4.3]

Personal Doctor Interaction

When identified as an emergency case, that case is forwarded to the doctor. Not only that when once a case is forwarded to the doctor, the doctor can accept that query and reply to it. else if the doctor rejects then that case is forwarded to another doctor, and even if the doctor accepts a case and then do not reply within a period of time, then that case is automatically forwarded to another doctor

Existing applications support the appointments fixing of the patients with the doctors, but it nowhere supports the online interaction with doctor personally

Nearby Hospitals, Pharmacies, Burn centers

When a medication is provided to the patient either by the doctor or some normal medicines by the application itself, our application suggests the location of nearby hospitals, pharmacies, bum centers which is very helpful for people living in villages or places (remote areas) where we cannot find the hospitals, pharmacies so easily

Existing applications show their own respective nearby clinic centers but do not show nearby hospitals, pharmacies, or bum centers [1.4.4]

Statistics

The application shows various statistics such as no of patients cured by that doctor or no of patients in a particular area cured by a particular medicine by doing analysis

No such statistics or analysis is provided in the existing applications [1.4.1,1.4.2, 1.4.3]

'Square brackets indicate figure numbers of the application user interface.

 
Source
< Prev   CONTENTS   Source   Next >